GPU questions

[billyzbear]

Hello, I've been having a rough time as of late. I play From the Depths and Forza Horizon 5. Sometimes others but mostly these two games. It's a long painful story that happened over a couple of months. The video card was artifacting pretty bad and the fans were randomly hitting 100% making a bad noise playing FH5. I RMAed the card. After they told me the fans were the problem. FH5 played fine but FTD would crash from time to time. FTD isn't known to be the most stable. I just deal with it. A couple of months ago I decide to figure it out because it was crashing after playing for an hour. When I do get back to Windows it says there is a problem with NV drivers. I didn't look before but it has been setting codes the whole time. I use DDU and do a clean install of latest drivers. Now I get a code 43. I end up updating the bios for video card which stopped code 43. I decide to do the MB as well. It then starts blue screening playing both games. During all of this I have noticed voltages were low under load. I up the voltage to CPU and that stopped blue screen. In the bios update there was a change to CPU voltage. Now looking into voltages under load PCIe slot 11.9v, 5v was 4.5v, 3.3v was 3.1v. I decide to get a new power supply. I put back CPU voltage to auto and voltages are good except 5v is 4.8v all the time but voltages aren't dropping underload. FTD is still crashing but it's random and I'm able to back desktop. Like sometimes it just slows way down and is fine again. This happens from time to time and sometimes it crashes.
Now for the reason I'm here. Today I noticed power consumption hits 102.9 % TDP. After GPU clock goes to 0%, memory clock almost stopped, GPU load went to zero, memory controller load zero. This seems to be the reason the game slows way down. Sorry I didn't take a screen shot. I do have the log. Under Perfcap reason it gives a number, 1-16. What do the numbers mean? 1 is load? 16 is idle? In the logs it's not like the graph. The graph showed the stats up and down a little then there is this major drop to 0 or really close to it for a couple of seconds. I saw this right after a major slow down in game. Temps were around 60c and hot spots where around 70c. CPU was around 80c so heavy load but not too high.
From what I understand it's normal for it to raise and lower watts, power consumption TDP. Why is it that at 102.9 and it goes into shutdown mode? Usually it's not that high more like 90 to 101.
As for Gigabyte. They're telling me I'm out of warranty and won't talk with me. It seems they did more than just the fan because of the code 43 and NV drivers and to get rid of them I had to update bios.
Is there a way to tell the card that using more than 102.9 TDP is ok?
Any help would be appreciated

 

[ir_cow]

Turn off your CPU / GPU overclock and disable memory XMP to see if the BSOD goes away.

 

[billyzbear]

After I upped the voltage to CPU it didn't blue screen again. So I installed a new power supply. I now have the CPU voltage back to auto and I'm still not blue screening.

I have a screen shot now. I was playing game. Major slow down. Paused game and tool a screen shot of the graph.

On the PerfCap reason says pwr, thrm in red for a second.

I went for upping the power limit with Afterburner. This is what I got.

 

[billyzbear]

I was really hoping I could get some thoughts on what is going on.
Today I was playing, I was getting slow downs some longer than others. You can see when it happens GPU load goes to near zero. Then game crashed. It took a bit to get back to Windows but I didn't have to reset. PerfCap showed none, it's in white. This makes no sense to me so if you have an idea I'd appreciate it.

 

[maxfly]

Troubleshoot with a different GPU. Or test your card in a back up rig or friends system. It sounds like your RMA should have been a card swap not a fan swap.

 

[billyzbear]

I saw on another post here, Perfcap none means there is nothing limiting the performance of the card, but fps drops accompanied by lower usage can be explained by CPU bottleneck.

It seems by this post the GPU isn't getting instructions on what to do, a bottleneck. The game was waiting and waiting then stopped.

When I came out of game it took forever for it to recover. I was looking at GPU-Z for awhile. I took a snapshot and it took minutes before it would let me save it. GPU-Z was running fine. I was at desktop game was closed. I was getting the little blue circle like it's doing something. I hit CNTL-ALT-DET to close whatever was running. It hung and hung. When the window did show to name the file to save it. It was acting like a whole lot of things were running in the back ground. Super slow mo. I've noticed this before and to prepair I've debloated Win10. I found a program that deleted all the background apps and programs. What is odd though task manager never came up. I hit CNTL-ALT-DEL a couple of times. I have been noticing that when I have several tabs open, YouTube and Google. Like you are listening to YouTube and surfing. When I click back on YouTube it takes a couple of seconds before it will refresh what is on that page. Lately I've been checking memory with AIDA64. The numbers I get vary. Read, write and copy I've seen as high as low 50,000 and lows 35,000. Memory latency varies as well. 48. - 42.. I've had Windows check memory and it says good. I've bench test, maybe I need to do it for hours. A while ago I did have a fan go bad on MB heatsink. It was making a bad noise and when I put my finger on it to tell which one it moved a little over. Broke the fan blade off. It shock the hell out of the MB for a bit.

I'm kind of leaning to it's the MB. I can switch video cards with my boy. He may not want to give it back...Teens. He is using my old 1060ti. I'm not sure if his power supply can run it.

What memory test would you run for several hours. The crashing seems to happen more after a couple of hours.

Is it really odd to see 5v line be 4.8v? I've never seen it except with this board. Reading on it and seams it's not a big deal.

 

[pressing on]

When you installed the latest motherboard BIOS did you also install the latest Intel INF (chipset driver) and Intel Management Engine (ME) Interface as well? If not I would suggest you do this. The ME interface or driver works with a matching ME firmware update that would have been part of the Gigabyte BIOS update.

 

[nomdeplume]

I'm kind of leaning to it's the MB. I can switch video cards with my boy. He may not want to give it back...Teens. He is using my old 1060ti. I'm not sure if his power supply can run it.

I did wonder while reading this far. Maybe the sane first step here would be uninstalling the gpu and (or borrowing his 1060ti if no igpu) hooking up his psu to your mobo. If you go this far might as well eliminate your RAM as well. At the very least it could be satisfying to remove and physically go over your board looking for defects.

Playing a game might actually be the preferred test here unless you can successfully isolate the problem through physical inspection or immediate difference after switching in psu or RAM. Which should be done in stages to eliminate making more than one change.

 

[billyzbear]

I did a little poking around in AIDA64. I was logging errors for virtual memory. I looked at the settings. I had max at 16000mb. I think way back I saw a video to set it at that or something. I set min. to Win recommended and max to 1.5x that.
I went back and downloaded all the drivers for board. I had them anyways but it can't hurt.
I ran prime for two hours with no hiccups. I played for two hours with no issues.
Back when I first looked into errors and codes. I looked into game logs and it showed I was getting memory errors. Talking with devs it does happen. I went and got a copy of MODS and MATS. It passed several rounds of MATS which is used to test memory of GPU, hot and cold. I don't know enough about MODS to test other parts of GPU. These cards have memory problems. Which I'm guessing was the problem with it when RMA. It was artifacting all the time and sometimes the fans hitting max because of hot spots I guess. From what I've read artifacting isn't a heat issue more a memory issue. The fans were making a clicking noise so I can understand they needed to be replaced.
When I started this post I was using a different bios for gpu. Since I was having so many issues with it power throttling I decided to go back to OG bios. Since it's not setting codes I'm going to say I had a PSU problem in the beginning. While changing the bios may have stopped it from setting code it wasn't the fix.
Another reason why the MB is an issue. Normally I set ram to pc3600 mild overclock. When looking into all this I went back and set to XMP. No OC of anything. A week or so ago I overclocked the ram again. It was fine no issues, games ran fine except the power throttle. Out of the blue it was just like the other day, super slow mo. Like something else was bogging the system down then it crashed with nothing else running. So I set to XMP and reboot. Won't go into Windows. I go back and look at all the settings, check voltages and temps, good. Try again, won't go. I set it to profile 1 which should be the same bare settings. It goes into Windows haven't had an issue until yesterday during playing three days later.
The other issue is the system temp is all over the place. I can't use it to set system fans. A little after the fan broke I went through all the fans, checking bearings. A couple were bad. I set a new PWM fan on system header. The temp was swinging +- 10c in seconds 30c to as high as 50c. I check with just putting finger there and seemed fine to me. I can't remember what was going on but the board didn't like having the fan on the system header.
I understand that an i7 7700k is old. I'm really thinking of getting an i7 13700 maybe an i5 13600 and 32g of DDR5. I've been looking into an MSI z790 Tomahawk or Pro. With video cards costing so much I don't really want to buy a new one unless I have to.

 

[billyzbear]

I wanted to give an update. I played and it crashed and couldn't get back to desktop. I was still getting errors with virtual memory. I followed long instructions from Microsoft. I played game and no issues. The code hasn't come back.

More testing.

I'm going to do all I can to fix problem with crashing and keep what I've got. If I don't get any errors or codes and it still crashes then I'm getting a new setup. Would it be worth the hassle to reinstall win10? I upgraded from before, I didn't pay for win10 so I'm not certain if I can. At some point I'll need to upgrade to use win 11. My CPU isn't approved for it. The longer I wait to upgrade the more performance I'll get for money.

 

[Solaris17]

Would it be worth the hassle to reinstall win10?

I mean from everything I have read you leave nothing stock long enough to get a good measurement. You have messed with windows settings, you still don’t have the bios at default you just turned up voltages, and now your chasing virtual memory “errors” but haven’t given us a stop code or event ID.

It seems like you might repeat all this after you fresh install so why bother?

The system has never been stock in anyway since this thread began and frankly I’m surprised no one has called you out on it. How are you supposed to know what variable cause what effect if you change like 3 things everytime you chase something new?

 

[wheresmycar]

The system has never been stock in anyway since this thread began and frankly I’m surprised no one has called you out on it. How are you supposed to know what variable cause what effect if you change like 3 things everytime you chase something new?

yep, i'm team Super Dainty Moderator on this one!

 

[billyzbear]

I'm sorry, I thought I explained things pretty good. All this has happened in a couple of months. I first was having issues with FTD crashing. I looked in Device Manager and there was a code for NV drivers. I ended up using DDU and clean install of latest drivers. Then a code 43 showed up. I updated the gpu bios. Code went away but still crashing. I updated MB bios. I decide to change power supply after getting back to desktop that some device stopped working error and BSOD with auto settings. Voltages were always a little low under load. 12v -11.9v, 3.3v - 3.2v, 5v - 4.7v. Now the voltages are good except 5v line reads 4.95v all the time. The other day, when I started this post, I was having issues with power throttling. You can see under "again" screen shot. So I flashed GPU bios back to OG. Not having throttling anymore. Still crashing. PerfCap - none - issue happened. I don't know anything about what this means. I found a post that says it's from a bottleneck from CPU to GPU and I guess the GPU is waiting for instructions. At this point instead of replacing or swapping parts I look for other errors that could be a problem. I find virtual memory errors. ID 2004 low virtual memory error and ID 157 disk 1 has been surprised removed as well as disk 2 and disk 3 have been surprised removed. I believe the disk error is about virtual memory. In fix of ID2004 they have a lot of steps to follow which I did and it passed.
As far as stock. I usually have memory boosted a little. First thing I did was just XMP. When it BSOD with auto voltage I upped voltage to CPU but after PSU I put back to auto. Later I boosted the memory a little and it was fine for three, four days and had a mini heart attack. I'm back to XMP 100% stuck.

 

[Deleted member 229121]

Swap the card out.

It is the fastest way to determine if the GPU is your problem by process of elimination.
Pop it in your son's rig and if it crashes there too you have your answer.

If not, then back to looking at your setup (clean install + rollback to previous BIOS and revert to stock).
Keep the 2070 in his rig and then test your old card with the above, if it still crashes only then would I say it's the board.

 

[billyzbear]

I hear you. My boy won't like it. I have to swap out his power supply for him to run the 2070s. I can run 1060ti no problem. I've never done a clean install.

I should add that From the Depths is a very CPU demanding game. There are a lot of calculations it has to do. I enjoy it. I can play Forza Horizon 5 all day long and it has never crashed. When I asked in FTD forum if the game was 100% stable 100% of the time everybody said yes or very rarely. The thing is they all had newer rigs than mine.

 

[billyzbear]

I've played for fourteen hours now and it hasn't crashed. I don't want to say it's fixed because I don't want to jinx it. Usually 1-2 hours and it would crash. It's been doing this for months. It was a little better after PSU but would still crash. Maybe the 2070s finally got scared and didn't want to go to my boys PC. I even got the go ahead from boy and we were set to switch. Other than the power supply and fooling around I followed all the steps for ID code 2004 low virtual memory from Google. I'm still getting major slow downs. The snap shot was after playing for a long while. FPS is low 40ish at 1444p. It's the size of my monitor. Someone stated with a newer rig around 80 fps with 4070 and i5 13600. I still don't understand what the whole PerfCap none thing is all about. If I had to guess the MB had another heart attack. I guess I'll go with it until she bucks me or I get the upgrade bug which ever comes first. Thanks all for the help I appreciate it.