MSI Z68A-GD80(G3) and Corsair's New RAM

[HU16E]

Yeah, ok, it all looks great, but how do they work together outside of a test laboratory in a real world home built pc?

That question was answered very quickly, without any special fiddling about required beyond normal setting changes.

When the Corsair kit was installed, the usual 'Memory Has Changed' screen popped up. No big deal, a simple check in the UEFI/BIOS looked like everything defaulted to ensure a successful post and bootup. Yup, that's exactly what took place. Corsair took the right approach IMO to set the profile to default to DDR3 1333. At the next restart, it was a simple matter to set the manufacturer's recommended rated speed of 2133, CR 1, 9 11 10 27, tRFC 171 on Auto DRAM V. Posted and booted right up.

Yeah, ok, how about if 8GB isn't enough RAM for me?

Figured that might be a question asked. Here is the finding. To make it more real, often pc builders buy components at different times. This can be problematic, but that is best discussed on a case by case basis. Back to the answer. Purchased on purpose a kit from a different batch, even the serial numbers were way apart. After they were installed, the usual 'Mem. Change' screen again, but at F10 - Enter, it posted and booted right up! No increase of DRAM V. was required. In fact, with it set to Auto, the 'Read Only' actually shows the voltage to be 1.472v's (by meter, it was just below 1.50v's).

Note that CPU-Z X.M.P. for this RAM shows wrong. In the UEFI/BIOS, Memory-Z, it shows correctly.

Conclusion? I would say MSI mainboards and this RAM are going to work out just fine together.

 

[xmad]

I still not sure that all of the new vengence and dominator don't have the same memory. I don't have any dominator to compare with my vengence system to system, but my stats are very much in line with yours.
Running currently at 9-10-10-21 / 1t
Dram v: 1.55 (works at 1.5, this is a habit /shrug)
I also have the tRFC at 91, although I haven't had a chance to really play with it since I stress tested it like this.

Perhaps I have a lucky set  or more likely they are putting the better memory chips in all thier vengence & dominator kits now. I will check the version to verify this in a bit. Doesn't take away from how pretty those sticks are 

Edit: I send Corsair pre-sales and tech support an email asking into which lines are they installing the Samsung chips. IC V 4.13.

 

[doubleohseven]

Thanks for the info on the chips guys!

As for the "sweet spot," raising the tRFC seemed to make little/no difference until it got to about 90, where I saw a decrease in performance.

As for tRAS, raising it at all gave me a decrease in performance. Now, I had read somewhere that tRAS should be CL+tRCD+2, to give enough time for the memory to cycle. Thats why I chose tRAS 21 (9+10+2). In seeing the performance drop from raising it, I decided to try to tighten it some more, and was able to get it down to 19 stable.

Out of 10 tests, the best for tRAS 21:

Out of 5 tests, the best for tRAS 19:

So now i'm thinking that whole simple math "equation" is BS, since it is 30 IBT Max tests stable, and a better performer at 19. Has anyone heard of this before?

 

[xmad]

Hehe, being that all of these are 1333 chip running so far out of spec that it isn't even in site I guess it's all on the table.  :lol_anim:

At 2133, memory is supposed to be clocked at 11-11-11-30. Check out the JEDEC table I have in that acticle.
https://forum-en.msi.com/index.php?topic=161322.0

This is what overclocking/ undertiming is all about 

 

[xmad]

These tests are fairly accurate, I just got new virus protection that caused the 114's on the burn test. Man do I hate Nortons, but we get it through Comcast for free so c'est la vie...

PS: Those extra threads sure make the test come out nice Froggy 

 

[Bernhard]

That formula may be a good rule of thumb, or may have worked with DDR and DDR2, but looking at JEDEC standard 1333, that would already be out. 9.9.9.24 is the standard. that would mean the 24 should be 20, which it's not. Maybe for OC'ing it may be an indicator of how far you can tighten them 
These are all tweaks that probably cannot be realised anywhere else but in benchmarks and tests as performed here.   
But its nice to compare similar or the same memory on similar platforms and see how close the performance figures are, yet highlight also how they differ, even at the same settings. Good to see for newbie OC'ers that even at the same settings, no two rigs will ever be the same.

 

[HU16E]

Wanted to thank all for contributing to this thread. I have most definately learned some things.

 

[doubleohseven]

That formula may be a good rule of thumb, or may have worked with DDR and DDR2, but looking at JEDEC standard 1333, that would already be out. 9.9.9.24 is the standard. that would mean the 24 should be 20, which it's not. Maybe for OC'ing it may be an indicator of how far you can tighten them 
These are all tweaks that probably cannot be realised anywhere else but in benchmarks and tests as performed here.   
But its nice to compare similar or the same memory on similar platforms and see how close the performance figures are, yet highlight also how they differ, even at the same settings. Good to see for newbie OC'ers that even at the same settings, no two rigs will ever be the same.

That makes sense, since I tried to look the article up again (from Mushkin actually), and the article appears to be so old, they took it down. :lol_anim:

It is quite surprising how close our bench results really are! It is also interesting how different systems with the same memory react to timings. Learned some good things in this thread, and once again, glad I made the purchase!

Wanted to thank all for contributing to this thread. I have most definately learned some things.

 

[Bernhard]

I think enthusiast tests like this add value in terms of being able to validate one's own obervations and results.
Once I have the new CPU installed I shall run a couple more tests to see how different the memory results may be influenced by a native 1600 IMC. Using the same platform, and same memory, it will be interesting to see if settings that are not achievable now may just be achievable.
I'll be back 

 

[doubleohseven]

I think enthusiast tests like this add value in terms of being able to validate one's own obervations and results.
Once I have the new CPU installed I shall run a couple more tests to see how different the memory results may be influenced by a native 1600 IMC. Using the same platform, and same memory, it will be interesting to see if settings that are not achievable now may just be achievable.
I'll be back 

Looking forward to it!

 

[Bernhard]

With the 3770K, surprise surprise, the maxxmem figures don't really change much, although I can boot with 1T, no problem, and it does improve the latency over the 2T settings.
I ran two tests at differing timings. The 7ZIP numbers are attributable to 8 threads, and are only slightly influenced by the timings. The mem allows me to run at 2200 as well, but there was no improvement. Suppose I'll do some more tests, but I don't think it will show any further improvements of significance.

     

 

[HU16E]

The brand new 32GB 4X8GB 2133 1.5V Platinum kits are now available. A kit is on the way and will be testing them in the near future on many MSI mainboard models other than just the Z68 GD80(G3).

 

[xmad]

Cool! let us know.

Be really cool if you could actually get it to run at 2133 on the dual channel machines.

 

[HU16E]

The testing is complete on the new 32GB 4X8GB 2133 Platinum kit. In order to stay with the topic as for mainboard used, these results are with the Z68 GD80 and a Sandy CPU. No major problems were encountered. The usual 'Memory Has Changed' warning as expected. For initial testing, DDR3 1333 was the 'default' for just 2 sticks. All was normal, so added the other 2 sticks from the kit. Everything was normal, so restarted and set the RAM as per the rated specs. To get all 32GB to run at 2133, I did have to raise the DRAM V. to 1.568v's (verified by DMM), which is within JEDEC specs for DDR3 RAM of 1.50v's + or - 0.075v's.

There is a pending matter of contacting Corsair though for clarification of their published information and a vendor's information (Newegg). The RAM label on the sticks physically shows the DRAM V. as 1.65v's! The new CMD64GX3M8A2133C9 64GB 8x8GB kit also has the same Tech Specs as the 32GB kit of 9 11 11 31 @ 1.50v's. Also worth noting is the Tech Specs 'Memory Configuration' for both 32GB and 64GB kits read Dual Channel rather than Quad Channel.

7-Zip 16GB @ DDR3 1333;

7-Zip 16GB @ 2133;

7-Zip 32GB @ 2133;

Z68A-GD80(G3), 2700k @ 4.5, 32GB @ 2133;

If any problems are encountered with this RAM on other MSI platforms such as Z77 or X79, I will follow up and report it.

 

[HU16E]

No problems to report, but wanted to confirm that '0' issues have been encountered with any current 6 or 7 series chipset MSI mainboards tested and these new Corsair RAM offerings. Just for fun, I ran a quicky test with the X79 BB X II with this 32GB 2133 Platinum kit and a CPU speed of 45X. After a little UEFI/BIOS problem was sorted out, bench results are very impressive even with a CR of 2.

   

 

[neoroy]

Is it unstable in IBT, Froggy gremlin? ... +007 looks like unstable, should it be ... -002?

 

[HU16E]

No BSOD's, restarts, or other problems when gaming, or when running apps and programs. Been running a couple weeks, so satisfied all is well.

 

[neoroy]

No BSOD's, restarts, or other problems when gaming, or when running apps and programs. Been running a couple weeks, so satisfied all is well.

OK then it quite stable, I guess ... but can you do re-test IBT again at least 2 or 3 loops?

 

[HU16E]

After almost 3 weeks running in the X79 with '0' problems, the kit has since been removed and found a permanent home in the topic's test Z68 GD80(G3). The X79 BB X II now has a different Corsair RAM kit in it running 2400, the platforms upper limit.

 

[HU16E]

Sorry about the lost images. PB account was renamed and restructured with the resulting loss of images posted here. I will see if an admin or mod can fix it once I compile all the right images for the right spots again. No longer having edit authority, I can't fix it myself.

 

[doubleohseven]

I can also report that i've had zero issues with this kit at well! Loving this kit of RAM!