Jump to content
9 hours ago, caldrin said:

Slow CPU, Slow ram.. :(
I know when you check CPU utilization its not maxed out but how about the cores that are actually getting used by the game.. are they maxed out ?

True, for todays standards the cpu is not the best and ram is very slow. I compared my cpu to one that was worse, older and was performing better for some reason in this video: View: https://www.youtube.com/watch?v=iz8ThsFOP3w

I know they use a rx 5700 xt, but the radeon vii in games performs very similar. The cpu is a year older and is worse in many ways such as having less cache. The cpu ran at 45% load while the gpu ran at 100% load which is quite odd even though I am getting half of the frames I am supposed to get. Once I get on my pc, I will run a couple benchmarks and see how the core utilization is doing. I really am confused by what might causing this. Can a broken memory channel cause this, I am starting to think a defected gpu, but this weekend I am going to swap the radeon vii for a 1080 and see how it performs.

Thanks for the help.

Link to post
Share on other sites

10 hours ago, Eigenvektor said:

It does, but minimal at best. It depends on how many PCIe lanes the slot has, e.g. if it's a x16, x8, ... etc. slot.

 

Here are two benchmarks by TPU:

https://www.techpowerup.com/review/nvidia-geforce-gtx-1080-pci-express-scaling/

https://www.techpowerup.com/review/pci-express-4-0-performance-scaling-radeon-rx-5700-xt/

 

Essentially both PCIe generation and lanes have very little impact.

 

Yeah, this weekend I am going to swap out the radeon vii for a 1080.

Link to post
Share on other sites

11 hours ago, artavasdes said:

True, for todays standards the cpu is not the best and ram is very slow. I compared my cpu to one that was worse, older and was performing better for some reason in this video: View: https://www.youtube.com/watch?v=iz8ThsFOP3w

I know they use a rx 5700 xt, but the radeon vii in games performs very similar. The cpu is a year older and is worse in many ways such as having less cache. The cpu ran at 45% load while the gpu ran at 100% load which is quite odd even though I am getting half of the frames I am supposed to get. Once I get on my pc, I will run a couple benchmarks and see how the core utilization is doing. I really am confused by what might causing this. Can a broken memory channel cause this, I am starting to think a defected gpu, but this weekend I am going to swap the radeon vii for a 1080 and see how it performs.

Thanks for the help.

Yeah only thing you can really do now is start swapping things around really especially if you have spare hardware you can test with :)

Link to post
Share on other sites

11 hours ago, caldrin said:

Yeah only thing you can really do now is start swapping things around really especially if you have spare hardware you can test with :)

Sounds like the best thing to do now, I will tell you guys my results once I test. I am going to swap for a 1080 and try removing ram. Hopefully I find out what is causing this.

Link to post
Share on other sites

  • 2 weeks later...
10 hours ago, artavasdes said:

@Eigenvektor and @caldrin Even Linus made a video featuring the e5-2667v2 and he is encountering no performance loss. I think it might be one of the cpus having a bad memory channel (or the motherboard?).

I've just watched the video. I don't know about no performance loss ? SoTR was mostly fine (40% vs 74% GPU bound), but Anno seems to have been quite slow. So anything that doesn't profit from a lot of cores and/or isn't aware of NUMA seems to suffer. Even Cinebench wasn't that much faster despite lots of cores (because the cores were slower).

 

I don't know about bad memory channel, you mean it doesn't run at full speed? Otherwise I'd expect you wouldn't see all of your memory. You could maybe try MemTest86.

Remember to either quote or @mention others, so they are notified of your reply

Link to post
Share on other sites

On 12/7/2019 at 4:01 PM, artavasdes said:

@Eigenvektor I tried running some games with different cores enabled, I got worse FPS with less cores. I got the highest FPS when I enabled all cores.

You are sure the cores enabled for the very game were on just one of the 2 cpus, not mixed?

You have to prevent that data gets shifted between both cpus.

Link to post
Share on other sites

11 hours ago, Eigenvektor said:

I've just watched the video. I don't know about no performance loss ? SoTR was mostly fine (40% vs 74% GPU bound), but Anno seems to have been quite slow. So anything that doesn't profit from a lot of cores and/or isn't aware of NUMA seems to suffer. Even Cinebench wasn't that much faster despite lots of cores (because the cores were slower).

 

I don't know about bad memory channel, you mean it doesn't run at full speed? Otherwise I'd expect you wouldn't see all of your memory. You could maybe try MemTest86.

Sorry for not being specific about having a broken memory channel. What I mean by that is that I cannot utilize some of the DIMMs (not because the motherboard is broken, I have tested it on other motherboards) because a memory channel is broken on the first cpu.

Link to post
Share on other sites

10 hours ago, Sir0Tek said:

You are sure the cores enabled for the very game were on just one of the 2 cpus, not mixed?

You have to prevent that data gets shifted between both cpus.

I don't know whether they are utilizing both cpus. I tried limiting the cores that could be used but I got worse performance. I don't think there is an option in BIOS to use only one cpu.

Link to post
Share on other sites

9 minutes ago, artavasdes said:

Sorry for not being specific about having a broken memory channel. What I mean by that is that I cannot utilize some of the DIMMs (not because the motherboard is broken, I have tested it on other motherboards) because a memory channel is broken on the first cpu.

Yeah, that is probably bad. I think Linus mentions it in the video: Accessing memory from the other CPU is slower. If some cores have no local memory it essentially has to go the long route all the time. Any threads that live there will have slow memory access.

 

Did the video show the time for Cinebench? You could run those benchmarks on your machine, see what times you get. I think rendering needs a good bit of memory, so if I'm correct it should have a fairly noticeable effect.

Remember to either quote or @mention others, so they are notified of your reply

Link to post
Share on other sites

1 hour ago, Eigenvektor said:

Yeah, that is probably bad. I think Linus mentions it in the video: Accessing memory from the other CPU is slower. If some cores have no local memory it essentially has to go the long route all the time. Any threads that live there will have slow memory access.

 

Did the video show the time for Cinebench? You could run those benchmarks on your machine, see what times you get. I think rendering needs a good bit of memory, so if I'm correct it should have a fairly noticeable effect.

When I ran cinebench, it said I only have 8 cores. My score on 8 cores was 956. I went into task manager to see all my 16 cores (hyperthreading disabled) and they were there. I opened up cpu-z to see that only 8 cores are detected which is very weird and it only gives options for one socket (even though I have two). Is there anyway to see if I have all my cores or something is wrong?image.thumb.png.c19d9e2282f7ea22e3a309ac9c9ca610.pngimage.thumb.png.e50f24e6ce3eff5b62701f1abcca3b2f.png

 

Link to post
Share on other sites

8 hours ago, artavasdes said:

When I ran cinebench, it said I only have 8 cores. My score on 8 cores was 956. I went into task manager to see all my 16 cores (hyperthreading disabled) and they were there. I opened up cpu-z to see that only 8 cores are detected which is very weird and it only gives options for one socket (even though I have two). Is there anyway to see if I have all my cores or something is wrong?

I suspect something is wrong. CPU-Z only shows one CPU at a time but you should be able to switch between them with the "Selection" drop down in the lower left (where it says Socket #1).

 

Do you see activity on all 16 cores in task manager or are 8 of them idle all the time? You could try enabling HT again, see if it changes anything in CPU-Z. You could also possibly try HWMonitor to see if it can see all 16 cores.

 

Did you have the option to reduce core count in BIOS? Check if you did. If that doesn't help, I guess you'll have to look at the CPUs. Maybe one of them just isn't seated properly or some of the pins are damaged.

Remember to either quote or @mention others, so they are notified of your reply

Link to post
Share on other sites

8 hours ago, Sir0Tek said:

You're not running it on Windows 10 Home, aren't you?

 

And how is the memory organized? Is the defective channel on cpu 1 or on cpu 0?

In case of memory size differences you usually have to give cpu 0 the better/larger amount.

I am running on Windows 10 Home. Below I put a diagram of the motherboard and how the memory dimm slots are configurated. I populated DIMM A1, B1, and D1 (they are for CPU 0). I couldn't populate C1 because I get memory errors when I do that. For CPU 0, I have E1 and F1 populated. So CPU 0 has 48gb of ram and CPU 1 there is 32gb of ram. I haven't tried populating DIMMs like A2, B2 or G1 (those are just examples). I will try it once I get a chance this weekend but I think I had tried with A2 and B2 and got memory errors (I will still try).

216829109_Screenshotfrom2020-01-0616-01-10.png.51ea446df6cbdf1640141b5df2254d06.png

Link to post
Share on other sites

You're trying to drive high refresh rates with a CPU that has a slightly lower IPC and clock than Zen+ chips. It's a CPU bottleneck, especially in something like R6S that'll run fast af with a Radeon VII. 

Gaming PC NAS Laptop Workstation

CPU: i5 12600KF 6P+4E Ryzen 7 3700X M4 SoC 4P+6E Xeon X5690 6c12t

Cooler: Noctua NH-D15S Wraith Stealth w/NF-A9 Passive Apple CPU Cooler

Motherboard: ASRock Z690 ITX/ax ASUS Pro B550M-C/CSM Apple J713AP Mac-F221BEC8 (Mac Pro 5,1)

RAM: 2x16GB 3600Mhz DDR4 2x16GB 2400MHz DDR4 24GB Micron LPDDR5 4x8GB 1333MHz ECC DDR3

GPU: Sapphire Pulse Radeon 9060 XT 16GB Radeon WX2100 M4 SoC 10C Radeon RX 5700

Storage: 1TB MP34 + 2TB P41 500GB SSD + 2x4TB IronWolf Pro in ZFS Mirror Apple AP0512Z 1TB Crucial MX500

ODD: LG WH14NS40 None LG GP65NB60 USB DVD Writer Don't know

PSU: EVGA 850W GM Silverstone SST-TX300 53.8Wh LiPo Battery Delta DPS-980BB

Case: Silverstone Sugo 14 Dell Inspiron 530S Mac16,12 chassis (13" MBA) 2009-2012 Mac Pro "Cheese Grater"

OS: Gentoo Linux TrueNAS Scale macOS 26 Tahoe Fedora Linux

 

Display: LG 27UK650-W (4K 60Hz IPS panel)

Mouse: EVGA X17

Keyboard: Corsair K55 RGB

 

Mobile/Work Devices: 14" M5P MacBook Pro (work) - iPhone 17 Pro - Apple Watch S11

 

Other Misc Devices: iPod Video (Gen 5.5E, iFlash Solo w/128GB SD Card, Rockbox), Nintendo Switch

 

Vehicles: 2002 Ford F150, 2003 Harley-Davidson Sportster 1200, 2022 Kawasaki KLR650, 1994 DR350SE

Link to post
Share on other sites

12 hours ago, Eigenvektor said:

I suspect something is wrong. CPU-Z only shows one CPU at a time but you should be able to switch between them with the "Selection" drop down in the lower left (where it says Socket #1).

 

Do you see activity on all 16 cores in task manager or are 8 of them idle all the time? You could try enabling HT again, see if it changes anything in CPU-Z. You could also possibly try HWMonitor to see if it can see all 16 cores.

 

Did you have the option to reduce core count in BIOS? Check if you did. If that doesn't help, I guess you'll have to look at the CPUs. Maybe one of them just isn't seated properly or some of the pins are damaged.

No dropdown appears when I click on the Socket option. Windows CMD shows only 8 cores which is very weird. When I go into task manager performance sections, only 8 cores appear. I haven't tried going into something like resource monitor but I will check that once I get a chance this weekend. I'll also try HWmonitor. There is no option in BIOS to reduce core count. If none of this works, I will try and a take a look at the CPUs. While I am looking at the cpus, is there anything you want me to do? I want to try maybe swapping the CPUs and putting them in different sockets.

Link to post
Share on other sites

2 minutes ago, Zando Bob said:

You're trying to drive high refresh rates with a CPU that has a slightly lower IPC and clock than Zen+ chips. It's a CPU bottleneck, especially in something like R6S that'll run fast af with a Radeon VII. 

Compared to other benchmarks like Linus's:

I am suffering a severe performance loss no one else with these dual cpus are suffering. There is a video in the first post, that one even had a worse and older cpu performing better in games than mine. He used a 5700 xt which is quite similar to the radeon vii in gaming performance. Also, the cpu runs at a 45% load so there can't be that great of a cpu bottleneck (except for bandwidth and IPC as you pointed out but that should not take a 50% toll on performance)

 

 

Link to post
Share on other sites

4 minutes ago, artavasdes said:

Also, the cpu runs at a 45% load so there can't be that great of a cpu bottleneck (except for bandwidth and IPC as you pointed out but that should not take a 50% toll on performance)

What's the usage on individual cores? Most games will not use all the cores, you'll likely have higher load on a few, but when averaging them out it comes to a low total. I could run Cinebench R20 on two cores and slap them both to max load, but overall CPU usage would still be low because I have 8c/16t and only 2 are being used. 

Gaming PC NAS Laptop Workstation

CPU: i5 12600KF 6P+4E Ryzen 7 3700X M4 SoC 4P+6E Xeon X5690 6c12t

Cooler: Noctua NH-D15S Wraith Stealth w/NF-A9 Passive Apple CPU Cooler

Motherboard: ASRock Z690 ITX/ax ASUS Pro B550M-C/CSM Apple J713AP Mac-F221BEC8 (Mac Pro 5,1)

RAM: 2x16GB 3600Mhz DDR4 2x16GB 2400MHz DDR4 24GB Micron LPDDR5 4x8GB 1333MHz ECC DDR3

GPU: Sapphire Pulse Radeon 9060 XT 16GB Radeon WX2100 M4 SoC 10C Radeon RX 5700

Storage: 1TB MP34 + 2TB P41 500GB SSD + 2x4TB IronWolf Pro in ZFS Mirror Apple AP0512Z 1TB Crucial MX500

ODD: LG WH14NS40 None LG GP65NB60 USB DVD Writer Don't know

PSU: EVGA 850W GM Silverstone SST-TX300 53.8Wh LiPo Battery Delta DPS-980BB

Case: Silverstone Sugo 14 Dell Inspiron 530S Mac16,12 chassis (13" MBA) 2009-2012 Mac Pro "Cheese Grater"

OS: Gentoo Linux TrueNAS Scale macOS 26 Tahoe Fedora Linux

 

Display: LG 27UK650-W (4K 60Hz IPS panel)

Mouse: EVGA X17

Keyboard: Corsair K55 RGB

 

Mobile/Work Devices: 14" M5P MacBook Pro (work) - iPhone 17 Pro - Apple Watch S11

 

Other Misc Devices: iPod Video (Gen 5.5E, iFlash Solo w/128GB SD Card, Rockbox), Nintendo Switch

 

Vehicles: 2002 Ford F150, 2003 Harley-Davidson Sportster 1200, 2022 Kawasaki KLR650, 1994 DR350SE

Link to post
Share on other sites

Just now, artavasdes said:

I want to try maybe swapping the CPUs and putting them in different sockets.

That sounds like a good plan. If it's not too much work, you could probably try them individually. If the PC fails to boot with one of them you'd know for sure. Unless you already know which one of them is having an issues, that would probably save some time.

 

Other than that, I'd inspect the CPU and socket for physical damage, bent pins that kind of stuff. It's also possible the board itself is having an issue. In that case both CPUs should work on their own. Or, when swapping them I'd expect the error to affect the other CPU that was working before.

Remember to either quote or @mention others, so they are notified of your reply

Link to post
Share on other sites

Afaik Windows 10 Home is 1 cpu only, so it's possible you can't make use of cpu1 anyway, unless going 1 step higher in version (or switch the OS).

Next, lga2011 does support dual and quadchannel for sure, but maybe triple channel will need a different setup where you need C1 populated, I guess. So populating D1 could have an adverse effect on speed.

 

Link to post
Share on other sites

3 minutes ago, Sir0Tek said:

Afaik Windows 10 Home is 1 cpu only, so it's possible you can't make use of cpu1 anyway, unless going 1 step higher in version (or switch the OS).

That seems correct. https://answers.microsoft.com/en-us/windows/forum/windows_10/windows-10-versions-cpu-limits/905c24ad-ad54-4122-b730-b9e7519c823f?auth=1. You need to look at all answers, the accepted answer is a bit to generic on that part and doesn't mention differences between editions.

 

That would definitely explain why OP can't make use of both CPUs and cores.

Remember to either quote or @mention others, so they are notified of your reply

Link to post
Share on other sites

Tbh, I didn't really expect the answer 'Yes, I'm using W10 home'. So, disabling cores doesn't help you for games but it is a very potent machine nonetheless. Lower IPC can't explain the behaviour on these. Someone just has to lift the brakes.

Getting cpu0 to use quad-channel will definitely help, cpu1 may have reduced dual channel access.

 

Link to post
Share on other sites

47 minutes ago, Zando Bob said:

What's the usage on individual cores? Most games will not use all the cores, you'll likely have higher load on a few, but when averaging them out it comes to a low total. I could run Cinebench R20 on two cores and slap them both to max load, but overall CPU usage would still be low because I have 8c/16t and only 2 are being used. 

The CPU usage is maxed out on the first two cores, but when I disable cores, the performance gets even worse. Even if I give it four cores, I loose like 50-60% of my FPS. Also, that doesn't explain why others are able to get good performance with a small toll. In the LTT video I replied to you with, there is a 6% difference in games between the 3rd gen ryzen cpu and the two xeons I have.

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×