03-03-2025 12:29 PM - edited 03-07-2025 01:25 AM
This post is meant to compare BIOSes 1501, 1401, and 1302 for Asus Z890 Hero in terms of memory stability, on a probably faulty memory set which managed to turn out interesting results. This seems like an issue affecting a wide range of motherboards, click here for another post on ASUS forums, having the same problem but on a Z890 Extreme board. The memories that will be used in this test are: mainly G-Skill 8200mhz 48GB (x2 24GB) CUDIMM RGB memory set that is most likely faulty, and a G-Skill 6000mhz 64GB (x2 32GB) UDIMM memory set made in 2022, which has been running flawlessly since then. All on XMP I and II profiles without additional tuning. The test program is Memtest86 version 11.2.
I built my desktop PC having a ROG Z890 Hero with the latest BIOS 1401 installed, latest should be the most stable, right? This doesn't seem to be the case with the Z890 Hero in terms of memory stability, at least according to the tests I did until now. First of all, I couldn't get my 8200mhz CUDIMM memory to work flawlessly in any of the BIOS revisions and the memory kit will most likely turn out faulty after I return it tomorrow. I actually first waited for version 1501, where things became even worse. But after downgrading to version 1302, I interestingly got the most stable results. Please note that I will get an 8400mhz TeamGroup Xtreem 48GB CUDIMM Kit and x2 TeamGroup Xtreem 48GB UDIMM Kits running at 7200mhz and 7600mhz speeds in a week, and do further tests about these BIOS revisions as detailed as time allows me. I mean, I don't want to ring any bells yet, at least not fully.
MEMTEST86 RESULTS (Exact Same Settings in all BIOS revisions):
I did 5 tests in each screenshot you will see here. And I took the ones having the error numbers that are in the middle, kind of an average of all. That's why I couldn't complete 4 passes in all tests, they take a lot of time.
BIOS 1501 with XMP II (1 and a half pass only):
(26 errors in 1 pass)
BIOS 1501 with XMP I (1 pass only):Sorry for the crappy photos. I didn't know I would create this thread while taking these. Trying 1302 idea came to my mind here, so did creating this thread, so the next photos are all well taken at least.
(21 errors in 1 pass)
BIOS 1401 with XMP II (2 passes):
(2 errors in 2 passes)
BIOS 1401 with XMP I (1 pass):
(3 errors in 1 pass)
BIOS 1302 with XMP II (All 4 passes!):
(1 error in 4 passes, complete test)
BIOS 1302 with XMP I (All 4 passes!):
(3 errors in 4 passes, complete test)
As you can see, even though the memory kit seems faulty, it is most stable on BIOS 1302, goes more unstable with BIOS 1401, and much more unstable with BIOS 1501. Even so much that I got a few blue screens on 1501 whereas I only got a single blue screen on 1401 while testing all hardware for days.
But, if you have upgraded from version 1302 to 1401 or 1501, and using high speed but low capacity memory kits(48GB is considered low capacity, yes) on XMP Profiles, or if you are fine-tuning every single detail about your memory the old-but-trusted way, I suggest running your memory test programs once again, just to be safe. As you know, a system can "seem" to be working stably, but you mostly get the errors when you do a real test, especially for memories. And we desktop users generally find our stable settings in our current BIOS first, then do a full test on every piece of hardware, and keep on upgrading to new BIOS revisions without doing further tests. And if you started with 1401 or 1501, you might reach to higher speeds or lower latencies with 1302 after fine-tuning.
G-Skill 6000mhz 64GB (x2 32GB) on BIOS 1302:It passes no matter how many times I run, also passed the OCCT Memory Test I ran for a short time. I also tested everything on the system with this kit to make sure every other piece of hardware is working perfectly.
I asked myself, Why? The answer might be in the BIOS fixes list:
1401: Support up to 5600MT/s when running the maximum memory capacity of 256GB. The exclusive ""AEMP III"" option will appear when four Kingston RAM DDR5 6400MT/s 64GB CUDIMMs are installed.
1501: This BIOS enhances system performance and improves high-capacity memory compatibility. This includes support for 2*DIMM memory configurations, achieving speeds up to 7000MT/s with AEMP III enabled.
As you have realized, none of these updates are for us low-capacity, but high-speed memory users. I have to mention that both 1401 and 1501 were first released as Beta and the Beta tags were removed only in a few days. Also, BIOS 1401 was made in the holiday period, only 10 days after the release of 1302.
Trying Other Options: I tried Dimm Fit and it only made Memtest give more errors, no matter which option out of 4 I picked (These 4 options are not present in ASUS's guide as well -ASUS Please update the guide-). I couldn't try Dimm Flex as well since the memory presets for Dimm Flex present on Apex boards since Z790 are not present on my Z890 Hero (-ASUS please add these presets to Hero and other boards with new BIOSes, not just Apex, and Apex is "white" and doesn't have a backplate-). Apex also has the best memory stability and can achieve the highest memory speeds amongst all ASUS Z890 motherboards, you can check here. I added a timestamp but the video is worth watching fully.
I will see how the results turn out after I get my hands on the memory kits on the way. On CUDIMMs, thinking that I already returned a 8400mhz G-Skill CUDIMM (Why I am getting a brand new soon, it's a replacement), and will most likely get a replacement on this 8200mhz one too. I mean 2/2 fail here, I also heard a lot of returns happened with CUDIMM memories where I live. I will keep on updating the thread as I progress.
03-04-2025 12:20 AM - edited 03-04-2025 12:21 AM
Very interesting results. There seems to be still a long road ahead for BIOS development to get CuDIMM memory stable running at the advertised (XMP) frequencies at Z890 boards. May I suggest though to use additionally another test? MemTest86 is insuffcient to prove stability / fault free operations at a given frequency and timings. For what is matters, I get 2 x 24 GB GSkill CUDIMM 8800 running stable at 8600 with BIOS 1501 (tested with KARHU). No way to get it running faultfree at 8800 MHz, even with relaxed timings. I still wonder who is to blame? (Faulty) memory or a non-optimized platform firmware?
04-25-2025 10:45 PM
Running bios 1801.gskill8400 ram. Can’t get it stable with xmp1/2/tweaked. Did you manually adjust other settings such as voltages? its stable when i remove the xml profile but then runs at 5600 on on “auto”
04-26-2025 02:35 AM
XMP I is the easier to run profile, by the way. I meanwhile changed the CPU, and with the new one I can now run the memory at 8800. Unfortunately it seems most 200 series CPU are not good enough to run memory above 8000 MHz. That is very apparantly the reason why Intel limited the new S BIOS to 8000 MHz. With other words, you need a far above average IMC on your die to run CUDIMN 8800 at 8800. No BIOS will change the headroom.
04-26-2025 11:30 PM
Thanks for the heads up. It’s strange for sure. I have a 285k processor btw
04-27-2025 01:10 AM
Have you checked/rated the SP of your IMC? A score around 75 seems to be average for 2xx CPU.
04-27-2025 01:14 AM
Nope. How do I do that?
04-27-2025 01:21 AM
You go to "Tweakers Paradise". Then "SP MC". Check it, "Get SP". The system us then rebooting and training the memory controller for a while, and then rebooting. In BIOS you can see the SP MC score then.
04-27-2025 01:39 AM
84
04-27-2025 01:57 AM - edited 04-27-2025 01:59 AM
AFAIK, this is a score above average. I think your IMC is not the bottleneck.
BTW, what is the MC voltage setting now? It is related to the SP score. Once your score is etsbalished, the BIOS is adjusting the memort controller voltage. What do you see?