MPG X870E CARBON WIFI Beta BIOS

[Svet]

Spoiler: Older Versions

>>> E7E49AMSI.1A1E <<<

>>> E7E49AMSI.1A1H <<<

>>> E7E49AMSI.1A21 <<<

>>> E7E49AMSI.1A22 <<<

>>> E7E49AMSI.1A24 <<<

>>> E7E49AMSI.1A25 <<<

>>> E7E49AMSI.1A26 <<<

>>> E7E49AMSI.1A28 <<<

>>> E7E49AMSI.1A29 <<<

>>> E7E49AMSI.1A31 <<<

>>> E7E49AMSI.1A32 <<<

>>> E7E49AMSI.1A33 <<<

 

[gael.blottier15a102de]

I wanted to ask if you've tried the drive itself in the top slot, and on another board to see if it still disconnects. It could be the board has an issue specifically with the crucial T705, but it would be good to eliminate the possibility that that specific drive has problems, and is properly cooled. Are your forcing Gen4 (instead of Auto) in the BIOS for those slots?

No didn't try the drive in the top slot, but it's a crucial T705 as well (I'm feeling lazy about installing a system again). I'm watching the temperature almost all the time. Since the new bios, full, auto for now the M2_1 reached 68°C top while playing a game streaming a lot of assets (was 65°C before the new bios but can't be sure it's related, or they did something in the game...) anyway it's below the warning thresholds.

M2_3 is often about 10 degrees less when in use, about 45 degrees when nothing is happening.

I'm currently trying a samsung 990 pro (gen4) on the M2_3, for now it's stable (but a bit early to really tell) (temperature is below the crucials' and stay below 56°C for now even writing on it).

Will try different parameters when I have time.

PS: I forgot, I lost the M2_1 after going to the bios and saving (no change), it happens quite often, but a shutdown, waiting for about 10 seconds, and start solve the issue.
PS2: half an hour later, lost he crucial T705 which was now on M2_4 doing.. nothing with it, it just silently disconnected

 

[Zippos]

Dang, good idea. Several of the USB ports are through both chipsets, i winder if that might eliminate some of the other USB issues folks are having?
Where specifically is the chipset voltage setting in the BIOS?
Have you noticed any heat issues with the increased voltage? The second chipset already runs fairly hot according to HWInfo, and there's no built-in active cooling.

I think it is at the bottom of the Overclocking section. You can also find it using the search functionality in the BIOS.

No problems with the heat. Chipset A and B temps in HWinfo hover around 60 deg C.

 

[capstan_siding_7152702d7]

Thanks

 

[Alyred]

I think it is at the bottom of the Overclocking section. You can also find it using the search functionality in the BIOS.

No problems with the heat. Chipset A and B temps in HWinfo hover around 60 deg C.

I'll take a look for it. Interesting about the temperatures, while my Chipset 1 hovers around 53 degrees with spikes up to 65, chipset 2 runs around 64 degrees and spikes up to 76. Averages hover in the 30's/40's at complete idle, but always did look at the second chipset with a little bit of sideeye.

 

[Alyred]

No didn't try the drive in the top slot, but it's a crucial T705 as well (I'm feeling lazy about installing a system again). I'm watching the temperature almost all the time. Since the new bios, full, auto for now the M2_1 reached 68°C top while playing a game streaming a lot of assets (was 65°C before the new bios but can't be sure it's related, or they did something in the game...) anyway it's below the warning thresholds.

M2_3 is often about 10 degrees less when in use, about 45 degrees when nothing is happening.

I'm currently trying a samsung 990 pro (gen4) on the M2_3, for now it's stable (but a bit early to really tell) (temperature is below the crucials' and stay below 56°C for now even writing on it).

Will try different parameters when I have time.

PS: I forgot, I lost the M2_1 after going to the bios and saving (no change), it happens quite often, but a shutdown, waiting for about 10 seconds, and start solve the issue.
PS2: half an hour later, lost he crucial T705 which was now on M2_4 doing.. nothing with it, it just silently disconnected

Temps look acceptable, though 68 is getting up there a little bit, but Crucial sates a temp tolerance up to 80 degrees before it'll throttle so you should be good. It's normal for the slot above the graphics card to get a bit warmer since there is a lot of heat coming off the card.
If your main drive is a T705 and is functioning well in the M2_1 slot, it leads a bit more evidence that the other T705 drive is having problems itself. I'm running Samsung 990 Pros in both slots myself, and aside from lower random IOPS in M2_1 than there should be (low by about 10%-ish), they're stable.
If you'd be willing to test, you should be able to put it in the M2_3 slot and still boot to it (just running at PCIev4 speeds, of course). If it's the slot you'll likely have full system crashes as the drive disconnects, however, if it's stable and your other T705 in Slot M2_1 disconnects, I'd definitely more suspect the drive itself. If you have another computer to test it in that would be the final test to see if the disconnects/issues follow.

 

[Alyred]

I can report on the A26 beta on one of my machines.
It seems to regularly restart once or twice while booting, dropping back to code 15 and re-training the RAM. I notice this when coming out of BIOS and having some savings set, it always seems to run through RAM training twice before it boots all the way. Once done, it's fine.
M2_1 Random IOPS is better than it was, but still not up to the speeds expected on the Samsung 990 Pro. Gained an average of 50k IOPS over A25, across 3 runs: 1.079M, 1.09M, and 1.153M IOPS, Same drive running on M2_3 gets a steady 1.45M IOPS.
Noticed there's now a setting to set slot E1 to Gen 5, so those lucky enough to get a 50 series Graphics card can run it ... assuming it doesn't burst into flames due to way too many amps being pulled over too few wires...
Otherwise, manually copied my CO, boost, and memory settings over from A25 and they're still showing just as stable after a full 3 hours OCCT Extreme CPU+MEM AVX512 run.
I'll likely update it on my daily driver workstation soon and see how it performs in games.

 

[Alyred]

Did have a core throw an error during extended testing, so backed off the curve on that core a little bit and went fine. Memory threw an error too so will back off the aggressiveness of that timing slightly as well and give it another long test.
Just goes to show that it's always necessary to check those extreme timings on new BIOSes.
Edit: the memory error looks like the mainboard dropped the A2 dimm completely under stress load. Never have seen this happen before on the other BIOS versions before A26, but maybe it was a one off. Windows loaded after a soft reboot so I didn't notice it at first. Had to completely shut down the computer and then power back on, didn't touch the RAM hardware. The BIOS said memory settings had changed and loaded both DIMMs. These are the non-binary 48gb Corsair sticks on this test bench. Will report back how the new tests run.

 

[vic123159d02e5]

Did have a core throw an error during extended testing, so backed off the curve on that core a little bit and went fine. Memory threw an error too so will back off the aggressiveness of that timing slightly as well and give it another long test.
Just goes to show that it's always necessary to check those extreme timings on new BIOSes.

I don’t understand you had an error with one core whereas it was stable with A25 ?

 

[Alyred]

I don’t understand you had an error with one core whereas it was stable with A25 ?

Yeah, with adjustments to the BIOS code, timings and stability can change.
Sure is taking a while to get this done, though. I'm lucky enough that my wife doesn't  need her new computer right away so I can run tests and beta BIOSes on it before trying things on my own, which is identical except for the RAM and case, but it sure is a drain on time. Never had this much trouble building a computer before, and I've built hundreds over my career.

 

[dlomond]

.26 works no problem

 

[Mr.Fox]

Yeah, with adjustments to the BIOS code, timings and stability can change.
Sure is taking a while to get this done, though. I'm lucky enough that my wife doesn't  need her new computer right away so I can run tests and beta BIOSes on it before trying things on my own, which is identical except for the RAM and case, but it sure is a drain on time. Never had this much trouble building a computer before, and I've built hundreds over my career.

Releasing half-baked and defective products and maybe fixing them later on is the new status quo. R&D and QC is a thing from the past. Consumers pay for the privilege of being a crash test dummy.

 

[RobsonByte]

Did have a core throw an error during extended testing, so backed off the curve on that core a little bit and went fine. Memory threw an error too so will back off the aggressiveness of that timing slightly as well and give it another long test.
Just goes to show that it's always necessary to check those extreme timings on new BIOSes.
Edit: the memory error looks like the mainboard dropped the A2 dimm completely under stress load. Never have seen this happen before on the other BIOS versions before A26, but maybe it was a one off. Windows loaded after a soft reboot so I didn't notice it at first. Had to completely shut down the computer and then power back on, didn't touch the RAM hardware. The BIOS said memory settings had changed and loaded both DIMMs. These are the non-binary 48gb Corsair sticks on this test bench. Will report back how the new tests run.

How did you come to the conclusion that the A2 dimm socket is falling out? In BIOS version A25 I noticed that sometimes I can't move the mouse or use the keyboard, so I have to turn off the power. I tried changing Nitro RAM settings, it improved a bit. Then I'll test BIOS version A26 to see if it fixes this. I'm in doubt if these freezes are related to RAM configurations.

 

[vic123159d02e5]

Yeah, with adjustments to the BIOS code, timings and stability can change.
Sure is taking a while to get this done, though. I'm lucky enough that my wife doesn't  need her new computer right away so I can run tests and beta BIOSes on it before trying things on my own, which is identical except for the RAM and case, but it sure is a drain on time. Never had this much trouble building a computer before, and I've built hundreds over my career.

That's what I was afraid of, so I'll pass on the A26 and stay with A25.
I also do tons of tests and spend a large amount of time trying to squeeze out every last drop of performance.
So now that I've finally found a stable settings at ddr5 8320mhz cl34 and 104 FSB with PBO +200 and curve -36, that would pass Y-Cruncher, Cinebench r23, Aida and OCCT ...I won't do it again.
In terms of performance, in benchmark like cpu-z or anything like this did you saw an improvement or not ?
(I'm asking this because my Cpu-Z numbers were a little better at same speed on the A23)

 

[RobsonByte]

We could be paid for the time spent doing the tests. I've tested all versions up to A25 and never managed to perform with stability. Now I'm trying to figure out why my mouse and keyboard stop randomly and the computer keeps running (A25 version). It would also be helpful to let you know what changes are made in the released versions.

 

[Alyred]

That's what I was afraid of, so I'll pass on the A26 and stay with A25.
I also do tons of tests and spend a large amount of time trying to squeeze out every last drop of performance.
So now that I've finally found a stable settings at ddr5 8320mhz cl34 and 104 FSB with PBO +200 and curve -36, that would pass Y-Cruncher, Cinebench r23, Aida and OCCT ...I won't do it again.
In terms of performance, in benchmark like cpu-z or anything like this did you saw an improvement or not ?
(I'm asking this because my Cpu-Z numbers were a little better at same speed on the A23)

Yeah, your numbers are insane - I have no idea how you were able to push it that far and keep it stable through those tests. What are your RAM latency numbers? Or are you mostly looking for raw throughput? I thought having to set it to a MEMCLK to 1:2 introduced latency that didn't make it worth running those speeds?
So far as performance on A26, I'm unable to see any difference but any benchmarks are currently synthetic on that workstation and just being used to run testing on, no daily usage or games as of yet. The only immediate difference I saw in A26 were the options to set PCIe_1 to Gen 5.
Stability difference was minor. I saw an instability at -25 on one core that I had to drop to -20 before it would go away. If you're able to maintain -36 Curve all-core I'd say you have a freakin' S-tier chip there...

 

[Alyred]

How did you come to the conclusion that the A2 dimm socket is falling out? In BIOS version A25 I noticed that sometimes I can't move the mouse or use the keyboard, so I have to turn off the power. I tried changing Nitro RAM settings, it improved a bit. Then I'll test BIOS version A26 to see if it fixes this. I'm in doubt if these freezes are related to RAM configurations.

When I rebooted after the memory error during the test cycle, it mentioned the memory size had changed. I didn't take notice of it at the time, but when I had gone into windows after saving the current configuration with slightly looser timings, I was missing half my RAM and then noticed Corsair iCue wasn't able to see the stick in A2, and while the profile was loaded the hardware profile for color was still on for that DIMM. Standard soft-reboot didn't get it back, but powering it down and back on did.
With the slightly looser timings (moved TRCDWR,tRCDRD abd tRP from 35 to 36, tRC from 68 to 70, and tRFC1 from 500 to 525) it was then able to pass a 6.5 hour stability test in OCCT. Unsure if it was just a fluke that the memory dropped that one time, or if the extreme settings pushed the new A26 BIOS to drop the DIMM, or what. May re-up those timings and try it again.

 

[vic123159d02e5]

Yeah, your numbers are insane - I have no idea how you were able to push it that far and keep it stable through those tests. What are your RAM latency numbers? Or are you mostly looking for raw throughput? I thought having to set it to a MEMCLK to 1:2 introduced latency that didn't make it worth running those speeds?
So far as performance on A26, I'm unable to see any difference but any benchmarks are currently synthetic on that workstation and just being used to run testing on, no daily usage or games as of yet. The only immediate difference I saw in A26 were the options to set PCIe_1 to Gen 5.
Stability difference was minor. I saw an instability at -25 on one core that I had to drop to -20 before it would go away. If you're able to maintain -36 Curve all-core I'd say you have a freakin' S-tier chip there...

MEMCLK 1:2 is catching up to 1:1 at around 8000 MHz vs. 6000 MHz, at least according to several experienced people like this guy that is a memory overclocking specialist :
https://www.youtube.com/c/ActuallyHardcoreOverclocking
In my real-world usage (mostly sim racing titles), running 6192 MHz at 1:1 CL30 or 8320 MHz at CL34 delivers similar performance, maybe slightly higher FPS at 6xxx, but better 1% and 0.2% lows at 8xxx. (note that for these timings I set 1.6v on the ram)
I had a -34 curve on A23 and could push -36 on A25. There are definitely better chips than mine, I can assure you. Also, to be fair, in single-core it goes all the way up to 5.642 GHz, but under full all-core load, it fluctuates between 5.585 and 5.612 GHz. On https://www.overclock.net, there are plenty of people with much better curves than mine, -45 is not uncommon there.
When I tried to find a per-core curve, I discovered that the chip had two 'bad' cores that wouldn't accept -45. However, it makes no difference whether I set two cores to -36 and six cores to -45, the chip behaves as if it were at -36 across all cores

 

[RobsonByte]

When I rebooted after the memory error during the test cycle, it mentioned the memory size had changed. I didn't take notice of it at the time, but when I had gone into windows after saving the current configuration with slightly looser timings, I was missing half my RAM and then noticed Corsair iCue wasn't able to see the stick in A2, and while the profile was loaded the hardware profile for color was still on for that DIMM. Standard soft-reboot didn't get it back, but powering it down and back on did.
With the slightly looser timings (moved TRCDWR,tRCDRD abd tRP from 35 to 36, tRC from 68 to 70, and tRFC1 from 500 to 525) it was then able to pass a 6.5 hour stability test in OCCT. Unsure if it was just a fluke that the memory dropped that one time, or if the extreme settings pushed the new A26 BIOS to drop the DIMM, or what. May re-up those timings and try it again.

The feeling I have is that every time the computer is turned on we have a surprise, I don't know if it has to do with the training of the DRAM that is trained differently at each boot start. After installing the a26 BIOS, I noticed that I didn't have a "d0" error yet.

 

[Alyred]

Think I'll be rolling back to A25 on my test bench. Even at the "loosened" timings I set, I'm still randomly getting test failures on memory. I can pass an 8 hour test one boot, but the next will fail after 35 minutes. There's also some USB issues that are cropping back in: one time I had my keyboard get "stuck" repeating a "1" while signing in, but there wasn't any physically stuck key. Only unplugging and re-plugging the keyboard seemed to fix it. I've also lost my Corsair loop several times under load, where the hardware stops responding/reporting but the software thinks it's still fine. This happened often under A21 as well, but on A25 I'm not seeing any of that. Since there's no AGESA update in A26 that I'm aware of, and the only major changes are the ability to set Gen 5 on PCIe_1 (and who knows low long it'll be before I can get a 50 series card), I think rolling back and re-entering the settings I had before is probably the best bet for now.

 

[Alyred]

MEMCLK 1:2 is catching up to 1:1 at around 8000 MHz vs. 6000 MHz, at least according to several experienced people like this guy that is a memory overclocking specialist :
https://www.youtube.com/c/ActuallyHardcoreOverclocking
In my real-world usage (mostly sim racing titles), running 6192 MHz at 1:1 CL30 or 8320 MHz at CL34 delivers similar performance, maybe slightly higher FPS at 6xxx, but better 1% and 0.2% lows at 8xxx. (note that for these timings I set 1.6v on the ram)
I had a -34 curve on A23 and could push -36 on A25. There are definitely better chips than mine, I can assure you. Also, to be fair, in single-core it goes all the way up to 5.642 GHz, but under full all-core load, it fluctuates between 5.585 and 5.612 GHz. On https://www.overclock.net, there are plenty of people with much better curves than mine, -45 is not uncommon there.
When I tried to find a per-core curve, I discovered that the chip had two 'bad' cores that wouldn't accept -45. However, it makes no difference whether I set two cores to -36 and six cores to -45, the chip behaves as if it were at -36 across all cores

Interesting to know. I'm not THAT hardcore into overclocking, just like to get performance that I paid for at a reasonable heat level. But it does sound that if the motherboard treats all cores at the least core's CO, that still sounds like a BIOS issue to me... otherwise, why even have a per-core setting?