why im getting some errors in HWinfo64 its that bad ? does my card is bad ?? i have a Sapphire r9 280 3GB factory oc 940/1250 dual-x my temps are fine and from time to time some games crash or stop working :banana:
Your card is probably getting to hot try down clocking or undervolting the memory,cleaning out dirt, and looking at your fan profiles. Moreover, if you really wanna go the extra step! say; for example, im playing warframe this game runs great you could probably lock your card to 700mhz and still play at 60fps. also reflash bios if all fails then id say its RMA or time for 470
well i just saw errors under the Sensors menu, and my card never get over the 67° with close case and ambient temp 32°, if i remove the case side pannerl it gets into the 65° ish i dont xp any artifacts, or any weird things so that's why im warried nothing shows that ther is any problem, i set a custom profile in MSI AF and same happens, also problems get worse with newer drivers, im usin 13.9 WHQL now, if i go to 16'ish its worse i have tp update my Specs tho Gigabyte z77x-ud3h Sapphire r9 280 3GB oc 940/1250 Dual-X CM600M Kingston HyperX 1866Mhz CPU core i5 3570k Stock, CM Tx3 EVO Cooler 2x1T Seagate 7200rpm
Request for input from AMD card owners I am bumping this thread for a valid reason. I know this number doesn't hold much of a significance to some of you, as some of you simply don't care, or feel like "bah this doesn't mean ****" and if that's the case, I respect that, but I'd like to request input from those of you who take interest and can spare some time into this. Firstly, download HWiNFO from here: https://www.hwinfo.com/download.php (This software is more accurate than Hardware Monitor Pro from my experience.) Keep it minimized and running in the background. Play some intensive games like Mass Effect Andromeda, Deus Ex Mankind Divided (DirectX 12) in particular for over half an hour. In fact, I recommend these titles in specific. Post the results under GPU Memory Errors from within the HWiNFO that can be found here: Please, do not be alarmed by the results. The more of your participate, the better. Also, mention whether you're running your card at stock/overclocked, and what kind of memory chip you have (Micron, formerly known as Elpida, or Hynix, Samsung, etc.) if possible. This is an AMD only feature so nvidia GPU owners don't need to bother with this. Thank you.
I found when i had RX480 that as long game does not crash, or you don't see any graphical glitches, you are fine. But if errors goes in 10000+ in few seconds, it will end soon with blackscreen, at least that what i got. I never get any glitches myself, if it was too much for graphic card, it just goes blackscreen then it reset graphic driver and leave graphic card in 2D clocks until you restarted windows.
Not the kind of input I was hoping for, but thanks. To everyone else: All I'm asking is a snapshot of what your memory error counter looks like after a 30 minute gaming session of the games mentioned in my post above.
I always will get a single memory error if I alt-tab during a fullscreen game, or I change its resolution. No real errors when gaming. The most I've seen were during BF1, but even these were always less than 10 for hours of play.
Firstly, thank you for your input, Pathfinder! Could you try Deus Ex: Mankind Divided in DirectX12 and Mass Effect Andromeda, please? I too thought it had something to do with Alt+Tabbing, but I was able to put that theory to rest for the games I have mentioned at least. For me, I get like 5 to 10 at stock while playing Deus Ex: Mankind Divided and up to 30 ish if I OC the memory. Mass Effect Andromeda won't show a single memory error while at stock memory speed, but will throw like 10~15 ish if I OC my memory. I don't want to kill my memory, and I know for a fact that one of the memory Chips has a missing pad. If you run into crashes during DE:MD DX12 mode, feel free to let me know. I think I have a few more tricks up my sleeve. BTW, did you tighten your 7970's Memory Straps in the BIOS? I think you'll find this interesting: Link to HWiNFO forums
Dont have ME Andromeda and dont have Deus Ex MD installed but this is mine after an hour of playing Quake Champions then hour of playing Prey.
Just gonna chime in here on GPU Memory Errors and HWINFO. I am just gonna go by my memory of all I have read and not bother to find and point people to these various actual threads and answers by both the aruthor of HWINFO and others...mainly all comes overclock.net and mainly in threads dealing with it, overclocking and bios modding. the GPU Memory Error counter in HWInfo reads both corrected and uncorrected error. We have no way to know which is which. GPU Memory Errors can be caused by memory chips overheating, memory overclocking, GPU Core overclock, too much or too little voltage and incorrect or too tight of memory strap timings. Unlikely but GPU memory errors can also be caused by faulty memory chips, just had to mention it as well. A few gpu errors should not worry you over a gaming sessions or even on the desktop not gaming...it happens there. Ideally you want 0 errors however a few every now and then is okay and should not affect performance. However a few every second to hundreds of thousands every second will GREATLY adversely affect your performance and lower your FPS. From this part you have to figure out what is the root cause and sometimes it more then one thing I mentioned. You should not get gpu memory errors at all when running everything at stock for the GPU you have. Hope this helps...I am sure if I get motivated I can find the specific threads...but I am just summing it up based on what I know with modifying my bios and overclocking. The GPU memory error counter is invaluable in helping to figure out why a higher overclock actually gives lower performance, as it takes the guesswork out.
GDDR5 on AMD GPUs will lower clocks if errors are detected, and it has mechanisms to detect them. As mtrai said, there is no way to know if the errors are corrected or not, and it's also my suspicion that since a lot of them occur when I alt-tab from fullscreen, some are not "real" errors but something to do with flipping framebuffers when switching. That's just my uneducated guess though. DXMD always gave me hard crashes at specific points with DX12. I play it using DX11. Andromeda gives hard crashes when VRAM is filled, but I never had the detector open in the background. When I'll play I'll inform you guys here. My GPU is flashed with a 280x bios. 280x cards have by themselves tighter memory timings than the 7000 series and I saw in various The_Stilt threads that my timings are as good as possible.
mtrai, I'm aware of the technicalities involved and I've linked to a forum post just above your post that discusses exactly what you've said. With that said, I appreciate you taking the time to respond, thank you. What I'm trying to figure out is if there is a difference or a similarity in behavior between cards and specific/different architectures in the aforementioned titles. There are tons of games that doesn't exhibit similar behavior hence I'm trying to narrow the issue down to being game specific or driver specific from or specific to my card from the inputs. Sadly, so far what I've gotten is no relevant input to go on with as most of you probably won't touch Mass Effect Andromeda or Deus Ex anytime soon. Undying, Thanks man! I guess Quake Champions and Prey are unaffected by this phenomenon (?) or maybe it's just my memory chips being flaky. Again, thanks for the input. Much obliged. PRMinisterGR, If your copy of DE:MD crashes in specific areas of the game while running DirectX12, one possible workaround is to locate a file called PSOCache.bin that's generated under %appdata%\Roaming\Eidos Montreal\Deus Ex Mankind Divided and delete it to see if that stops the crashes. If that doesn't help, there's another PSOCache.bin file right next to the executable within the game's retail directory, rename it PSOCache.bin.bak and that'll definitely fix the issue. But if this is the fix that works for you, I recommend that you update the game to the latest build possible because latter builds have fixed the area specific crashes therefore the PSOCache.bin file within the retail folder can be left alone.
