PRO Z890-A WIFI - iGPU hardware video acceleration issue with RAM clocks above 5600MT/s

P

persiz154202d8

Member
Joined
Aug 22, 2024
Messages
69
Recently got a Pro Z890-A WIFI but any RAM clocks above 5600MT/s and NGU clock above 3000MHz breaks the hardware video acceleration.
For example that video drops a lot of frames on 4K, even on 1440p. At 8K is unwatchable and sometimes crashes.

Any voltages, timings or gears doesn't helps.
Tried by removing the old drivers with UDD and reinstalling the new ones but it didn't help either.
RAM clocks at or under 5600MT/s and NGU clock at or under 3000MHz works fine and the video is smooth with no frames drop at any resolutions.
Using the latest 7E32v2A73 BIOS, haven't tried with different ones.

The CPU is 265K, the RAM is CMK64GX5M2B6600C32
 
Last edited:
Sort by date Sort by votes
gert.riegle124a0299 said:
Could there be correlation that with XMP enabled, memory is showing as quad-channel in hwinfo?
Click to expand...

DDR5 splits the memory module into two independent 32-bit addressable subchannels to increase efficiency and lower the latencies of data accesses for the memory controller. The data width of the DDR5 module is still 64-bit. However, breaking it down into two 32-bit addressable channels increases overall performance, also see here.

It's possible that HWinfo went the CPU-Z route of naming things for DDR5 now (i can't check, i have DDR4). Because CPU-Z showed "4x 32-bit" for dual-channel DDR5 (since each DDR5 RAM module also has two internal channels, as mentioned), but since many people didn't know that fact, they have gone back to showing "2x 32-bit" to avoid confusion. Which of course adds confusion for those who got used to their "4x 32-bit" from before.

And before all that, for a while they showed "Quad channel" with DDR5, which was totally confusing of course. They want to be too exact for their own good, nobody would (or should) call it quad-channel on a dual-channel board. So it's possible HWinfo also does something like that now. But the weird thing is how it changed with and without XMP, while presumably nothing was physically changed by you, so this might simply be a bug in HWinfo then.

gert.riegle124a0299 said:
Only thing I would like to get XMP working without issues.
Click to expand...

It's best to open a ticket, only when MSI get enough tickets about a problem, there will be some movement on it.
 
Upvote 0 Downvote
gert.riegle124a0299 said:
Is there a "proper" fix for this already?
2x32GB Kingston Fury with XMP causes the same problem for me.

Could there be correlation that with XMP enabled, memory is showing as quad-channel in hwinfo?
Without XMP, dual-channel, iGPU working fine
265k on a Z890 Tomahawk Wifi

Everything running at default Bios settings as of now, which is stable and no issues.
Only thing I would like to get XMP working without issues.
Click to expand...
Did setting the SOC SA voltage to 0.900v solve the problem?
 
Upvote 0 Downvote
I did not try that yet, should this be the only thing that's needed until there is fixed Bios available?
Or you think it's more of a combination on your other undervolts?
Really don't know that's why I'm asking again, last OC things were back on 1155 for me
 
Upvote 0 Downvote
gert.riegle124a0299 said:
I did not try that yet, should this be the only thing that's needed until there is fixed Bios available?
Or you think it's more of a combination on your other undervolts?
Really don't know that's why I'm asking again, last OC things were back on 1155 for me
Click to expand...
Only adjusting the SOC SA should fix the problem with the video acceleration. MSI automatically raises the SOC SA voltage when you increase RAM or NGU clocks which apparently is unnecessary. I'm not sure how much the default SOC SA voltage is but my guess is about 1.020v. My tests confirmed anything below 1.020 fixes the problem. I've tested down to 0.800 and there was no instability or any issues whatsoever. Never tested below 0.800v. So in my opinion 0.900 should be perfectly fine but you can test by yourself.
 
Upvote 0 Downvote
Since NPU error 43 with XMP is also seen on different manufacturer boards and I think iGPU is related...
It makes me think not only MSI boards set wrong value.
Wondering if this is wrong documentation of Intel?

Majority probably uses dedicated GPU so a they won't even be directly affected.
 
Upvote 0 Downvote
Only response to my ticket was "try increasing tLC and tRCDWR"
After responding that it's fixed by loweing SOC SA, no answer...
Honestly idc anymore as long as this workaround works.
 
Upvote 0 Downvote
FYI: That's a known issue from older generation(12gen) where MSI already reported the issue to Intel.
And MSI reported the issue again for the current case in this topic with this new generations,
expecting a solution by Intel.
Will update this topic when solution/update is available.

Until wait as work around is set SOC SA voltage to below 1.1V.
Note that lowering SOC SA voltage might affect overall system stability if RAM is overclocked,
make sure to run full memory stability test after the tuning.
 
Upvote 0 Downvote
Svet said:
Note that lowering SOC SA voltage might affect overall system stability if RAM is overclocked
Click to expand...

Yes, and via firmware Intel always set that value to 1.1V so far.
Intel Core Ultra does not behave like Gen 14 here, so probably 0.90-1.00 might be the right value.
Again, all their calculations are for DDR5-6400, CL52 at 1.1V
No overclocking and overvolting (XMP) of any kind!
 
Upvote 0 Downvote
SOC SA on Arrow Lake is not the same as the SA voltage on Alder/Raptor lake and in my experience has nothing to do with the RAM clock.
VccSA on Arrow Lake is closer to the SA voltage on Alder/Raptor lake but it's still not quite the same.
Dual rank RAM at 7333MT/s and NGU at 3200 works absolutely stable with SOC SA at 0.800v
So automatically raising SOC SA on MSI doesn't make any sense. They could just leave it on Default which as I said I believe is 1.020v.
 
Upvote 0 Downvote
persiz154202d8 said:
VccSA on Arrow Lake is closer to the SA voltage on Alder/Raptor lake but it's still not quite the same.
Click to expand...

As I said before, that's the main problem with the Intel's firmware.
SOC SA is just another source of confusion. :biggrin:
The second problem is the different native memory speed:
DDR5-5600, CL46 at 1.1V for Gen 14
DDR5-6400, CL52 at 1.1V for Core Ultra.
All these things are interlinked and the CPU IMC is affected by all of them.
 
Upvote 0 Downvote
@Svet Any chance the effective SOC SA voltage could be displayed in a future BIOS update on the Z890 boards while we wait for Intel to fix the issue?

At least then people wouldn't have to guess its default value, and could potentially make more informed choices about where to set it to work around the hardware acceleration problem reported in this thread. On my Z890 Tomahawk WIFI with the .1A90 BIOS, the SOC SA Voltage Limit is reported as 1.1V at the "Auto" setting, which gives some indication of approx. where the actual voltage might be.
 
Upvote 0 Downvote
Back
Top