i9 7900x delidded by Overclockers, one core runs 20c hotter + issues

Chaoss · 27 Nov 2017 at 11:44

I've bought an Asus Rampage VI Apex x299 and Intel i9 7900x from Overclockers which is cooled by a fractal design Celsius 360mm aio, with 6 noctua fans in push/pull configuration but I'm running into heat issues when leaving the pc compiling and rendering. I've had the CPU delidded by Overclockers due to overheating issues.

Since I got this machine it's had nothing but problems, I've had to RMA so many peices of it to overclockers (and subsequently needed to take the time off work) that my boss has said there can be no more downtime as I'm really far behind on my tickets.

I've finally gotten the system "working" but there are 2 distinct problems.

1: the cpu temps hit 100c after a few minutes of compiling Unreal Engine 4. I knew it was getting hot as the heat coming off the pc could be felt 2 meters away!

2: When running Cinebench (or indeed any other cpu intensive task that uses all cores) the entire system will lock up, but then will unfreeze a few moments later, the task that the pc was doing when it locked up has progressed during this lockup. I've RMA'd my TeamGroup ram and had it replaced with some G.Skill I had the problem on another system and that was the only part I had that came from that system.

I'm about ready to either sell this on eBay as faulty and buy another system, I've had months and months of these problems and it's caused me far too much stress.

Ideas on how to fix this would be welcomed or advice on what to do.

Here's my HWinfo about 90% of the way through a Cinebench run.

gavinh87 · 27 Nov 2017 at 11:50

That is seriously toasty for a cinebench run.
Not what you want to hear but I'd be sending that back.

Unless I'm blind I cannot see the vcore on those screens.

Chaoss · 27 Nov 2017 at 11:51

gavinh87 said:
That is seriously toasty for a cinebench run.
Not what you want to hear but I'd be sending that back.

Unless I'm blind I cannot see the vcore on those screens.

vcore is 1.2v manually set. core 2 goes up to 120c before the pc shuts down but only after a few long hours and only sometimes.

Radox-0 · 27 Nov 2017 at 11:52

Lower the overclock and voltages slightly which should improve temps, unless you brought a specifically binned chip? double check the mounting / AIO.

I knew it was getting hot as the heat coming off the pc could be felt 2 meters away!

Laws of physics at play here. Your CPU will be dumping out X amount of heat anyways into the air so even if your had the PC had the best cooling solution on it and core temps were only hitting 70 degrees, with the same parameters as you have now, same amount of heat would be chucked into the air.

Chaoss · 27 Nov 2017 at 11:54

Radox-0 said:
Lower the overclock and voltages slightly which should improve temps, unless you brought a specifically binned chip? double check the mounting / AIO.

Laws of physics at play here. Your CPU will be dumping out X amount of heat anyways into the air so even if your had the PC had the best cooling solution on it and core temps were only hitting 70 degrees, with the same parameters as you have now, same amount of heat would be chucked into the air.

What settings would you reccomend trying?

Radox-0 · 27 Nov 2017 at 11:59

Chaoss said:
What settings would you reccomend trying?

Not an expert on that chip, but following typical principles, would drop the voltage down to 1.15v and see what clock you can stabilise with that, should shave some degrees off and tweak as needed.

Chaoss · 27 Nov 2017 at 12:07

Radox-0 said:
Not an expert on that chip, but following typical principles, would drop the voltage down to 1.15v and see what clock you can stabilise with that, should shave some degrees off and tweak as needed.

OK I've set the vcore to 1.15 & cpu multiplier to 45, I've also followed a recommendation someone else suggested on Discord and manually set the cache voltage to 1 and the multiplier to min 28 max 30 (I don't know what the default values are or why that matters/what they do?) Now I'm seeing 83c tops after repeated Cinebench runsand a slightly higher than avarage Cinebench score of 2445 (it was 2330 before)

I have another pesky problem with this system as well that happens regardless of what settings I use, when running Cinebench (or rendering/compiling) it will freeze for a moment and then unfreeze. There is no help anywhere online at what could be the cause of that issue, would you have any ideas? the freezing does not affect score and if rendering, it will have "progressed" while it was frozen, so it'd have taken the same amount of time to complete whether it froze or not

gavinh87 · 27 Nov 2017 at 12:20

Freezing usually points to ram.
Might be worth doing a memtest.

Chaoss · 27 Nov 2017 at 12:21

gavinh87 said:
Freezing usually points to ram.
Might be worth doing a memtest.

I've just had to send my old sticks of ram in as they did this. the ram I've got back is a different brand G.Skill TridentZ are you sure?

gavinh87 · 27 Nov 2017 at 13:02

Chaoss said:
I've just had to send my old sticks of ram in as they did this. the ram I've got back is a different brand G.Skill TridentZ are you sure?

I'm not sure, no. I said usually.
Running a memory testing program will tell you if something is amiss. Otherwise it could be storage related

Radox-0 · 27 Nov 2017 at 13:10

Chaoss said:
I've just had to send my old sticks of ram in as they did this. the ram I've got back is a different brand G.Skill TridentZ are you sure?

I see same issue with cinebench now and then. It seems to freeze but continues and completes perfectly fine and other stress tests pass fine so I consider it a non issue. If everything else is performing fine and the system is stable i would say it's not a problem. Do just run memtest as suggested to rule it out, but it it passes and everything else is fine. Happy days i say.

Silent_Scone · 27 Nov 2017 at 16:06

Those temps are what one should expect from a CPU that hasn't been delidded. For what it's worth, comparing to an 8700K is fruitless as the current drawn will be greater. Some CPU will have higher core delta than others, though.

Bailey · 27 Nov 2017 at 16:49

Hey there

Sorry to hear of you having issues, a couple of things I would try are:

1 - AVX Instruction Core Ratio Negative Offset - 3
2 - Disable XMP and try mem at stock

Try that and see if that helps. I have also alerted 8 pack to have a look on this thread too

Cheers

Bailey

Chaoss · 27 Nov 2017 at 17:10

Bailey said:
Hey there

Sorry to hear of you having issues, a couple of things I would try are:

1 - AVX Instruction Core Ratio Negative Offset - 3
2 - Disable XMP and try mem at stock

Try that and see if that helps. I have also alerted 8 pack to have a look on this thread too

Cheers

Bailey

I have AVX set to -5 and enabling or disabling XMP doesn't yield any differences. I've followed some advice from another forum and set the cache voltage to 1.0v, cache ratio min to 28 and max to 30, set the vcore to 1.15 and cpu ratio to 45 and so far it seems to run A LOT cooler. Core 2 still runs ~20c hotter than the other cores and I still have a bsod about once or twice a week problem but I guess thats just how PC's work now and they aren't built like they were 10 years ago. I'll have to learn to live with it/get used to it which is a shame since I've spent upwards for £5k on this (including monitor etc.)

I might contact the buyer I sold my old core 2 quad to and ask to buy it back and use both system, the core 2 quad for most of the work and just use the 7900x for the heavy workloads. I can't play "wait for the RMA" any more as I've been told I WILL loose my job if I fall any further behind and I've already had to RMA so many parts from this thing.

awaybreaktoday · 27 Nov 2017 at 17:24

What happens if you run it at stock settings?

Chaoss · 27 Nov 2017 at 17:26

awaybreaktoday said:
What happens if you run it at stock settings?

If i reset the bios (which I've done a few times now) it is pretty much the same, it actually runs cooler with the settings I've put in now. Either way stability doesn't change either way

Chaoss · 27 Nov 2017 at 19:00

Bailey said:
Hey there
...Try that and see if that helps. I have also alerted 8 pack to have a look on this thread too

Cheers

Bailey

You have 8 pack on these forums?! Nice. Yeh I could definately use some enterprising help. I mean it seems to at least stay relatively cool now if it bsods while I'm working again though I'll probably snap my keyboard in half in frustration lol

Figured I'd show you what I do with all this hardware (or at least what I'm supposed to be doing I've fallen quite far behind because of these problems)...

Link is here if you want to know more: https://playdeadmatter.com/forums/index.php?/topic/1005-development-diary-4/

It's quite important I get this working, the amount of times I've lost several hours of work due to an unexpected bsod corrupting my (currently open) 3dsmax file and having not saved it in a while is more than annoying hence why I'm thinking of asking the guy I sold my Core2Quad to if I can buy it back and just use that for most of the 3dsmax work (and just use the 7900x for rendering and compiling) since the 7900x is not a CPU that's very good for game development

awaybreaktoday · 27 Nov 2017 at 19:23

Chaoss said:
You have 8 pack on these forums?! Nice. Yeh I could definately use some enterprising help. I mean it seems to at least stay relatively cool now if it bsods while I'm working again though I'll probably snap my keyboard in half in frustration lol

Figured I'd show you what I do with all this hardware (or at least what I'm supposed to be doing I've fallen quite far behind because of these problems)...

Link is here if you want to know more: https://playdeadmatter.com/forums/index.php?/topic/1005-development-diary-4/

It's quite important I get this working, the amount of times I've lost several hours of work due to an unexpected bsod corrupting my (currently open) 3dsmax file and having not saved it in a while is more than annoying hence why I'm thinking of asking the guy I sold my Core2Quad to if I can buy it back and just use that for most of the 3dsmax work (and just use the 7900x for rendering and compiling) since the 7900x is not a CPU that's very good for game development

Your work looks great.. how long has this game been in development for now and when is it due for release?

Danny75 · 27 Nov 2017 at 20:50

Chaoss said:
I have another pesky problem with this system as well that happens regardless of what settings I use, when running Cinebench (or rendering/compiling) it will freeze for a moment and then unfreeze. There is no help anywhere online at what could be the cause of that issue, would you have any ideas? the freezing does not affect score and if rendering, it will have "progressed" while it was frozen, so it'd have taken the same amount of time to complete whether it froze or not

Radox-0 said:
I see same issue with cinebench now and then. It seems to freeze but continues and completes perfectly fine and other stress tests pass fine so I consider it a non issue. If everything else is performing fine and the system is stable i would say it's not a problem.

I've seen this on Cinebench (with a different processor - 4770K) when experimenting with changing CPU priority through Task Manager to High or Realtime. It would freeze, fail to show the rendering progress smoothly, but ultimately it would show the run completed and with a slightly higher score. So it's doing the rendering (and when done it shows the full render), but it's not showing all of the display/frames as it's taking place, because too many of the resources are going to the rendering itself. Not sure how helpful that is in your case since I highly doubt priority is changing automatically to High/Realtime but that's one cause of what you describe.

ZXSpectrum · 25 Jun 2018 at 14:05

I normally deal with Workstation boards but it should apply all the same. They often do have a core that runs hotter than the rest but that is excessive that would suggest that the thermal compound isn't covering the die or there is contamination in the liquid metal.

Was the CPU binned?

I would try some simple settings such as:

Ai Overclock Tuner [XMP]
ASUS MultiCore Enhancement [Disabled]
AVX Instruction Core Ratio Negative Offset [2]
AVX-512 Instruction Core Ratio Negative Offset [2]
CPU Core Ratio [Sync All Cores]
ALL-Core Ratio Limit [42]
DRAM Frequency [DDR4-3000MHz]
TPU [Keep Current Settings]
CPU SVID Support [Disabled]
Extreme Over-voltage [Disabled]
BCLK Aware Adaptive Voltage [Disabled]
CPU Core Voltage [Manual Mode]
CPU Core Voltage Override [1.10]
CPU Input Voltage [1.900]
CPU Current Capability [140%]
CPU VRM Switching Frequency [Auto]
VRM Spread Spectrum [Disabled]
CPU Power Phase Control [Optimized/Extreme]
CPU Power Thermal Control [140]
DRAM Current Capability [100%]

See if that exhibits the same issues. If the answer is yes then you may have to try that CPU in a different board.

If it doesn't then bump the core ratio up incrementally and when it becomes unstable add a little voltage.

Hope that helps