Overclock.net banner

1 - 20 of 23 Posts

·
Registered
Joined
·
1,370 Posts
Discussion Starter #1 (Edited)
Hi there,

So a week ago, I ordered my new computer parts.

8700k, that I have OCed with just a modest OC to 4700 CPU, 4400 Cache at 1.24v (Ran Prime95 small test for an hour with no issues)
G.Skill 4133mhz RAM running at that speed
And the Gigabyte Gaming 7 Z370 board.
1080TI @ 2020/6100
My cooling is a Corsair H110 with Noctua Fan upgrades.

So here is where I am having the issue,
My roommate has the EXACT same computer that I built him, except that his ram is G.Skill 3200mhz vs my 4133Mhz model.
He also didn't want me to overclock his computer, so he is running it at stock.
Thing is, he consistently gets better FPS than me in PUBG, Fortnite and R6 (Those are mostly the games we play together.) Yet he plays on a 1440p res, and I play on a 1080p res. Setting are the same for both in games.
I have not been able to figure out why that is.

I ran a 3Dmark test in Spy Extreme and scored a 3421 on the CPU test, but when comparing those results online, most people with my type of system scored around 3750-3850.
Cinebench score single core of 200 and multi-thread of 1411.

I'm not hitting thermal throttle, as my CPU never goes above 80, usually staying around 76C during heavy gameplay.

My question is, did I get a faulty CPU? Or could it be the Motherboard causing this, or the RAM? Should I try running my ram underclocked to 3200 and see if my results improve?
Or should I just send this CPU back while it's still within it's return policy, and just buy a different one, or just upgrade to the 8086k.
 

·
Registered
Joined
·
1,370 Posts
Discussion Starter #2 (Edited)
Just to show the difference in tests, ran a few games in the exact same area with same settings.

PUBG In certain areas my FPS will go down to 100 while my friends stays around 120.

R6 FPS will go down to 150 while looking at a specific spot on the skyrise map, and in the EXACT same spot, my friends FPS is 174 average.
 

·
Registered
Joined
·
1,370 Posts
Discussion Starter #5
Check your pcie speed, force it to x16 or x8
It's not a GPU issue, my gpu scores higher than his in every GPU bench we ran.
My CPU however scores below his, despite running at 4.7Ghz on every core, and his runs at 4.3 on every.
 

·
Registered
Joined
·
5,374 Posts
Is your ram stable on your board? Maybe try setting it to 3200Mhz and see what happens. While I'm sure your ram is capable of what it is rated for, changing the cache will affect stability as well as the board. Maybe your board doesn't like the ram?

Are you sure this is a CPU issue instead of a GPU issue?

Are your motherboard temps OK?
Did you clean install the GPU driver or install it over the default driver? Has windows updated since then?

If you got your score in TimeSpy while others scored higher, it's possibly many people running 3DMark are overclocked to 5Ghz.
What are your friend's results in TimeSpy?
Can your pass Prime95 blend for a few hours?
Is your vram actually stable? With Pascal ram you can run it unstable just fine but the error correction is working overtime and slows the GPU down.
Does your board need a BIOS update?
Maybe try clearing the CMOS and compare your stock results against others.



If it doesn't get figured out this should fix anything unless something is actually broken:

I'd go back to square one, clear CMOS, clean install of Windows (unless you just installed it) then let it fully update to the latest. Install Prime95, HWInfo, GPUZ, then any bench type software you like. Avoid installing anything from Gigabyte or whatever bloat you might come across.

Now that you're fully updated get into the BIOS and set your XMP but pull the speed down to 3200Mhz after.
Set MCE or manually set the cores to 4.7Ghz (or whatever speed you want), set the voltage you think it needs but a little less. If you want 4.7 I'd set 1.25v manually (not adaptive) then stress with Prime95 small FFT without AVX for now. My logic on stress testing is I don't want to run the stress test for very long since it'll wear it down faster than anything so I pick a multiplier and undershoot the voltage so I get a fast crash. If it happened instantly I'll give it close to 0.1 but if it took a few minutes I only do +0.05, if it took 30 minutes to crash I'll get down into the thousandths with the vcore or just add 0.01 so I don't need to worry about it crashing again. Then I'll run small FFT for 2 hours, minimum 1 hour.
Then I do the cache the same way, usually I'd do the ram first but since there's really no benefit to ram faster than 3800Mhz I'd say it's ok to start with cache. Do it the same way with the same test method. After that set the ram, the only thing I do different with ram is test with blend but for much longer. If you can pass blend for +4 hours that combination should be stable AF. After that you can set adaptive voltages.
I set Windows 10 to it's performance setting with the minimum CPU speed at 5% so it will still downclock and undervolt but the clock speed will stay pegged when under load.

Now DDU the windows driver and install the latest nvidia driver change to optimal performance and restart. Proceed to small core bumps while gaming with unlimited FPS till you find your sweet spot. For ram I run something like superposition (it's short) and bump the ram till the score stops rising then take it down a little more.


EDIT: You answered some questions while I was typing this.
 

·
Registered
Joined
·
3,747 Posts
Try running BIOS at default F7 with memory clocked to 3200 and see what you get for FPS in PUBG. From what you describe it is a GPU issue.
 

·
Registered
Joined
·
491 Posts
One quick and dirty test to see if its the ram is pop the 3200 out of your mates PC and stick it in yours and see if it makes a difference.

BTW it is worth noting that the Intel processors are far less dependant on RAM speed for performance. Ryzens depend on good fast RAM to get optimum speed but Intel chips really do not care too much about RAM speed, any reasonable RAM will do.
 

·
Registered
Joined
·
1,370 Posts
Discussion Starter #9
Is your ram stable on your board? Maybe try setting it to 3200Mhz and see what happens. While I'm sure your ram is capable of what it is rated for, changing the cache will affect stability as well as the board. Maybe your board doesn't like the ram?

Are you sure this is a CPU issue instead of a GPU issue?

Are your motherboard temps OK?
Did you clean install the GPU driver or install it over the default driver? Has windows updated since then?

If you got your score in TimeSpy while others scored higher, it's possibly many people running 3DMark are overclocked to 5Ghz.
What are your friend's results in TimeSpy?
Can your pass Prime95 blend for a few hours?
Is your vram actually stable? With Pascal ram you can run it unstable just fine but the error correction is working overtime and slows the GPU down.
Does your board need a BIOS update?
Maybe try clearing the CMOS and compare your stock results against others.



If it doesn't get figured out this should fix anything unless something is actually broken:

I'd go back to square one, clear CMOS, clean install of Windows (unless you just installed it) then let it fully update to the latest. Install Prime95, HWInfo, GPUZ, then any bench type software you like. Avoid installing anything from Gigabyte or whatever bloat you might come across.

Now that you're fully updated get into the BIOS and set your XMP but pull the speed down to 3200Mhz after.
Set MCE or manually set the cores to 4.7Ghz (or whatever speed you want), set the voltage you think it needs but a little less. If you want 4.7 I'd set 1.25v manually (not adaptive) then stress with Prime95 small FFT without AVX for now. My logic on stress testing is I don't want to run the stress test for very long since it'll wear it down faster than anything so I pick a multiplier and undershoot the voltage so I get a fast crash. If it happened instantly I'll give it close to 0.1 but if it took a few minutes I only do +0.05, if it took 30 minutes to crash I'll get down into the thousandths with the vcore or just add 0.01 so I don't need to worry about it crashing again. Then I'll run small FFT for 2 hours, minimum 1 hour.
Then I do the cache the same way, usually I'd do the ram first but since there's really no benefit to ram faster than 3800Mhz I'd say it's ok to start with cache. Do it the same way with the same test method. After that set the ram, the only thing I do different with ram is test with blend but for much longer. If you can pass blend for +4 hours that combination should be stable AF. After that you can set adaptive voltages.
I set Windows 10 to it's performance setting with the minimum CPU speed at 5% so it will still downclock and undervolt but the clock speed will stay pegged when under load.

Now DDU the windows driver and install the latest nvidia driver change to optimal performance and restart. Proceed to small core bumps while gaming with unlimited FPS till you find your sweet spot. For ram I run something like superposition (it's short) and bump the ram till the score stops rising then take it down a little more.


EDIT: You answered some questions while I was typing this.
Just to answer a few, and than I will try the others and get back.
The windows install is NOT fresh, I used the old windows that I just migrated from my old SSD to the new M.2 drive. Then manually uninstalled all the old drivers from device manager and installed the new ones from the website. But the windows itself is still the old one. The drivers for the GPU are clean, but I had this 1080ti even before my new mobo/ram/cpu came in.

The GPU is stable for sure, I ran the benchmarks for 2h+ while messing with the vram clocks and this was the highest it could go, while still stable. BUT, even with both of our 1080tis set to stock, I still get the lower FPS, but mostly in CPU intensive games. In GPU only benchmarks, my 1080ti actually scores about 2-3% higher at same clocks.

I ran Prime95 for 2 hours and 11 minutes total while I was away from home, and when I got back it was still stable with 0 errors or crashes. Max temps did go up to 90C tho, but XTU didn't show thermal throttling.

Also just to update, I ran cinebench on both of our machines, and he actually scored slightly slower in both single and multi. I scored 1441/200, he scored 1315/185.
My friend's timespy bench is actually almost identical to mine, he scored just 5 points under me, but he is running his CPU at 4.3 with 4.0 cache, and mine is 4.7 with 4.5 cache, so I think I should have definitely gotten more than 5 points difference on the CPU test.

So I'm starting to think it might be the RAM.

I am running the latest version of the bios availble.

I will be trying to drop the ram down to 3200 and seeing how that goes, and what my difference is. Despite the fact that I ran the blend for 2+ hours and it passed at 4133(which is what the ram is rated for), but I just can't see what else it could be.
 

·
Registered
Joined
·
1,370 Posts
Discussion Starter #10
One more update, I just tried running OCCT test, and while I was passing it just fine earlier, it appears that bumping my cache from 4.4 to 4.5 has made it so that It gives me an error on core 1(2,4,8) about 20-30 seconds into the test. Does this mean that it IS a cpu issue, or ram?
 

·
Registered
Joined
·
1,370 Posts
Discussion Starter #11
So, I set my ram to 3866 and tried to change the timings down to 18-19-19-38 from 19-19-19-39. and when I tried to reboot, the mobo went in to infinite boot and would never actually boot. Finally got it to boot, and it would freeze while in the bios. So I reset the CMOS and set everything to stock, but set the ram to 3866 with the default timings of 19-19-19-39, it booted just fine in to windows. However, running the OCCT test again, this time 5 minutes in to it, it gave me an error again on core 2. So then, I reset the cmos again, and again set the ram to xpm default but lowered it to 3200, and OCCT again crashed about 5 minutes in giving me an error on core 1. It looks like changing the RAM doesn't change the fact that it crashes. So maybe it isn't the RAM's fault?

I am running OCCT on small batch.
 

·
Registered
Joined
·
5,374 Posts
If increasing the cache speed hurt something or gave an error it's probably the cache. That's why I wrote the thing up there about stressing, it's not hard to get something that won't crash but hurts performance. Running Prime small FFT is great for crashing cores and cache but if OCCT works then so be it.

Try a lower ram speed after bringing the cache down/voltage up but if that doesn't work I'd blame the dirty Windows install, the upgrades almost always come at a cost to gamers and overclockers. You could actually make a Windows To-Go stick and boot with that so you don't have to nuke your OS to test, or you can dual boot with a parted drive or a different drive. Restarting your PC in Windows recovery may do the same thing but I've never tried it to find out.

This won't solve your issue but when you get it sorted out you can use O&O Shutup to kill all the Windows services you don't want or that hurt performance. After you've used that to prevent Windows from installing new stuff you can clean up the mess with CCleaner.
 

·
Registered
Joined
·
1,370 Posts
Discussion Starter #13
If increasing the cache speed hurt something or gave an error it's probably the cache. That's why I wrote the thing up there about stressing, it's not hard to get something that won't crash but hurts performance. Running Prime small FFT is great for crashing cores and cache but if OCCT works then so be it.

Try a lower ram speed after bringing the cache down/voltage up but if that doesn't work I'd blame the dirty Windows install, the upgrades almost always come at a cost to gamers and overclockers. You could actually make a Windows To-Go stick and boot with that so you don't have to nuke your OS to test, or you can dual boot with a parted drive or a different drive.

This won't solve your issue but when you get it sorted out you can use O&O Shutup to kill all the Windows services you don't want or that hurt performance. After you've used that to prevent Windows from installing new stuff you can clean up the mess with CCleaner.
I think you were writing this as I posted my other post. I did lower the memory speeds with default cache. Let me know what you think?
 

·
Registered
Joined
·
1,370 Posts
Discussion Starter #14
One more update, even tho OCCT crashes, my PUBG is actually now getting MORE fps than my friends computer at the exact same settings, but that's just because he plays on 1440p and I play on 1080p. But, it seems that running my CPU at stock and the ram at 3866 actually increases my performance in games. While still crashing in OCCT. I have the ability to send this CPU back, as it's only been a week, so I think I'm going to send it back and get the 8086k and then test with that.
 

·
Registered
Joined
·
1,370 Posts
Discussion Starter #15
Update:

After doing some more tests, it seems that raising my cache even 100mhz from the stock 4.0, causes my OCCT test to spit out an error. I'm thinking the CPU might be faulty, or it could be the motherboard, but probably the CPU. I have a 30 day return on all 3 components, cpu/ram/mobo, so I went ahead and ordered a 8086k, it will be in on Wednesday, I will test it out with the 8086k at stock, and if I'm still having issues than it's probably the RAM or the Mobo. The RAM I can test by using my buddies RAM, so that will determine if it's the RAM or Mobo, IF the new CPU also fails.
 

·
Registered
Joined
·
3,747 Posts
Have you tested at BIOS default F7 without changing anything in BIOS?
 

·
Newb to Overclock.net
Joined
·
4,168 Posts
Update: Ran Prime95 for about 30 minutes on small batch, and was watching the XTU throttling, on all accounts it said No.
I'm inclined to think that XTU is not representative of your gaming load on the CPU, and at some point, it does throttle because the VRM gets too hot. This seems to be a common issue with this particular motherboard if you google:

What you're seeing is largely a VRM heatsink limitation. That's why to fully utilize the potential of Gaming 7 , ROG Apex/Code/Hero, Taichi, Godlike , etc you would need to cool the VRM with a monoblock or better surface area VRM heatsink than the RGB monstrosity on the Gaming 7.
At least one user reported VRM temperature drops on the Gaming 7 from switching thermal pads and torquing down the VRM heatsink more.
Also, your PSU is not mentioned. What is it?
 

·
Registered
Joined
·
1,370 Posts
Discussion Starter #20
I'm inclined to think that XTU is not representative of your gaming load on the CPU, and at some point, it does throttle because the VRM gets too hot. This seems to be a common issue with this particular motherboard if you google:



Also, your PSU is not mentioned. What is it?
I've watched my VRM temps and they stay under what is considered to be throttling, they stay around 92C even after hours of gameplay.

As for the PSU it is the EVGA 1300G2
https://www.amazon.com/EVGA-SuperNO..._rd_t=40701&psc=1&refRID=SKDH8A0ZPYT4HPCJTNWV
 
1 - 20 of 23 Posts
Top