Probably not the first defective iGPU they’ll have seen, all told. Kinda makes me wonder if it’s a more common problem that just goes unnoticed due to relatively low use rates.
No from the GPU. I have my 4k reconnected for a few days and I have not had any freezes so far.
Puget is misinformed. CPB is AMD’s name for their light-load clock boost behavior. Running with CPB off is equivalent to locking an Intel chip to base clock. CPB on is 100% in-spec, in warranty, stock operation, and not an overclock.
Puget’s benchmark results are what they are because P ∝ f^3 and marginal IPC goes down with frequency because memory latency and bandwidth don’t scale. They discovered the open secret that x86 desktop chips are clocked well past the point of diminshing returns even at stock.
I really wish they’d add a correction at the top of that article. I have to reply to people citing it like twice a month.
(Sorry about the ridiculous delete/undelete/edit spam.)
And thats why I miss original anandtech so much, they would have would have discovered and published something like this already. RIP
So the last “guarantee” of stable by default freq/v table from vendor no longer exists? Great, thats just great.
I have exactly the same problem with my 7950X, it worked without problems with Linux and Windows for over a year until I used the iGPU as a passthrough GPU with Proxmox.
Currently the IGPU is no longer recognized by the AMD driver under Windows, unless the iGPU clock is fixed via the BIOS, but like you I have to disable the hardware acceleration of Firefox.
I have switched to the 9950x and will look at the problem with the 7950x again this week in another board
I’ve had this issue with my 7800X3D as well, issue is the iGPU firmware or something being unstable. However after normally 1-8 GPU fails! it will come good. It’s a VERY weird sort of HW issue because of that.
I was thinking of getting the next gen X3D sometime (9000 is it?) but I suspect the price will be too much… wait until a special.
This is pretty interesting, and even more so that the iGPU was the cause.
Ran into a similar problem with my sister’s build a few years ago (Ryzen 5 5600G). The problems presented as driver crashes while gaming. The memory also passed the MemTest, but it wasn’t until I replaced the memory that the crashes resolved.
If I had to guess, using main memory for VRAM imposes a sort of load that MemTest doesn’t appear to account for.
Hi All,
I currently have what I think is the same problem with the 9750x.
Asus b650 E-F Mobo with current non beta firmware. I also downloaded all the drivers including the chipset driver from Asus. Asus drivers seem to be just rolling out.
Pushing 5 monitors total, i need to use the iGPU to support one monitor as a 2080TI can only push 4 at once.
I’d like to assume at this point this is a bios/software thing and not defective CPU. If you could please keep posting trouble shooting attempts here I’d appreciate it. I will do the same.
Similar, and more problems (not just gpu) with my 7950x. I get garbled text / output similar to below. This can span the entire screen / multiple screens, etc.
This happens across distros (Fedora,Arch,Nix) desktop environments (sway,gnome, kde, cosmic, i3) and windows. Cables/displays HDMI, DisplayPort, using the USB-C alt dp on my motherboard causes the amdgpu
module to crash. For reasons I’ve replicated this across memory modules, motherboards, etc. i.e. only commonality was the cpu. This garbling spans even sunshine/moonshine, or video encoded remote desktop streams.
You get these graphical artifacts when using iGPU, right?
I’m done with troubleshooting since I bought a 2nd CPU (7600x) to confirm that the 1st one (7950x) was faulty and 7600x works perfectly fine for 2+ weeks now. I’ve sent 7950x back and waiting for their feedback. Hopefully they will reproduce this on their end and replace my CPU.
When I received 7600x and inserted it worked fine with no issues. After this I inserted 7950x back into the socket and glitches/instabilities returned immediately. When I switched it to 7600x again - issues were gone. In my case it was definitely a faulty CPU.
I get the impression that issues with AMD’s iGPU are actually somewhat widespread and the reason why it’s not making as much noise as it should is because not that many people use iGPU. In my case the CPU worked fine as long as I don’t use iGPU.
Hmm, I got this garbled text issue at some point too, but I thought it was a software issue… Lately it is gone again though…
At some point also I got regular “amdgpu ring timeout” faults in dmesg on linux which crashed the DE on linux. I struggled with this issue for weeks until I power cycled my PSU, and then it was gone again. No idea why…
it actually doesn’t confirm anything since its a totally different chip
Yeah, those glitches are iGPU only.
Similar with the amdgpu ring issue for me my issues are much wider than iGPU. Full list as curious if other people are encountering similar
AMD Random Reset verified the model matches in the kernel patch. While I’m “stable” now I have noticed a slight performance regression
BiFurcation issues, primary motherboard is the gigabyte b650i itx board. With the 7950x I get a hard shutoff on heavy load with bifurcation to 2 gpus (x8 / x8) via vfio, or gpu and nvme (x8/x4/x4). I know splitters et. al are finicky. But this worked on a X570, and Epyc Rome build. This was “working” at one point.
Around the time of the voltage kerfufle is when the cpu started getting wonky. I am negative curve PBO, pre voltage bios fixes all core -18, post patch -10. As part of the testing, I’ve also removed PBO.
Very memory sensitive. I’ve tried a number of kits from the motherboard QVL and they won’t post. My kit is a basic corsair 5200 32GBx2. I’ve found upgrading is impossible.
I’ve tested with new motherboard, psu, ram, etc. The only other ram I got to post was another corsair 16x2 at 5200. The only thing I didn’t swap was the cpu, as I didn’t see the value in investing in AM5 anymore.
As a side I’ve also encountered the garbled text on the 7840u/780M. I thought it was some plugin or font I was using, but I tried stock fedora and ubuntu for several weeks and still encountered it. Across a P14s and several framework motherboards. Internal display, external, various docks, and testing several monitors. I’ve seen similar comments from others encountering with mobile iGPU as well.
It confirms something was wrong with the first CPU, no?
Yes, that’s my thinking as well. If a different CPU (7600x being from the same line of CPUs and with the same iGPU) works fine but 7950x has problems then I think it’s safe to conclude that 7950x is faulty (or at least the iGPU part is).
If after replacing the CPU I still had problems then I would start to look in the direction of RAM and Motherboard but since this RAM and Motherboard works fine with another CPU then it is the CPU causing trouble.
It sounds like your CPU might be the rootcause.
It’s worth adding that even though I was planning to isolate my dedicated GPUs from the Linux host in order to pass it to a QEMU guest VM along with one of the USB controllers. And then use Sunshine/Moonlight to remotely controll the host OS, so that’s why I needed a working iGPU.
But before I got into any of that my system didn’t “pass” a stability “test” with my 1st CPU (7950x) and iGPU started glitching a few seconds after logging into operating system (both Linux and Windows installed on bare metal, no VMs or PCI isolation yet).
Right now as I’m temporarily using 7600x in that exact system I’m able to do all that. AMD RX 6600 and Nvidia RTX 4070 isolated from the Linux Mint host and passed-through to QEMU VMs. Sunshine running on the host (as well as the guests VMs) so I’m able to access those and control them remotely. This setup works well (2+ weeks in) and I was able to transition my existing VMs from my old (Intel 8700K / AMD WX 7100 + GTX 970) system to the new one with little to no effort.
It’s kind of amazing actually to be back in my working VM after just a few minutes and continue my work in the same environment exactly where I left it even though the underlying hardware is completely different. All I had to do was to just copy qcow files and call virsh define vm.xml. Great stuff!
@animusnull I was also worried that maybe my RAM (Kingston 6000 (KF560C30BBEAK2-64)) was the rootcause of my instabilities (even though I also got it from motherboard’s QVL list) or that maybe my motherboard was messing with the memory (since for some years now Motherboard manufacturers tend to make “safe default” not so safe and default just to score better in reviewer’s charts) but since this RAM at default settings works fine with my 2nd try at AMD (7600x) which has the same RAM requirements I’ve concluded that the RAM and Motherboard are most likely fine in my case.
When investigating this I’ve found that setting “Core Performance Boost” to Disabled in the BIOS and changing GPU frequency setting from Auto to 2200 which is what CPU specs declare (idea of this guy on AMD forums). make iGPU way more stable. I wonder if doing this will make your iGPU artifacts go away? (note that this will affect your CPU’s overall performance since it basically disables frequency boosting and on your Gigabyte motherboard “Core Performance Boost” might have a different name)
HI Just wanted to follow up here with my issue. My issue of the iGPU hanging has been fully resolved. It was a memory config issue on my b560E-F. I trusted one of the Auto settings and I shouldn’t have, I’m new with Asus mobos and usually build with intel.
When i reset the mobo to optimized settings then manually timed the memory to the advertised speed and latency shes purring like a kitten. I ran two benchmarks targeting the igpu specifically and even ran 6 videos at once all on the monitored powered by the igpu not even a hickup.
This issues are usually about isolating variables. If you pop in a new chip, that new chip could be totally cool with a setting on the mobo that was causing the issue with the 7950x. Also you are assuming the issue is CPU and not memory. In my experience when a CPU has a issue it starts freaking out, this wasn’t the case here, the CPU would still have low utilization, during the freezing. Could the chip be defective? Sure, but you’d have other problems not just the igpu, IMHO.
Neither the CPU or gpus are overclocked but I will see the cpu boost to 5.5ghz when needed thanks to the the AMD equivalent of Intels turbo feature. The 7950x is a nice chip that runs cool with a artic III aio cooler. Haven’t gone above 15% utilization and above 50celcius where as the intel I9-9900k was at 80%. It’s a work horse.
Good Luck!
I should say i still have hardware acceleration disabled on Firefox, I don’t need it, i dont think this was part of the problem, but im not about to test it.
So does it mean that with default JEDEC RAM timings presented by your motherboard you still had stability issues or your motherboard’s BIOS didn’t offer you an option for default timings so you had to type in each value manually?
And regarding hardware acceleration in Firefox: do you mean that you don’t want to enable it because you worry that it might make your system/CPU unstable again?
P.S. refering to both your and my case: an average user shouldn’t have to tweak this much just to be able to boot his system and use a web-browser. I get the impression that a manufacturer messed up.