Close

Register Now!

To Get More Info and Daily Reward.

Please login or register.
Pages: [1] 2  All   Go Down

Author Topic: P67 & RAM Performance -or- Why DDR3-1333 RAM is enough -> Save the money!  (Read 52028 times)

0 Members and 1 Guest are viewing this topic.

JackTopic starter

  • Ex Forum moderator
  • GENERAL OF THE ARMY
  • ******
  • Offline Offline
  • Posts: 14939
    • A user initiative by MSI fanatics to support all other MSI product users

Dear P67 owners and potential P67 buyers,

ever since I read an Intel "Sandy Bridge"/iX 2000 review on the German HardTecs4U [thanks to BikerKS/DH3AAK for bringing it to my attention] I wanted to start this thread to point out that spending a lot of extra bucks on so-called DDR3-1600, DDR3-1866, DDR3-2000 or DDR3-2133 memory modules instead of just sticking to DDR3-1333 (the defined native maximum according to Intel's CPU specifications) turns out to be a rather useless investment. System builders could really rather save the money here and/or use it to either buy more memory or an SSD, because these seem to be far better options to increase actual system performance.

The review I just mentioned included a thorough comparison of memory related benchmark results intended to illustrate the impact of higher memory speeds than DDR3-1333 on system performance.  The reviewers compared DDR3-1333 and DDR3-2133 results using the following benchmarks, tools, games and usage scenarios:

- PCMark [CPU]
- PCMark [Memory]
- Cinebench 11.5 [1 Core]
- Cinebench 11.5 [all Cores]
- ScienceMark
- 7 Zip
- WinRar
- WinZip
- Blender
- POV-Ray
- GIMP
- IrfanView
- Handbrake x264 [1920x1080]
- Mainconcept [H264 HQ]
- iTunes
- Nero AAC
- Lame
- OggEnc
- Anno 1404 [800x600 LQ]
- Colin McRae: Dirt 2 [800x600 LQ]
- Far Cry 2 [800x600 LQ]
- Mafia II [800x600 LQ]
- Anno 1404 [1280x1024 HQ]
- Colin McRae: Dirt 2 [1280x1024 HQ]
- Far Cry 2 [1280x1024 HQ]
- Mafia II [1280x1024 HQ]


You can find the the test results >here<. As you can see, the only really significant performance increase that was observed with DDR3-2133 over DDR3-1333 operation is a 11 second(!) bonus in 7ZIP performance and a 7 seconds(!) bonus in WINRAR performance.  

... and that is basically it.  There seems to be absolutely no significant performance bonus regarding all the other benchtests when DDR3-2133 is chosen over DDR3-1333. Now, if those 7-11 seconds in RARing and ZIPing stuff mean a fortune to you, then you'll definately need to go for DDR3-2133 RAM.  However, in case you believe that those 7-11 seconds are really not worth the extra money, then -- and this is what the evidence suggests -- then you are really better off with DDR3-1333.

-----------------------------

I have chatted with Bernhard and HU16E about those benchmark results a little bit and since they both call a P67 mainboard their own, they have gotten on to doing a bunch of tests of their own in order to check the above HardTecs4U results and were basically able to confirm the mentioned conclusion.  In the next post or two, you can see their test results.

-----------------------------

For the remainder of this thread, you guys are all invited to share your own observations and test results in case you have any to present to us.  But please don't reply to this thread when the only thing you have to say are things like "DDR3-1600/1866/2000/2133 it feels faster, though".  In case you decide to share your own results, please do so with the help of screenshots that backup whatever findings you wish to present. Thank you.
Logged

Bernhard

  • Support the MSI Red Rockets Team Members 나는 한국을 사랑
  • GENERAL
  • *****
  • Offline Offline
  • Posts: 6248
  • Read the user manual first. Properly

P67 chipset boards and the myth around performance memory

After Jack provided a link by a German Magazine that had conducted tests to verify the real performance gain on the P67 platform by using memory speeds higher than the native 1333 for the 1155 socket CPU,  I went back to previous posts from HU16E who had been comparing FPS tests independently.

After having seen a never ending stream of  users on this forum that have been convinced by media marketing and other sources that heatsinks on DRAM modules are cool and promise to be the ultimate in improving PC performance over and above a multiplier increase on the CPU, we needed to verify if this general perception contains any inkling of proof. The amount of posts are all evidence of never ending instabilities associated with these overrated, overvolted and overclocked profiles of these modules.

We decided to do a test battery of our own on our current P67 rigs to try and establish the real world benefit  of these Overclocked DRAM speeds on this platform. By real world we mean programs that would provide a visual performance increase. To be fair, we also included a theoretical/synthetic  benchmark tool for each step along the way.  We appreciate that the cynics amongst you will no doubt point out that your tests showed this marked increase in performance.  We remind you that we are fully aware of these theoretical increases in performance, but that none of these seem to translate into anything meaningful in your everyday gaming / other usage.

A couple of PMs between us and we decided to use our available hard and software without having to spend additional money. I didn't have any performance memory, so I was prepared to buy a 2 X 2GB kit. When I told my supplier about the purpose, he agreed that I can return it after the weekend, as long as it's not damaged in any way. No cost. Thanks go to Rune from Rebeltech.co.za for sponsoring the RAM for the tests. Thus I got a 4GB DDR3-2000 kit at no cost.

We have 4 batches of results: All batches contain the results obtained with our theoretical benchmark tool "MaxxMEM v1.95".

Batch 1 through to Batch 3; the FPS tests  were conducted with the Crysis Tool  and a GTX480 GPU at stock clocks with a 1680 X 1050 screen resolution.

Batch 4: the FPS tests were conducted using "Furmark" and the "FarCry2" bench tool , using the "Ranch long" option.  FarCry2 settings were all maxed out and both FPS tools were run at 1920X1080 screen resolution. The GPU for batch 4 is a GTX460 at 880/1760/1800 clocks.

The CPU frequency for all the tests was at 4,5GHz.

Now lets get to the results:
I included a trend line in the FPS graphs and you will notice that it's about as level as a highly polished Mahogany table ready for the big dinner party.

Batch 1:  2 X 2GB config with the Crysis tool tests:
This batch was done to test very tight timings at both 1333 and 1600.


Batch 2:   2 X 4GB config with the Crysis Tool tests.


Batch 3:   4 X 4GB config with the Crysis Tool tests.
These tests were not able to be conducted higher than 1866 due to memory restrictions.


Batch 4: 2 X 2GB config with FarCry2 and Furmark results. Furmark provides no decimals.
The 2133 test done at 11.11.11.30 timings was abbreviated as 11.30 in the column.


Conclusion and thoughts:
As can be seen through our own testing, the theoretical / synthetic benchmark increases proportionally with an increase in mem speed.  However all our FPS benchmarks conducted, confirm the test results in the article that Jack has already provided a link for.

My memory had one anomaly in terms of the CL8 timings where the gain over CL9 was minimal and in certain respects less. At CL7 though, where I was able to achieve it, it was markedly better. Whether that is the specific chips  that are in use, I don't know and I did not do any further investigation into that. HU16E believes that if I had changed the tRFC at the same time it may have improved.  I did not go as far as changing the tRFC values.

Further than that, the only difficulty experienced was trying to get the 2000 rated mem to run at 2133 (No real surprise). Blue screen after blue screen, until I increased the voltage to one notch above 1,65V and loosened the timings to all 11 and 30. That did the trick, but as can be seen from batch 4, the gain in even the theoretical benchmark between 2133 and 1866 is insignificant.

1333 CL7 or CL6 if one could achieve that would be the most sensible memory setting, and even then we are talkiong about less than 1 FPS difference at a 124FPS.
We really appreciate the effort and personal  time that HU16E put into this project of ours.

HU16E did some tests with lower tRFC values but it did not change the FPS scores noticably. Based on his own tests, here are some comments from him:
•   tRFC settings at 1333, 1866 & 2133 seem to be responsible to mess up the works. Set too low and all hell breaks loose.
•   It is best at the onset , to be liberal with the tRFC setting (stick to whats displayed in CPU-Z SPD Tab tables or the manufacturers recommendations) then tighten up a little at a time after having established a stable RAM setting first, to find the optimum tRFC.
•   A word of warning. Be careful that if you go too far with the tRFC, disaster can strike and you may end up bricking the board. I had the BIOS fail once and go beserk a second time with a double boot behaviour becoming evident. Once an ideal tRFC was reached though, everything was fine. For mainboards that don't have a backup BIOS chip, this can spell disaster, so be careful trying to tweak tRFC. It's best left alone.
•   A  lot of the problems are messed up manufacturer timings/SPD on their RAM. I have yet to have problems (X58 or P67) with RAM that have standard 1333, CR 1, 9 9 9 24, tRFC 88 @ 1.50v's 'ever' fail to work.
•   The only memory that I experienced problems with were the 4GB sticks as those were the only ones I have that don't have a standard 1333 SPD profile.

For all the forum users and others that may read this,  the real purpose of this excercise was to end this misinformed quest of wanting the highest rated memory for this platform. You are wasting your money and you are wasting unnecessary time trying to get this to run on this platform. Your money would be better spent on an SSD if you want to realise a performance gain, and a proper PSU to ensure an ample and stable supply.

Logged
1.B350F Gaming; R5-1600; Crucial MX300 275GB, Ballistix Sport LT BLS16G4D240FSC 2X8gb; Sapphire RX580 Pulse 8gb; Seasonic750W; Win10-64; LG DVD
2.FUJITSU LIFEBOOK S751 QM67;i5-2520M;2 X 4GB 1333Crucial CT256 MX100; Win10-64; Orico Dual Bay USB3 Dock



"veni vedi doctus"
MSI 미친

Froggy Gremlin

  • GENERAL OF THE ARMY
  • ******
  • Offline Offline
  • Posts: 11111
  • Master Air OC'er

Wanted to add that all the Crysis Tool tests were averaged from three runs of 3 passes in each configuration. Setting too low a tRFC in conjunction with other tightened timings can be a concern if you have a mainboard without having a dual BIOS chip to back you up if it all goes badly. In this extra test shown here with the CPU running at 4.7 to compare to running at 4.5, it didn't produce any significant increase in performance results. Actually, the results were rather disappointing. Changing the tRFC from 88 to 74 actually lost a .1 Mem Latency in ns. Only a small overall Mem Latency gain of .5 ns. FPS was actually a fraction worse as compared with most of the the 4.5/1866 results. 







(DRAM V. as per digital multi meter for 1870 was 1.600v's on the button).
Logged
The Last Hoorah! I Had Retired From All This! Can't Believe I'm Really Doing A Platform Test One More Time!  :shocked:

Z97 Gaming 7, 4790k, Zalman CNPS9900Max-R, Corsair AX-850, Corsair Dominator Platinum 2800's, WD 600 SATA III V.Raptor, Lite-On iHAS 424, GTX 680's In SLI, Creative X-Fi Fatal1ty Champion, Win 7 Ultimate 64 Bit    

Life Can Be Very Fulfilling Without Win 8/8.1, SSD's (Exception: Testing PCI-E M.2), RAID, Current AMD GPU's, G.Skill RAM, Hackers, Crackers, & Crypto Mining!

IXOYE

dilburt

  • SECOND LIEUTENANT
  • **
  • Offline Offline
  • Posts: 174

Sandy Bridge 1155-based processors contain a dual-channel memory controller with a maximum transfer rate of 1333 MT/s so I'm pretty sure there is a relation between that and the results you are finding.  The Nehalem X58-based processors contain a triple-channel memory controller that can move data over 2000 MT/s if I recall correctly so these higher memory clock-rates were probably developed originally for 1366 systems.  Of course, marketers are in the business of driving more sales so that probably slipped their minds when they made up the advertising materials for Sandybridge systems.

Here is another interesting article regarding memory selection for Sandy Bridge systems:

http://www.bit-tech.net/hardware/memory/2011/01/11/the-best-memory-for-sandy-bridge/1

- D
Logged
MSI Big-bang XPower, BIOS Ver 1.6
Intel I7-950, Corsair H80
Corsair 12g DDR3 CMX6GX3M3C1600C7
Antec Ninehundred case, AC PWM PST 120mm fans
EVGA GTX 570
2 - WD 1TB 7200 Sata 3 Drives
2 - LG WH10LS30 Blu-ray RW Drives
Silverstone 1100 PS

JackTopic starter

  • Ex Forum moderator
  • GENERAL OF THE ARMY
  • ******
  • Offline Offline
  • Posts: 14939
    • A user initiative by MSI fanatics to support all other MSI product users

Quote
The Nehalem X58-based processors contain a triple-channel memory controller that can move data over 2000 MT/s if I recall correctly

X58 / Socket 1366 processors have an integrated memory controller with a max. rating of DDR3-1066 [1066 MT/s] according to Intel's specifications.  However, X58/Socket 1366 based boards are not really the concern of this thread, even though similar results could probably be observed there.
Logged

Englishman

  • STAFF SERGEANT
  • *
  • Offline Offline
  • Posts: 93

I did a few tests with different memory frequencies using SuperPi (results averaged over 5 runs at each frequency). There is a difference, but not worth the effort of screenshots:

1333MHz - 7.97sec
1600MHz - 7.89sec
1866MHz - 7.80sec
2133MHz - 7.71sec

Also tried @32M and the scaling was almost linear.

You certainly won't notice much difference in the real world, but I like getting the machine as fast as possible, so happy to spend the little bit extra (£52 for 2133MHz RAM vs. £40 for 1600MHz RAM) to achieve 0.1sec - sad really ;o)
Logged
MSI P67A-GD65 (B3), Intel 2500K @ 4.8Ghz (1.385V), Corsair H100, 2*2GB GSkill F3-17000 @2133MHz (9,11,9,28 1.6v), 50GB OCZ Revo. 2*WD5000AAKX (6Gbs) RAID 1, ATI HD 4890, ASUS Xonar 2DX

EVGA 680i, Intel E8600 @ 4.5GHz, 4*1GB OCZ 2N1066SR1G @1066MHz, 2*Corsair X32 RAID 0, ATI 1950XT

Bernhard

  • Support the MSI Red Rockets Team Members 나는 한국을 사랑
  • GENERAL
  • *****
  • Offline Offline
  • Posts: 6248
  • Read the user manual first. Properly

The purpose of the post was to measure real world gains with applications/games. We did include the synthetic benchmarks to show that although the theoretical memory read/write and copy numbers increase, that the performance gain in gaming as an example is negligible. If someone wants to show their theoretical (potential RAM) numbers, then by all means blow your money on the memory. The real world effect is near 0 (zero).
Logged
1.B350F Gaming; R5-1600; Crucial MX300 275GB, Ballistix Sport LT BLS16G4D240FSC 2X8gb; Sapphire RX580 Pulse 8gb; Seasonic750W; Win10-64; LG DVD
2.FUJITSU LIFEBOOK S751 QM67;i5-2520M;2 X 4GB 1333Crucial CT256 MX100; Win10-64; Orico Dual Bay USB3 Dock



"veni vedi doctus"
MSI 미친

Englishman

  • STAFF SERGEANT
  • *
  • Offline Offline
  • Posts: 93

Agreed that for gaming there is no difference, but that isn't everyone's world. (I don't possess a single game)

Quoting from the link dilburt posted 3 entries above, 'Multi-tasking test showed a noticeable benefit in faster, lower latency memory, though. This is worth noting, since these are quad-core/eight-thread thread CPUs designed to do a multitude of things at once. The difference between the 1,333MHz kit at a sluggish CL9, and the 2,133MHz kit at CL8 was over 200 points.

As far as gaming goes, though, there's very little point in buying anything faster than 1,600MHz memory. Crysis couldn't care less about what memory you use, while Civ 5 ran equally as fast on the 1,600MHz kit at CL9 as the 2,133MHz at CL8. Only with the memory set to 1,333MHz did we see a slowdown, and even then by 1-3fps (around 5 per cent) at most.'

Logged
MSI P67A-GD65 (B3), Intel 2500K @ 4.8Ghz (1.385V), Corsair H100, 2*2GB GSkill F3-17000 @2133MHz (9,11,9,28 1.6v), 50GB OCZ Revo. 2*WD5000AAKX (6Gbs) RAID 1, ATI HD 4890, ASUS Xonar 2DX

EVGA 680i, Intel E8600 @ 4.5GHz, 4*1GB OCZ 2N1066SR1G @1066MHz, 2*Corsair X32 RAID 0, ATI 1950XT

JackTopic starter

  • Ex Forum moderator
  • GENERAL OF THE ARMY
  • ******
  • Offline Offline
  • Posts: 14939
    • A user initiative by MSI fanatics to support all other MSI product users

Quote
This is worth noting, since these are quad-core/eight-thread thread CPUs designed to do a multitude of things at once. The difference between the 1,333MHz kit at a sluggish CL9, and the 2,133MHz kit at CL8 was over 200 points.

That is exactly the kind of arguement that should be backed up with either a link to a specific review or some sort of screenshot that actually illustrates your point. Currently you gave us no real idea what that "200 points" difference actually means.  200 points of what/in what/what benchmark/usage scenario, whatever.  And also, what makes DDR3-1333 @CL9 "sluggish" in that picture.  Don't get me wrong, but I simply don't understand what exactly it is you are talking about. 
Logged

Englishman

  • STAFF SERGEANT
  • *
  • Offline Offline
  • Posts: 93

That is exactly the kind of arguement that should be backed up with either a link to a specific review or some sort of screenshot that actually illustrates your point. Currently you gave us no real idea what that "200 points" difference actually means.  200 points of what/in what/what benchmark/usage scenario, whatever.  And also, what makes DDR3-1333 @CL9 "sluggish" in that picture.  Don't get me wrong, but I simply don't understand what exactly it is you are talking about. 

Please read my post again - I was quoting (hence the quotation marks) a previous link that does indeed explain what '200 points' means and provides details of the benchmarks performed - much like your results in fact, they are very similar!
Logged
MSI P67A-GD65 (B3), Intel 2500K @ 4.8Ghz (1.385V), Corsair H100, 2*2GB GSkill F3-17000 @2133MHz (9,11,9,28 1.6v), 50GB OCZ Revo. 2*WD5000AAKX (6Gbs) RAID 1, ATI HD 4890, ASUS Xonar 2DX

EVGA 680i, Intel E8600 @ 4.5GHz, 4*1GB OCZ 2N1066SR1G @1066MHz, 2*Corsair X32 RAID 0, ATI 1950XT

Froggy Gremlin

  • GENERAL OF THE ARMY
  • ******
  • Offline Offline
  • Posts: 11111
  • Master Air OC'er

I do believe, with this platform, we are 'all' in agreement that for gaming, it is almost a dead heat across the board in regards to FPS, be it 1333 or 2133. As for real world copy, read, write, I had a video project that I could have tested the results of at 1333 versus 2133. Did it awhile before these posted tests were conducted. Maybe will when the next audio/video project comes up. Would be nice to know if any 'significant' minutes, seconds, or fractions of seconds are really, in practice, being gained with higher RAM speeds.  
Logged
The Last Hoorah! I Had Retired From All This! Can't Believe I'm Really Doing A Platform Test One More Time!  :shocked:

Z97 Gaming 7, 4790k, Zalman CNPS9900Max-R, Corsair AX-850, Corsair Dominator Platinum 2800's, WD 600 SATA III V.Raptor, Lite-On iHAS 424, GTX 680's In SLI, Creative X-Fi Fatal1ty Champion, Win 7 Ultimate 64 Bit    

Life Can Be Very Fulfilling Without Win 8/8.1, SSD's (Exception: Testing PCI-E M.2), RAID, Current AMD GPU's, G.Skill RAM, Hackers, Crackers, & Crypto Mining!

IXOYE

Bernhard

  • Support the MSI Red Rockets Team Members 나는 한국을 사랑
  • GENERAL
  • *****
  • Offline Offline
  • Posts: 6248
  • Read the user manual first. Properly

Here is another bunch of tests conducted by Hexus, that pretty much confirm similar results. http://www.hexus.net/content/item.php?item=29833
Logged
1.B350F Gaming; R5-1600; Crucial MX300 275GB, Ballistix Sport LT BLS16G4D240FSC 2X8gb; Sapphire RX580 Pulse 8gb; Seasonic750W; Win10-64; LG DVD
2.FUJITSU LIFEBOOK S751 QM67;i5-2520M;2 X 4GB 1333Crucial CT256 MX100; Win10-64; Orico Dual Bay USB3 Dock



"veni vedi doctus"
MSI 미친

sdk1985

  • SERGEANT
  • *****
  • Offline Offline
  • Posts: 39

I'm not sure why this post is sticky. It's not such a done deal as proposed in this thread...

To keep it simple and understandable i will be to the point:

Statement 1) Faster memory than 1333 is useless

Please checkout a more thorough test by bit-tech; http://www.bit-tech.net/hardware/memory/2011/01/11/the-best-memory-for-sandy-bridge/12

Highlights (I added percentages):

Memory Bandwith (MB/s)
Intel Core i7 (Sandy Bridge) 1,333MHz CL9   16985   100,00%
Intel Core i7 (Sandy Bridge) 1,600MHz CL9   20467   120,50%
Intel Core i7 (Sandy Bridge) 1,866MHz CL9   23725   139,68%
Intel Core i7 (Sandy Bridge) 2,133MHz CL9   25969   152,89%

Memory latency (nano seconds)
Intel Core i7 (Sandy Bridge) 1,333MHz CL9   76,3   100,00%
Intel Core i7 (Sandy Bridge) 1,600MHz CL9   68,2   89,38%
Intel Core i7 (Sandy Bridge) 1,866MHz CL9   62,1   81,39%
Intel Core i7 (Sandy Bridge) 2,133MHz CL9   58,8   77,06%


Real world performance (multitask 7zip+hd file)
Intel Core i7 (Sandy Bridge) 1,333MHz CL9   1449   100,00%
Intel Core i7 (Sandy Bridge) 1,600MHz CL9   1511   104,28%
Intel Core i7 (Sandy Bridge) 1,866MHz CL9   1596   110,14%
Intel Core i7 (Sandy Bridge) 2,133MHz CL9   1629   112,42%

Gaming crysis (fps)
Intel Core i7 (Sandy Bridge) 1,333MHz CL9   43,7   100,00%
Intel Core i7 (Sandy Bridge) 1,600MHz CL9   44,2   101,14%
Intel Core i7 (Sandy Bridge) 1,866MHz CL9   44,3   101,37%
Intel Core i7 (Sandy Bridge) 2,133MHz CL9   44,7   102,29%


Bittech conclusion:
"Conclusion of what to Buy
If you're the type of person that runs dozens of applications all at once, then a higher memory frequency does help, particularly when you're running demanding software. However, our testing shows that memory rated at over 1,866MHz doesn't give much extra performance. Worse still, in some applications only 1,333MHz memory gives a performance penalty, meaning that 1,600MHz memory is fine.
 
If you're doing anything other than heavy multi-tasking - this goes for gamers in particular - then a 1,600MHz or 1,866MHz kit is plenty. You could opt for CL8, as we saw some advantage in the video encoding test, but we wouldn't obsess over this factor, especially if a CL9 kit is much cheaper. "
'

Also keep in mind that overclocking your (k) cpu can also increase memory bandwith demands. More importantly getting lower CL modules seems to be a waste of money.

Statement 2) you can save money

Actually in the Netherlands 1600mhz and 1333 mhz do not differ that much in price, corsair modules are both around 87 euro for 8 GB. Valueram is 78 euro, so you could save 10 euro... Really you cannot get an SSD disk for that.

Statement 3) it is advisable to get an ssd for your system,

Not sure if the current price/performance of ssd's is that interesting. If you get a raid 0 array with 2x spinpoint f3 500 GB (2x 44 euro) you will reach 260 mb/s on linear read/writes, so for extracting and installing applications. This will destroy a lot of last year ssd's which have a write limit of around 80 MB/s and furthermore get less performance over time (performance degredation). Furthermore your burst rate is 4400 MB/s on GD65 ^^. Boot time and acces time will suffer compared to ssd but it will not impact your FPS. Furthermore boot time can be avoided completely by making use of S3 state.

My f3 raid 0 on sata II 3 Gb/s (port 3/4) http://min.us/mveD4gI .

Conclusion:
In the end if you want to spend 10 euro for a 20% in memory bandwith increase and an 11 % decrease of latency is your own call but it's certainly not a big mistake to make (!). Also keep in mind that most games currently are gpu limited therefore the true impact of cpu and thus memory performance in games is not yet completley visible.

ps if you do your own testing please include the number of tests you performed to get each result (you should perform more then one and have an acceptable standard deviation between different tests before a test means anything statistically).
Logged
I5 2500K, Scythe Mugen Rev B., P67-GD65, 2x 4GB Corsair XMS3, MSI 460 Hawk 1024MB, Corsair HX 750W, CM 690 II Advanced, 2x Samsung Spinpoint F3 500 GB (1 platter) in raid 0, 1x Samsung spoinpoint F3 1 TB, 1x Samsun Spinpoint F1 1 TB.

Froggy Gremlin

  • GENERAL OF THE ARMY
  • ******
  • Offline Offline
  • Posts: 11111
  • Master Air OC'er

Quote
Also keep in mind that most games currently are gpu limited therefore the true impact of cpu and thus memory performance in games is not yet completley visible.
When raising a GTX 480 from stock recently, setting 850/1700/2000, the FPS jumped to a 96.125 average. This was regardless of RAM speed 1333 or 2133, or CPU speed from 4.5 to 4.7. CPU speed is definately a factor, but VGA settings and/or quality have a pretty good impact as far as gaming is concerned. Having pointed that out, how many FPS is really enough to run any game on the market today? It starts to become nothing more than a numbers game. I see no difference in game performance if the average FPS is 80 or 130. 1333 RAM can still get the job done with less cost, less voltage, & in many situations, less compatibility problems. For whatever reason, RAM speed is rather insignificant when it comes to gaming with this platform. For those that don't game, the benefits of faster RAM is left up to them, but at least an attempt to get as much information out about it to make an informed decision is a good thing, no? 
Logged
The Last Hoorah! I Had Retired From All This! Can't Believe I'm Really Doing A Platform Test One More Time!  :shocked:

Z97 Gaming 7, 4790k, Zalman CNPS9900Max-R, Corsair AX-850, Corsair Dominator Platinum 2800's, WD 600 SATA III V.Raptor, Lite-On iHAS 424, GTX 680's In SLI, Creative X-Fi Fatal1ty Champion, Win 7 Ultimate 64 Bit    

Life Can Be Very Fulfilling Without Win 8/8.1, SSD's (Exception: Testing PCI-E M.2), RAID, Current AMD GPU's, G.Skill RAM, Hackers, Crackers, & Crypto Mining!

IXOYE

sdk1985

  • SERGEANT
  • *****
  • Offline Offline
  • Posts: 39

Yes informed decisions are always good:). The topicstarter is completely right that paying premium for 2000+mhz modules is a waste of money in terms of gaming. Just pointing out that it actually can increase performance according to respected web sites and that in general 1600 mhz is recommended for this platform and not 1333 mhz. So please don't forget 16000mhz, that seems to be the sweet spot for this chipset. (ps intel will officially support 1600 mhz in the new cpu's comming 2012 if you didnt already know know :) ).

Anyways according to bit-tech for gaming it adds 1%, so regarding that you are completely right; it's not that important. However, that should not mean 1333 mhz is recommended for all users of a p67 chipset. It's is an increase, it costs only a few euro's, you gain multitasking performance and the fps is still not 60 for this insane game ^^.  For me personally I am expecting to replace my graphics card once every year (for the minimum of 3 years that I want to keep this platform) thus eventually my cpu will become the bottleneck of my system. At this time every percent in performance I gain will be welcome since upgrading a graphic card at this point will no longer result in higher fps.

When raising a GTX 480 from stock recently, setting 850/1700/2000, the FPS jumped to a 96.125 average. This was regardless of RAM speed 1333 or 2133, or CPU speed from 4.5 to 4.7. CPU speed is definately a factor, but VGA settings and/or quality have a pretty good impact as far as gaming is concerned. Having pointed that out, how many FPS is really enough to run any game on the market today? It starts to become nothing more than a numbers game....

You can also check this with gpu-z and notice your gpu load is probably 99% all the time. But here the discussion becomes difficult... Statistically speaking you do not want the gpu to limit and influence your results, it should be a contant not a limiting factor. The way I interpret your results in your case the gpu is the limiting factor; If you get higher fps with a faster gpu, but similar fps with 4.5 ot 4.7 mhz then this must be the limiting factor. This probably explains why the ram also has no influence:). To prevent this issue often tests are run at e.g. 800x600. Then it is likely the gpu does not get 99% loaded. Even call of duty 4 at 1024x768 still shows signs of gpu limitation (http://www.techpowerup.com/reviews/MSI/GTX_460_HAWK/9.html ). However running tests at low resolutionis not a real every day application so one could argue about the use of this.

Regarding the fps, my brand new gtx460 is an improvement over my gtx 260 (192core) but still I can see 30-40 framerates at some parts of just cause 2. This is a game from march 2010. For those who aren't so luck to have the most expensive hardware, the framerate is still something to tweak our hardware for ^^.
Logged
I5 2500K, Scythe Mugen Rev B., P67-GD65, 2x 4GB Corsair XMS3, MSI 460 Hawk 1024MB, Corsair HX 750W, CM 690 II Advanced, 2x Samsung Spinpoint F3 500 GB (1 platter) in raid 0, 1x Samsung spoinpoint F3 1 TB, 1x Samsun Spinpoint F1 1 TB.

JackTopic starter

  • Ex Forum moderator
  • GENERAL OF THE ARMY
  • ******
  • Offline Offline
  • Posts: 14939
    • A user initiative by MSI fanatics to support all other MSI product users

Quote
I'm not sure why this post is sticky.

To provoce critical and thorough discussion.
Logged

Bernhard

  • Support the MSI Red Rockets Team Members 나는 한국을 사랑
  • GENERAL
  • *****
  • Offline Offline
  • Posts: 6248
  • Read the user manual first. Properly

Quote
intel will officially support 1600 mhz in the new cpu's comming 2012 if you didnt already know
So it seems from the latest information available. Currently though its still 1333 for the Sandy Bridge.
We welcome anyone that can provide us with stats on their own experience where memory speeds higher than 1333 have provided them with a marked increase in performance of a specific application or game on the P67 platform, which is proportionate to the percentage increase in memory speed above 1333.
Logged
1.B350F Gaming; R5-1600; Crucial MX300 275GB, Ballistix Sport LT BLS16G4D240FSC 2X8gb; Sapphire RX580 Pulse 8gb; Seasonic750W; Win10-64; LG DVD
2.FUJITSU LIFEBOOK S751 QM67;i5-2520M;2 X 4GB 1333Crucial CT256 MX100; Win10-64; Orico Dual Bay USB3 Dock



"veni vedi doctus"
MSI 미친

sdk1985

  • SERGEANT
  • *****
  • Offline Offline
  • Posts: 39

So it seems from the latest information available. Currently though its still 1333 for the Sandy Bridge.
We welcome anyone that can provide us with stats on their own experience where memory speeds higher than 1333 have provided them with a marked increase in performance of a specific application or game on the P67 platform, which is proportionate to the percentage increase in memory speed above 1333.
@Jack Let's hope Im adding to the discussion then ^^.

The marked increase has been proven by bit-tech. The second requirement would be ridiculous to expect  :rolleyes_anim: (!). The only real proportionate variable linked to pc performance is cpu frequency. My 3dmark cpu score was 45% lower in both scores when my cpu was stuck on x16 multiplier (16/34=0,47).  Also my superpi 1M was 23,384 (16x multiplier) compared to 10,701 stock which is 0,43% of the previous time ^^ (turbo also didnt work @ x16, so its x38-x34 for 10,701). So cpu frequency is proportionate. But that is the only value I ever found to be directly proportionate.

You have to think proportionate in terms of added cost (!). That is what really matters. The added cost for 1600mhz is 0-10 euro for branded ram (corsair) depending on your store and 7-10 euro compared to valueram at 1333mhz.

This works for most pc components, for example the gpu. If you dont believe me then install msi kombustor (or furmark) and try for yourself. I made an example run:
I changed my gtx460 to 700 mhz, and run kombustor. The result was avarage 284 fps after 30 seconds. Now I did a 10% core increase to 770 mhz. The result was 292 fps. This is an increase of 292/284=2,8%. So you actually gain 2,8% after increasing clocks by 10%. So that is why the mhz of ram is irrelevant, you cannot expect a 20% performance boost just because the frequency is 20% higher. What matters is performance + added cost. This is true for every component of the pc... I dont expect my 780/1800 hawk to perform 15% better just because it has 15% higher clocks ;).

Personally I dont mind the 10 euro investment for a 1,4% increas in gaming and 4 % increase in multitasking, for this 10 euro I could not have upgraded any other component in my system:). A gtx560 would have cost me €61. And for the lucky few that have 400 euro cards, why would you want to save 10 euro and have a performance penalty??? Even if it's 1-4% if you buy a graphics card for 400 euro it seems a bit odd to save 10 euro on memory.

I dont think I will be benchmarking 1333 vs 1600 mhz soon because for this you have to disable eist and the turbo to exclude variations and then test 2x 10x on a clean OS, also I thrust bit-tech.net. Interesting to see that hexus tested 1600 mhz vs 2133 and not 1333. Maybe some more reviews exist:).

ps I still think in general we completely agree on the not needing ddr 1866/2133 however I am hoping to convice you that 1600 mhz is usefull for sandy bridge:).
Logged
I5 2500K, Scythe Mugen Rev B., P67-GD65, 2x 4GB Corsair XMS3, MSI 460 Hawk 1024MB, Corsair HX 750W, CM 690 II Advanced, 2x Samsung Spinpoint F3 500 GB (1 platter) in raid 0, 1x Samsung spoinpoint F3 1 TB, 1x Samsun Spinpoint F1 1 TB.

Bernhard

  • Support the MSI Red Rockets Team Members 나는 한국을 사랑
  • GENERAL
  • *****
  • Offline Offline
  • Posts: 6248
  • Read the user manual first. Properly

Quote
I still think in general we completely agree on the not needing ddr 1866/2133 however I am hoping to convice you that 1600 mhz is usefull for sandy bridge:).
  :biggthumbsup:
The intent was to have a realistic overview of the real gains above 1333. Those remain insignificant excepting benchmarks and the odd program like zip. Furthermore, the amount of troubles and instabilities experienced by the user community of trying to run at 1600 or above, had us weigh up the potential gains vs the troubles to get the systems stable. For those that believe that they gain performance with the overclocked RAM, there is no convincing needed, and if it works for you, great. For those that intend to buy memory for this platform, the advice remains, stick with the native 1333 as the marginal overall gain for overclocked memory is not worth the potential instabilities. This sticky is after all for the majority and not for individuals that have mastered their memory overclocking skills.
Logged
1.B350F Gaming; R5-1600; Crucial MX300 275GB, Ballistix Sport LT BLS16G4D240FSC 2X8gb; Sapphire RX580 Pulse 8gb; Seasonic750W; Win10-64; LG DVD
2.FUJITSU LIFEBOOK S751 QM67;i5-2520M;2 X 4GB 1333Crucial CT256 MX100; Win10-64; Orico Dual Bay USB3 Dock



"veni vedi doctus"
MSI 미친

sdk1985

  • SERGEANT
  • *****
  • Offline Offline
  • Posts: 39

 :biggthumbsup:
The intent was to have a realistic overview of the real gains above 1333. Those remain insignificant excepting benchmarks and the odd program like zip. Furthermore, the amount of troubles and instabilities experienced by the user community of trying to run at 1600 or above, had us weigh up the potential gains vs the troubles to get the systems stable. For those that believe that they gain performance with the overclocked RAM, there is no convincing needed, and if it works for you, great. For those that intend to buy memory for this platform, the advice remains, stick with the native 1333 as the marginal overall gain for overclocked memory is not worth the potential instabilities. This sticky is after all for the majority and not for individuals that have mastered their memory overclocking skills.

Hmm i havent heard about memory issues in my community. But maybe everyone I know just was smart enough to get memory from the vendor qualified list on msi.com ^^. Anyways I would advice those that do want 1600mhz or above touse modules with advanced SPD (XMP etc) since this is very easy to enable in bios (xmp profile>enable  :lol_anim:) and requires no further configuration (xmp includes voltage, commandrate and all required latencies):). Really with xmp you dont need any experience and it is guaranteed to work.

edit: ps are you sure your memory settings are ok, I have 8.5 ns with 19.12 GB/sec @ 1600mhz XMP (9-9-9-24,41, 2T).
http://min.us/mPsjSrUTMrJJc#1
Logged
I5 2500K, Scythe Mugen Rev B., P67-GD65, 2x 4GB Corsair XMS3, MSI 460 Hawk 1024MB, Corsair HX 750W, CM 690 II Advanced, 2x Samsung Spinpoint F3 500 GB (1 platter) in raid 0, 1x Samsung spoinpoint F3 1 TB, 1x Samsun Spinpoint F1 1 TB.

Bernhard

  • Support the MSI Red Rockets Team Members 나는 한국을 사랑
  • GENERAL
  • *****
  • Offline Offline
  • Posts: 6248
  • Read the user manual first. Properly

Quote
ps are you sure your memory settings are ok, I have 8.5 ns with 19.12 GB/sec @ 1600mhz XMP (9-9-9-24,41, 2T).
8.5ns ?  :think: mmmm exceptional.  typo perhaps ?
even your pasted bit tech reference posted as follows. Intel Core i7 (Sandy Bridge) 1,600MHz CL9   68,2ns
Logged
1.B350F Gaming; R5-1600; Crucial MX300 275GB, Ballistix Sport LT BLS16G4D240FSC 2X8gb; Sapphire RX580 Pulse 8gb; Seasonic750W; Win10-64; LG DVD
2.FUJITSU LIFEBOOK S751 QM67;i5-2520M;2 X 4GB 1333Crucial CT256 MX100; Win10-64; Orico Dual Bay USB3 Dock



"veni vedi doctus"
MSI 미친

Froggy Gremlin

  • GENERAL OF THE ARMY
  • ******
  • Offline Offline
  • Posts: 11111
  • Master Air OC'er

Quote
even your pasted bit tech reference posted as follows. Intel Core i7 (Sandy Bridge) 1,600MHz CL9   68,2ns
68.2 ns? Used command rate 2 maybe? My tests with CR 1 were 49.9 ns & 50.6 ns depending which modules installed. Very respectable 48.8 ns when at 1333C6. :think:
Logged
The Last Hoorah! I Had Retired From All This! Can't Believe I'm Really Doing A Platform Test One More Time!  :shocked:

Z97 Gaming 7, 4790k, Zalman CNPS9900Max-R, Corsair AX-850, Corsair Dominator Platinum 2800's, WD 600 SATA III V.Raptor, Lite-On iHAS 424, GTX 680's In SLI, Creative X-Fi Fatal1ty Champion, Win 7 Ultimate 64 Bit    

Life Can Be Very Fulfilling Without Win 8/8.1, SSD's (Exception: Testing PCI-E M.2), RAID, Current AMD GPU's, G.Skill RAM, Hackers, Crackers, & Crypto Mining!

IXOYE

sdk1985

  • SERGEANT
  • *****
  • Offline Offline
  • Posts: 39

8.5ns ?  :think: mmmm exceptional.  typo perhaps ?
even your pasted bit tech reference posted as follows. Intel Core i7 (Sandy Bridge) 1,600MHz CL9   68,2ns

I included a screenshot it really says 8.5 ns, I used the tool that was listed in this topic. Mabye it's a bad tool,  never heard of it before. That could explain why the website says it will be released in 2010;). Everest used to have a nice tool but aida no longer has it. Can you recommend another tool  :bonk:?
Logged
I5 2500K, Scythe Mugen Rev B., P67-GD65, 2x 4GB Corsair XMS3, MSI 460 Hawk 1024MB, Corsair HX 750W, CM 690 II Advanced, 2x Samsung Spinpoint F3 500 GB (1 platter) in raid 0, 1x Samsung spoinpoint F3 1 TB, 1x Samsun Spinpoint F1 1 TB.

Froggy Gremlin

  • GENERAL OF THE ARMY
  • ******
  • Offline Offline
  • Posts: 11111
  • Master Air OC'er

Quote
Can you recommend another tool?
MaxxMEM2, V1.95 seems to be very accurate, & makes a small footprint in the system. http://www.maxxpi.net

For all published averages with my tests, it was run three times. The screenshot in Reply #2 was an average, so I included it in the post. I had one just .1 above it, & one with just .1 below it. :-))
Logged
The Last Hoorah! I Had Retired From All This! Can't Believe I'm Really Doing A Platform Test One More Time!  :shocked:

Z97 Gaming 7, 4790k, Zalman CNPS9900Max-R, Corsair AX-850, Corsair Dominator Platinum 2800's, WD 600 SATA III V.Raptor, Lite-On iHAS 424, GTX 680's In SLI, Creative X-Fi Fatal1ty Champion, Win 7 Ultimate 64 Bit    

Life Can Be Very Fulfilling Without Win 8/8.1, SSD's (Exception: Testing PCI-E M.2), RAID, Current AMD GPU's, G.Skill RAM, Hackers, Crackers, & Crypto Mining!

IXOYE

boby7

  • STAFF SERGEANT
  • *
  • Offline Offline
  • Posts: 61

MaxxMEM2, V1.95 seems to be very accurate
Sadly enough this program seems to have compatibility problem and can not run properly on my pc. It gives me 150ns latency and leaves hanging CPU-Z process running in background. Restart needed :/
Logged
MSI P67A-GD65(B3) BIOS beta V4.39, 2500k@4.2, MSI GTX560ti TFIIOC, Cogage Arrow, Geil 2x4GB DDR3 1600 9-8-8-26-1T, Samsung 840 EVO, LG DVD-RW, W10 Pro, X-Fi ExtremeGamer.

Bernhard

  • Support the MSI Red Rockets Team Members 나는 한국을 사랑
  • GENERAL
  • *****
  • Offline Offline
  • Posts: 6248
  • Read the user manual first. Properly

Quote
Mabye it's a bad tool,
MaxxMEM V1.95 is needed for the Sandy Bridge Platform. Earlier versions are not compatible and provide erroneous results.
Logged
1.B350F Gaming; R5-1600; Crucial MX300 275GB, Ballistix Sport LT BLS16G4D240FSC 2X8gb; Sapphire RX580 Pulse 8gb; Seasonic750W; Win10-64; LG DVD
2.FUJITSU LIFEBOOK S751 QM67;i5-2520M;2 X 4GB 1333Crucial CT256 MX100; Win10-64; Orico Dual Bay USB3 Dock



"veni vedi doctus"
MSI 미친

boby7

  • STAFF SERGEANT
  • *
  • Offline Offline
  • Posts: 61

Report:
CPU: 2500K@4500
RAM: 2x4GB Geil, tested at 1333 9-9-9-28-2T and 1600 9-9-9-28-1T. Timings the same to check only bandwidth benefit.
Tested: 3DMark06 (640x480px), Cinebech (multi and single), SuperPi (2MB) and short AfterEffects CS5 composition.
It seams that only SuperPi can get some benefits from 20% more memory bandwidth. Here my results:

                       1600     1333       gain
3DMark Score   25258   25204     100,2%
SM2.0 Score       9089     9095       99,9%
HDR/SM3.0       12163   12120     100,4%
CPU Score           7901    7869     100,4%

CINEBENCH M      7,15         7,13       100,3%
CINEBENCH S      1,81         1,82         99,5%

super pi            43,68 sec     44,97 sec    103,0%

AE CS5 single      106 sec      107 sec   100,9%
AE CS5 multi        94 sec        95 sec    101,1%

PS: It is not proffessional test but I was just curious :)

Logged
MSI P67A-GD65(B3) BIOS beta V4.39, 2500k@4.2, MSI GTX560ti TFIIOC, Cogage Arrow, Geil 2x4GB DDR3 1600 9-8-8-26-1T, Samsung 840 EVO, LG DVD-RW, W10 Pro, X-Fi ExtremeGamer.

RemusM

  • Memory Expert
  • LIEUTENANT COLONEL
  • *
  • Offline Offline
  • Posts: 2171
    • Necromanthus

There seems to be absolutely no significant performance bonus regarding all the other benchtests when DDR3-2133 is chosen over DDR3-1333.

 :biggthumbsup:
Above DDR3-1333 you'll get 2 cents performances boost and 200 bucks troubles.
Something more:
- most of the DDR3-1600 modules on the market use overvolted DDR3-1333 chips (9-9-9-24)
- most of the DDR3-1866, 2000 & 2133 modules on the market use overvolted DDR3-1600 chips
(9-9-9-27, 10-10-10-27, 11-11-11-30)
- there're 2 ways to overclock & overvolt DDR3-1333 chips (by default 9-9-9-24 at 1.5V):
a) faster access: 1333MHz at 8-8-8-24 or 7-7-7-20
b) increased speed: 1600MHz at 9-9-9-27 (or even at CAS = 8)
In both cases you need voltages above 1.60V
Stay away from "tight timings".
In most of the cases: insignificant performances boost, but serious stability issues, abnormal behaviours and shorter lifetime.
 :beerchug:
Logged

Mainboard: HC85
Processor: Zilog Z80 3.5MHz
System RAM: 48KB
Video RAM: 16KB (only 6912 bytes are used for Pixel Shader effects)
Video Card: integrated (16 colors)
PSU: 5V/3A
OS: BASIC Spectrum Sinclair

note: NO overclocking!

Aaron

  • GENERAL
  • *****
  • Offline Offline
  • Posts: 5690
  • BOOST!

Remus, I thought you'd disappeared!  I should've known this thread would draw your attention. :lol_anim:



I think that the wrong crowd have been reading this thread.  I don't believe the idea was to challenge the verdict, rather to make a decision from it.  If you think DDR-1333 isn't enough then you're clearly not the type of person this thread should appeal to.

For all the trouble overvoltaged and overclocked memory causes, I can personally say that it's not worth it.  And believe me and Jack - we've been active members here for 5 years; Jack is a current moderator and I was one previously.  The amount of "Can't get memory to work at <insert overvoltaged and overclocked rated speed>" threads we've witnessed is incredible.  Simply because these users do not even know they have to overclock said memory to begin with.

If you know what you're doing - sure it's probably worth it.  If you're an overclocker that's what you want to do: overclock.  If you want to overclock, you'd buy the tweaking-required higher specification memory.  If you just want it to work, you'll by the plug and play standard modules.

I too was in many users' opinions, obsessed with numbers and nanoseconds difference.  Wanting to make every MHz, every FPS and MB/s count.  I do understand the points sdk1985 has thrown into the discussion, but I do not believe they serve the same purpose.

To summarise:

- The point of this thread was does DDR-1333 versus DDR-1600+ create any real world difference?  Excluding compression/decompression, the answer is no.
- Is the ~1% performance gain worth the 8-15% extra price?  The answer is no.

If the nanoseconds difference matters to you... why are you reading this thread to begin with? ;-))

Thanks for reading. (If you even got this far.)
Logged
Refrain from having to solve most of your problems by not creating them in the first place.

DDR3 Memory Standards: Blurry Lines
Integrated Memory Controllers (IMC)
Intel EIST and Processor C-states
Overclock Database 2.0

sdk1985

  • SERGEANT
  • *****
  • Offline Offline
  • Posts: 39

To summarise:

- The point of this thread was does DDR-1333 versus DDR-1600+ create any real world difference?  Excluding compression/decompression, the answer is no.
- Is the ~1% performance gain worth the 8-15% extra price?  The answer is no.

If the nanoseconds difference matters to you... why are you reading this thread to begin with? ;-))

Thanks for reading. (If you even got this far.)
Don't make the mistake to compare 2 relative values, it has no meaning. If memory would cost 1 cent and ddr2133 would cost 4 then it would be 400% of the cost... However I would like a few percent of performance increase for just 3 cents... Read above it has been noted before ^^. You should compare the 1% performance boost to the 5-10 euro (which completely depends on luck and daily ram pricing) cost and then decide if thats worth it for you (given the chance of stabilty issues if you chose the wrong ram). Personally I had to spend 165 euro on a gtx460 which gave me a 30% performance increase compared to the gtx260 I already owned. Thus I payed 5,5 euro for each percent. I payed a 5 euro premium for ddr1600 low volt from corsair which translates in 4,38 euro per percent (if we go on the 1.14% performance increase in crysis noted earlier). Of course this is an example which is not applicable globally but you need to understand that comparing two different relative values means nothing at all... You can only comprare relative values between differently priced memory modules , but you should not forget about the actual monetary value involved compared to the other components in the system...

Doesnt mean that I do not agree with the topic title atm, 1333 is indeed enough (and so is a 550W power supply ,4 GB ram, a mid ranged graphics card and a 30 euro ms-tech case ^^).

Anyways @ Remus dont forget that if you do get faster rated memory that there are special 1.5v editions and that it is not recommended to buy the memory modules from the previous platform that run at 1.65v.

ps
I actually payed less per overall performance for the type of graphics card I chose, checkout this calculation
http://min.us/ljRLOu
At the time I made the calculation it would have been 18 euro for 10% avarage performance comparing the HAWK with the default gtx 460 thus €1,80 per percent which of course destroys the added value of memory (but still it's not 5 euro but 20) :). But the default gtx460 is more cost efficient at 1680x1050 (-1,56% for the hawk)... However factoring in the chance that I would probably switch to a new display somwhere in the next 2 years and the overclock ability I ended up chosing the Hawk. In the end you can do a lot with numbers and the most important thing is to know how to interpret them:).
Logged
I5 2500K, Scythe Mugen Rev B., P67-GD65, 2x 4GB Corsair XMS3, MSI 460 Hawk 1024MB, Corsair HX 750W, CM 690 II Advanced, 2x Samsung Spinpoint F3 500 GB (1 platter) in raid 0, 1x Samsung spoinpoint F3 1 TB, 1x Samsun Spinpoint F1 1 TB.

Aaron

  • GENERAL
  • *****
  • Offline Offline
  • Posts: 5690
  • BOOST!

Don't make the mistake to compare 2 relative values, it has no meaning. If memory would cost 1 cent and ddr2133 would cost 4 then it would be 400% of the cost... However I would like a few percent of performance increase for just 3 cents... Read above it has been noted before ^^. You should compare the 1% performance boost to the 5-10 euro (which completely depends on luck and daily ram pricing) cost and then decide if thats worth it for you (given the chance of stabilty issues if you chose the wrong ram). Personally I had to spend 165 euro on a gtx460 which gave me a 30% performance increase compared to the gtx260 I already owned. Thus I payed 5,5 euro for each percent. I payed a 5 euro premium for ddr1600 low volt from corsair which translates in 4,38 euro per percent (if we go on the 1.14% performance increase in crysis noted earlier). Of course this is an example which is not applicable globally but you need to understand that comparing two different relative values means nothing at all... You can only comprare relative values between differently priced memory modules , but you should not forget about the actual monetary value involved compared to the other components in the system...

Not exactly.  I'm not trying to look at this in a mind state of dissecting everything.  The only statistic that matters to the average user (you know, the one I said this thread should appeal to - not the obvious overclocker that you are) is the price difference.  The nanoseconds difference is not worth the extra price as far as I am concerned.

I shall now start using numbers; since you insist.

Cheapest DDR-1333 kit:  £34.99

Cheapest DDR-1600 kit:  £34.79

Cheapest DDR-2133 kit:  £54.98

The £20 price over the DDR-1333 and 1600 modules is for what?  It is not worth the price premium, and the difficulties some users to go through to get it working.  Same for the DDR-1600 - since no desktop CPU currently supports anything above DDR-1333.

I would like to refer to the post you quoted:

If the nanoseconds difference matters to you... why are you reading this thread to begin with? ;-))

I'm glad you've formed your own opinion on this, but I don't see anyone else changing to your point of view.

Have a nice day.
Logged
Refrain from having to solve most of your problems by not creating them in the first place.

DDR3 Memory Standards: Blurry Lines
Integrated Memory Controllers (IMC)
Intel EIST and Processor C-states
Overclock Database 2.0

TT101

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 19

Is it worth upgrading from 2x4gb 1333mhz to 2x 4gb 1600mhz for my system in my sig? will i see any significant gains? I am running the cpu at stock but want to overclock in future , maybe it will be easier to overclock with 1600 mhz ram with a heatsink than my current 1333 mhz without a heatsink?
Logged
i5 3470
Gigabyte B75M-D3H
2 x 8GB G.SKILL Ares 1833 Mhz
Sapphire HD6850 GPU
1tb Seagate HDD
Corsair 400cx PSU
Windows 8 pro x64
LG DVDRW


....................................... ................
Intel i3 550
MSI H55M-P33 ( Bios Ver 1.A )
1x4gb corsair Value select ddr3 1333
1tb Seagate HDD
Corsair 430cx PSU
Windows 8 pro x64
Sony DVDRW AD-7260S

JackTopic starter

  • Ex Forum moderator
  • GENERAL OF THE ARMY
  • ******
  • Offline Offline
  • Posts: 14939
    • A user initiative by MSI fanatics to support all other MSI product users
Logged

TT101

  • CORPORAL
  • ****
  • Offline Offline
  • Posts: 19

^ please see updated post
Logged
i5 3470
Gigabyte B75M-D3H
2 x 8GB G.SKILL Ares 1833 Mhz
Sapphire HD6850 GPU
1tb Seagate HDD
Corsair 400cx PSU
Windows 8 pro x64
LG DVDRW


....................................... ................
Intel i3 550
MSI H55M-P33 ( Bios Ver 1.A )
1x4gb corsair Value select ddr3 1333
1tb Seagate HDD
Corsair 430cx PSU
Windows 8 pro x64
Sony DVDRW AD-7260S

Froggy Gremlin

  • GENERAL OF THE ARMY
  • ******
  • Offline Offline
  • Posts: 11111
  • Master Air OC'er

Quote
Is it worth upgrading from 2x4gb 1333mhz to 2x 4gb 1600mhz for my system in my sig? will i see any significant gains?
Answer is for all practical purposes, still 'no'.
Logged
The Last Hoorah! I Had Retired From All This! Can't Believe I'm Really Doing A Platform Test One More Time!  :shocked:

Z97 Gaming 7, 4790k, Zalman CNPS9900Max-R, Corsair AX-850, Corsair Dominator Platinum 2800's, WD 600 SATA III V.Raptor, Lite-On iHAS 424, GTX 680's In SLI, Creative X-Fi Fatal1ty Champion, Win 7 Ultimate 64 Bit    

Life Can Be Very Fulfilling Without Win 8/8.1, SSD's (Exception: Testing PCI-E M.2), RAID, Current AMD GPU's, G.Skill RAM, Hackers, Crackers, & Crypto Mining!

IXOYE

tolyaasch

  • SERGEANT
  • *****
  • Offline Offline
  • Posts: 33

I do the same mistake: bought Corsair XMS3 1600C9 modules and yesterday I have problem with my system and after 2 days trying to find the reason one of these was the 1600 modules are going "stable" on 1333 mode :/
And seems to me the CMX6GX3M3A1600C9 modules are partly incompatible of MSI P67-C43 boards :)
Have a nice day :)
Logged

monographix

  • STAFF SERGEANT
  • *
  • Offline Offline
  • Posts: 57

I know of an application - Game where memory speed used to play a very significant role until P35 at least. Flight simulators and more particularly FSX which was directly and noticeably affected by memory speed (and size) . And people who are interested in flight simulators have permanently installed more than 2 of those the least and our demands in lack of stutter and smoothness , besides high FPS are a lot higher than any other simulator (driving) or gaming. I am curious what the differences would be between 1333/1600/1800/2000 with current chipsets @ FSX
Logged
Q6600 @ 3 :: 2x1Gb Transcend Axeram 800 @ 1066 5-5-5-15 2.10v:: 8800GT @ 700/1000
MSI P35 Neo2-FR :: Hiper 480 Type-R :: Raptor 36 :: WD2500YS

dick123

  • PRIVATE E-2
  • **
  • Offline Offline
  • Posts: 2

yes its a nice post so thanks for share with me.
Logged

SonDa5

  • SERGEANT
  • *****
  • Offline Offline
  • Posts: 49

Will be interesting to see how IB on S1155 will perform.
Logged
Intel i7-3770k
MSI Z77 MPower
OCZ Agility 4 RAID 0
Windows 7 64bit Pro
Seasonic Platinum 860W PSU
BenQ XL2410T Display
HD 7950
Samsung Green 2x4GB 2200mhz DDR3, CL9 1.52v

Bernhard

  • Support the MSI Red Rockets Team Members 나는 한국을 사랑
  • GENERAL
  • *****
  • Offline Offline
  • Posts: 6248
  • Read the user manual first. Properly

Quote
Will be interesting to see how IB on S1155 will perform.
That depends all on the memory controller that will be used. Perhaps it will be true 1600 rated controller. But when the Ivy Bridge CPUs are released, you are welcome to conduct the testing and provide feedback.  ;D
Logged
1.B350F Gaming; R5-1600; Crucial MX300 275GB, Ballistix Sport LT BLS16G4D240FSC 2X8gb; Sapphire RX580 Pulse 8gb; Seasonic750W; Win10-64; LG DVD
2.FUJITSU LIFEBOOK S751 QM67;i5-2520M;2 X 4GB 1333Crucial CT256 MX100; Win10-64; Orico Dual Bay USB3 Dock



"veni vedi doctus"
MSI 미친

SonDa5

  • SERGEANT
  • *****
  • Offline Offline
  • Posts: 49

IB S1155 is supposed  to have improved memory performance.  I plan on getting i5-3570k when it is available.  Hope it lives up to the hype.  I will keep you guys posted.
Logged
Intel i7-3770k
MSI Z77 MPower
OCZ Agility 4 RAID 0
Windows 7 64bit Pro
Seasonic Platinum 860W PSU
BenQ XL2410T Display
HD 7950
Samsung Green 2x4GB 2200mhz DDR3, CL9 1.52v

Froggy Gremlin

  • GENERAL OF THE ARMY
  • ******
  • Offline Offline
  • Posts: 11111
  • Master Air OC'er

Not sure there is confirmation that Ivy will be supported on the P67 mainboards yet even with a UEFI/BIOS update. More information should be available before Ivy's launch time in March of 2012.
Logged
The Last Hoorah! I Had Retired From All This! Can't Believe I'm Really Doing A Platform Test One More Time!  :shocked:

Z97 Gaming 7, 4790k, Zalman CNPS9900Max-R, Corsair AX-850, Corsair Dominator Platinum 2800's, WD 600 SATA III V.Raptor, Lite-On iHAS 424, GTX 680's In SLI, Creative X-Fi Fatal1ty Champion, Win 7 Ultimate 64 Bit    

Life Can Be Very Fulfilling Without Win 8/8.1, SSD's (Exception: Testing PCI-E M.2), RAID, Current AMD GPU's, G.Skill RAM, Hackers, Crackers, & Crypto Mining!

IXOYE

maxx_tools

  • PRIVATE E-2
  • **
  • Offline Offline
  • Posts: 2

hi,

well i saw a lot of Referring url (to maxxpi.net) back to this thread...

i can say, as the coder behind maxxmem and all maxxpi tools,
that the memory performance massively increase the FPS for example.

BUT, not the maximum reached ones, or the average... but rather the MINIMUM reached FPS
and this much more interesting as the both other ones.

and of course all applications that use much memory like databases, modelling/CAD tools get a huge +performance.

if you take a look here *Memory multi* benchmarks link, you will see the sandys seperatet from the other memory typs
(i will do this in all other memory-diagramm's too)

the lowest result starts at ~11gbyte/sec. and goes up to ~22gbyte/sec.
do you really think that this increase of 100% ! will be really not noticeable ? ...

think about it.





Logged

JackTopic starter

  • Ex Forum moderator
  • GENERAL OF THE ARMY
  • ******
  • Offline Offline
  • Posts: 14939
    • A user initiative by MSI fanatics to support all other MSI product users

Quote
do you really think that this increase of 100% ! will be really not noticeable ? ...

think about it.

To "think about it" is really the wrong approach here.  To test it in real-life application (not synthetic/theoretical benchmark) is the point of this thread, not what people think.
Logged

maxx_tools

  • PRIVATE E-2
  • **
  • Offline Offline
  • Posts: 2

To "think about it" is really the wrong approach here.  To test it in real-life application (not synthetic/theoretical benchmark) is the point of this thread, not what people think.

well...   ;-))

your answer shows me only one thing. i'm sorry.

this is my job, i work longer with machines, that are under really heavy memory load,
and write software for them as you are maybe old...

try it. get games, databases or other memory-based software...
let them run with 10g/byte per second and then with 20g...

look at my bench that i took, take a *good look... and you will see,
that the benchmark that i choosed is an multithreaded bench...
this exactly emulate a modern application with an huge memory usage.
Logged

Froggy Gremlin

  • GENERAL OF THE ARMY
  • ******
  • Offline Offline
  • Posts: 11111
  • Master Air OC'er

The 'point' is getting off track. Sandy Bridge seems to suffer from the lack of much benefit from faster RAM speeds over it's native DDR3 1333. Especially in gaming.

If you wish to debunk the findings, please open a new thread & enlighten us detailing the most current information in your knowlwdge base for P55, X58, P67, Z68, & X79. Many of us tend to believe 'real world' experience beats theory & how things should be on paper. It becomes a joke after awhile. How much CPU speed, RAM speed, or FPS are needed past a certain plateau? Applications, games, & software can't hardly keep up as it is.

Off topic, but 7 series chipset mainboards are now the latest thing, so let's see if they are more responsive to RAM speed & settings. :-))
Logged
The Last Hoorah! I Had Retired From All This! Can't Believe I'm Really Doing A Platform Test One More Time!  :shocked:

Z97 Gaming 7, 4790k, Zalman CNPS9900Max-R, Corsair AX-850, Corsair Dominator Platinum 2800's, WD 600 SATA III V.Raptor, Lite-On iHAS 424, GTX 680's In SLI, Creative X-Fi Fatal1ty Champion, Win 7 Ultimate 64 Bit    

Life Can Be Very Fulfilling Without Win 8/8.1, SSD's (Exception: Testing PCI-E M.2), RAID, Current AMD GPU's, G.Skill RAM, Hackers, Crackers, & Crypto Mining!

IXOYE

Bernhard

  • Support the MSI Red Rockets Team Members 나는 한국을 사랑
  • GENERAL
  • *****
  • Offline Offline
  • Posts: 6248
  • Read the user manual first. Properly

This sticky is around  Sandy Bridge with JEDEC standard DDR3-1333 that yield around 19-20GB/sec in practice and not a comparison of DDR3 in general that can have a range of 11 - 24GB/s.
We conducted these tests purely on Sandy Bridge platforms. The results were ~19-24GB/s in theoretical benchmarks, testing DDR3-1333 through to DDR3-2133. The real world performance is contained within the tests.
Nowhere was there any mention of DDR3 at 11GB/s on a different platform, nor an attempt at comparing that to the Sandy Bridge performance. We are not disputing that there may/will be a difference if one were to conduct these tests on differing platforms with differing mem controllers at the spread (11-22GB/s) that you now want to add to this thread. If a X48 chipset board with DDR3-1066 or 1333 were to provide 11GB/s then obviously there will be a difference.
This remains a sticky with tests on Sandy bridge only and you are welcome to compare your own real world results using that hardware.
Logged
1.B350F Gaming; R5-1600; Crucial MX300 275GB, Ballistix Sport LT BLS16G4D240FSC 2X8gb; Sapphire RX580 Pulse 8gb; Seasonic750W; Win10-64; LG DVD
2.FUJITSU LIFEBOOK S751 QM67;i5-2520M;2 X 4GB 1333Crucial CT256 MX100; Win10-64; Orico Dual Bay USB3 Dock



"veni vedi doctus"
MSI 미친

Aaron

  • GENERAL
  • *****
  • Offline Offline
  • Posts: 5690
  • BOOST!

BUT, not the maximum reached ones, or the average... but rather the MINIMUM reached FPS
and this much more interesting as the both other ones.

But by how much?  I would be very interested in seeing these statistics.  My opinion/guess is also that it would be of an insignificant or extremely minimal amount [1-2FPS maximum] increase.  Which once again, as per this thread, is pointless.  It's also a waste of money.

I have to admit that I'm still dumbfounded by the amount of people who cannot see beyond the scope of synthesised results and analysis.  The idea is what YOU can perceive as an increase, not the machine. 
Logged
Refrain from having to solve most of your problems by not creating them in the first place.

DDR3 Memory Standards: Blurry Lines
Integrated Memory Controllers (IMC)
Intel EIST and Processor C-states
Overclock Database 2.0

Hak Foo

  • SERGEANT
  • *****
  • Offline Offline
  • Posts: 23

One angle to consider though-- if the cost difference to go up one speed grade is trivial-- there could be a better 'binning' of parts sold for higher speeds

I recall there used to be a lot of weak DDR2-800 out there-- really intended to be DDR2-667 at reasonable voltages but sold as 800 if you run it at 2 volts or higher-- but the kits rated 1066 could reliably hit 800 at stock voltage

By getting 1600 you are really just ensuring it's overqualified to be reliable at 1333 :)

I figure the discussion is largely futile now-- as the cost difference is USD10 on 16Gb
Logged

neoroy

  • FIRST LIEUTENANT
  • ***
  • Offline Offline
  • Posts: 360

Wow very good information about RAM in this thread :) thx alot guys, especially Jack, HU16E & RemusM ^_^ ... Now  I think I better off from 2133mhz since it forcing me to use Vdimm higher at 1.645volt from default 1.50volt. Eventhough its quite stable 2133mhz 7-10-7-27 2T 1.645v for my RAM but I guess I'm gonna let it run at stock 1600mhz 6-8-6-24 1T 1.50volt or 1866mhz 7-9-7-27 1T 1.55volt :)
As far as I know performance memory in gaming from 1333mhz to 1600mhz may gain about 3 - 4FPS and 1600mhz to 1866mhz/2133mhz only gain 1 FPS or 2 FPS ^_^
Logged
Intel Xeon E5-2623 V3 @3.3GHz
MSI X99A SLI Krait Edition
GALAX GTX980Ti GOC LN2 @1500MHz stock volt
Gskill Ripjaws V 4x4G 2400MHz DDR4
Pioneer SSD APS-SL3N 240GB
Seagate HDD 1TB 7200rpm SATA3
Seagate HDD 500GB 7200rpm SATA3
DELL U2713HMT
CoolerMaster SilentPro M2 1000W Silver
Cooler Master Hyper 212X Turbo
Razer Mamba TE Chroma
Razer Goliathus Extended
SAMSUNG Bluray Combo SH-B123L
Windows 10 Pro 64bit
Pages: [1] 2  All   Go Up