Shadow Of The Tomb Raider - CPU Performance and general game benchmark discussions

[Taraquin]

16-17-16, I can put 16-16-16, but its not really 100% stable. If you got some advice on how to tighten even more, I am ready to listen

Are the xmp-timings 16-17-16? That is very unusual. What are you able to run tRFC at? Post your zentimings and I can try to help

 

[Felix123BU]

Are the xmp-timings 16-17-16? That is very unusual. What are you able to run tRFC at? Post your zentimings and I can try to help

"Are the xmp-timings 16-17-16? That is very unusual", yup, I know, these timings are unusual, which would make me think they are some B-dies that where deemed not the best, hence some looser timings where applied. Trfc I can go as low as 290. Here are also the zen timings:

 

[Taraquin]

"Are the xmp-timings 16-17-16? That is very unusual", yup, I know, these timings are unusual, which would make me think they are some B-dies that where deemed not the best, hence some looser timings where applied. Trfc I can go as low as 290. Here are also the zen timings:

View attachment 198164

That is B-die, no other can run tRFC that low at 3733. What agesa are you on? 1.2.0.1? Tried running IF above 1866? Your ram voltage is very low, try upping voltage as tRCDRD scales positively with more voltage, tRFC also likes higher voltage.

Try the following: 1.45Vdimm, 900mv vddp, 940mv vddg ccd, 1020 vddg iod, 1110mv soc, 2t instead of gdm and 1t, CL15, tRCD 16, tRP 15, tRAS 31, tRC 46, tRFC 276, tRTP 6, tRDWR 9 or 10. Disable spread spectrum.

If that works and you want to try even tighter try tRCDRD 15, tRP 14, tRC 44, tRFC 264.

If you don't wanna go that much higher on volt set it to 1.4V and do 16-16-16-32-48 288 tRFC.

 

[Felix123BU]

That is B-die, no other can run tRFC that low at 3733. What agesa are you on? 1.2.0.1? Tried running IF above 1866? Your ram voltage is very low, try upping voltage as tRCDRD scales positively with more voltage, tRFC also likes higher voltage.

Try the following: 1.45Vdimm, 900mv vddp, 940mv vddg ccd, 1020 vddg iod, 1110mv soc, 2t instead of gdm and 1t, CL15, tRCD 16, tRP 15, tRAS 31, tRC 46, tRFC 276, tRTP 6, tRDWR 9 or 10. Disable spread spectrum.

If that works and you want to try even tighter try tRCDRD 15, tRP 14, tRC 44, tRFC 264.

If you don't wanna go that much higher on volt set it to 1.4V and do 16-16-16-32-48 288 tRFC.

Thanks for trying to help , will probably try it later today when time allows. Agesa is 1.2.0.1
Yes, tried running IF above 1866, wont work, cant run it without some indecent voltages on the CPU and even then its not even close to stable (though have not tried only IF without the Mem clock in sync, might be that the CPU can run higher IF but the ram cant, will test that) . Spread spectrum is always disabled.
"2t instead of gdm and 1t" hmm, did not think of trying that, but its worth a try
There is one weird thing that happens with my ram, namely, if I up the voltage over 1.4, it becomes unstable, I have to keep it under 1.4 for 100% stability, that is really weird, though I read some people have the same "issue".

The only reason I am trying to tune everything just a bit more is for this test exclusively, bugs me that I cant go over 230FPS in 1080 Lowest, even though I don't play anything at 1080

Had 10 minutes free, GDM off now, and 1T cause with GDM off it goes to Auto only, and it stuck with 1T, tested again, behold..... 230 again

Will play some more around later

tRCD 16 brought me finally over 230fps, to 237 though the CPU scores remain basically identical

Its not 100% stable at the current voltage, though game stable.

 

[Taraquin]

Thanks for trying to help , will probably try it later today when time allows. Agesa is 1.2.0.1
Yes, tried running IF above 1866, wont work, cant run it without some indecent voltages on the CPU and even then its not even close to stable (though have not tried only IF without the Mem clock in sync, might be that the CPU can run higher IF but the ram cant, will test that) . Spread spectrum is always disabled.
"2t instead of gdm and 1t" hmm, did not think of trying that, but its worth a try
There is one weird thing that happens with my ram, namely, if I up the voltage over 1.4, it becomes unstable, I have to keep it under 1.4 for 100% stability, that is really weird, though I read some people have the same "issue".

The only reason I am trying to tune everything just a bit more is for this test exclusively, bugs me that I cant go over 230FPS in 1080 Lowest, even though I don't play anything at 1080

Had 10 minutes free, GDM off now, and 1T cause with GDM off it goes to Auto only, and it stuck with 1T, tested again, behold..... 230 again

View attachment 198173

Will play some more around later

tRCD 16 brought me finally over 230fps, to 237 though the CPU scores remain basically identical

Its not 100% stable at the current voltage, though game stable.

View attachment 198175

Try 1.39V, that will probably stabilize tRCDRD. Try tRP 15, tRC 47 and tRFC 282, tRTP 6.

 

[toilet pepper]

Not sure, only got one soc voltage at mye setup.

Done any ram tuning? 250fps+ should be doable with a bit of tuning

Here's what I have currently. They are bdies and I just upgraded from a Ryzen 3600 and know the RAM and mobo are perfectly capable of going higher.

For some reason the 5800x doesnt want to go any higher. I tried DRAM calculator and it wouldnt boot. I just plugged this in and called it a day.

 

[Felix123BU]

Try 1.39V, that will probably stabilize tRCDRD. Try tRP 15, tRC 47 and tRFC 282, tRTP 6.

The voltage weirdness continues, at 1.37 I could run the benchmark, at 1.39 the game crashed completely within seconds put it back to 1.37, game runs fine.
And I would go higher with voltage for testing purposes, only that as said if I go higher, the higher I go, the more unstable it gets, weird. Tried 1.4, then 1.41, got worse each time
What about MEM VTT voltage?

 

[Taraquin]

Here's what I have currently. They are bdies and I just upgraded from a Ryzen 3600 and know the RAM and mobo are perfectly capable of going higher.

For some reason the 5800x doesnt want to go any higher. I tried DRAM calculator and it wouldnt boot. I just plugged this in and called it a day.


View attachment 198179

Try trrds 4, trrdl 6, tfaw 16, trc 48, twr 12, trtp 6. If that works you can try trp 15, tras 31, trc 47 and trfc 282.

The voltage weirdness continues, at 1.37 I could run the benchmark, at 1.39 the game crashed completely within seconds put it back to 1.37, game runs fine.
And I would go higher with voltage for testing purposes, only that as said if I go higher, the higher I go, the more unstable it gets, weird. Tried 1.4, then 1.41, got worse each time
What about MEM VTT voltage?

Dunno about mem vtt. Are rgb on ram active? If it is and you dont care about it, try disabling it as it increases temp on ram, this might make it possible to run higher voltage.

 

[Felix123BU]

Try trrds 4, trrdl 6, tfaw 16, trc 48, twr 12, trtp 6. If that works you can try trp 15, tras 31, trc 47 and trfc 282.


Dunno about mem vtt. Are rgb on ram active? If it is and you dont care about it, try disabling it as it increases temp on ram, this might make it possible to run higher voltage.

As far as temps go, they are pretty cool, during normal operations it never goes above 55c, that should not be an issue. Anyhow, i fixed my 'issue' of not seeing anything over 230fps, for now the itch is scratched.

Thanks a lot for the advices, will probably revisit the ram oc topic one day.

And speaking of Cpu performance in SOTR, whatever i did, the cpu stats basically remain the same, when there was a Fps increase, the section that increased was the Gpu section, i wonder if thats a difference of architectures since you have a Amd Nvidia combo, i have a Amd Amd combo, so the game engine might behave differently.

 

[rtwjunkie]

I thought we already had a very multipage SOTTR benchmark thread?

 

[Deleted member 202104]

I thought we already had a very multipage SOTTR benchmark thread?

I believe this new thread was set up to focus on CPU (1080 Lowest) vs overall system (1080/1440 Highest) in the other thread.

(There was a rather aggressive user who had popped into the other thread disregarding the parameters for posting results, prompting this one.)

 

[rtwjunkie]

I believe this new thread was set up to focus on CPU (1080 Lowest) vs overall system (1080/1440 Highest) in the other thread.

(There was a rather aggressive user who had popped into the other thread disregarding the parameters for posting results, prompting this one.)

I’m confused why a benchmark would focus on how no one actually plays the game? The best purpose of a BM is to test somewhat how a system plays in the associated game, including visuals.

 

[Felix123BU]

I’m confused why a benchmark would focus on how no one actually plays the game? The best purpose of a BM is to test somewhat how a system plays in the associated game, including visuals.

Because some people think something is good, others think something else is good. We had the same discussion on the other Shadow of the Tomb Raider benchmark, where 1080 Highest was the norm, some people where not pleased and wanted a benchmark for Lowest, hence I created this opposite threat so we can compare and discuss setting that let the CPU shine instead of the GPU. But there will always be people who don't like either one or the other, and in the end not all can be pleased

And to answer your question, why? Because its fun to test the same thing from a different perspective

I believe this new thread was set up to focus on CPU (1080 Lowest) vs overall system (1080/1440 Highest) in the other thread.

(There was a rather aggressive user who had popped into the other thread disregarding the parameters for posting results, prompting this one.)

To quote the user you are referring to: "The whole world uses this game as a CPU benchmark", we can not disappoint the whole world

 

[Taraquin]

As far as temps go, they are pretty cool, during normal operations it never goes above 55c, that should not be an issue. Anyhow, i fixed my 'issue' of not seeing anything over 230fps, for now the itch is scratched.

Thanks a lot for the advices, will probably revisit the ram oc topic one day.

And speaking of Cpu performance in SOTR, whatever i did, the cpu stats basically remain the same, when there was a Fps increase, the section that increased was the Gpu section, i wonder if thats a difference of architectures since you have a Amd Nvidia combo, i have a Amd Amd combo, so the game engine might behave differently.

Temp is your stability issue. B-die prefers beliw 50C and tends to be unstable if timings are tight above 50C. When I put a fan directly at ram I could up voltage to 1.5V and lower several timings. With no fan ram overheats above 1.45V. I changed chassis now with better cooling so it might be better now.

 

[Deleted member 202104]

I’m confused why a benchmark would focus on how no one actually plays the game?

I think W!zzard asked the same question as well (or at least how it related to game play).

I think it's in the same spirit of why 720p benchmarks are posted - CPU performance.

The best purpose of a BM is to test somewhat how a system plays in the associated game, including visuals.

I agree that's a really good purpose. I play 1440p and as high quality settings as decent frame rates allow.

From doing this, I saw the actual results in my system from swapping to a lower frequency (3600 to 3200) but lower latency (18-22-22-42 to 14-14-14-34) RAM and changing CPU power limits.

It was more interesting than a Cinebench run.

 

[rtwjunkie]

Because some people think something is good, others think something else is good. We had the same discussion on the other Shadow of the Tomb Raider benchmark, where 1080 Highest was the norm, some people where not pleased and wanted a benchmark for Lowest, hence I created this opposite threat so we can compare and discuss setting that let the CPU shine instead of the GPU. But there will always be people who don't like either one or the other, and in the end not all can be pleased

Thanks for the thorough explanation. I went to see what was up. It seems rather than report and threadban someone who can’t follow the thread rules and test his system on a benchmark that TPU uses as a GAME benchmark, he gets catered to and gets the CPU benchmark thread he wants created?

Ok, then, don’t worry I’ll just see myself out.

 

[Felix123BU]

Temp is your stability issue. B-die prefers beliw 50C and tends to be unstable if timings are tight above 50C. When I put a fan directly at ram I could up voltage to 1.5V and lower several timings. With no fan ram overheats above 1.45V. I changed chassis now with better cooling so it might be better now.

It could be, you might be onto something, I never bothered about temps since 55 was the absolute max in a stress test after 1 hour (and that's max, normally its around 45 ingame), the one thing that sort of would not apply to this theory is that after a ram setting change and a reboot my ram idles at around 33c, and it does not heat up that fast, but who knows, the temp sensors on the sticks could be on one part of the stick and another gets hot really fast and is not covered by the sensors. But I like the idea, have tried so many things, putting a fan above the sticks for a couple of minutes to see if there is a difference would be a worthy test

I think W!zzard asked the same question as well (or at least how it related to game play).

I think it's in the same spirit of why 720p benchmarks are posted - CPU performance.



I agree that's a really good purpose. I play 1440p and as high quality settings as decent frame rates allow.

From doing this, I saw the actual results in my system from swapping to a lower frequency (3600 to 3200) but lower latency (18-22-22-42 to 14-14-14-34) RAM and changing CPU power limits.

It was more interesting than a Cinebench run.

Deffo more interesting that a CB run

Thanks for the thorough explanation. I went to see what was up. It seems rather than report and threadban someone who can’t follow the thread rules and test his system on a benchmark that TPU uses as a GAME benchmark, he gets catered to and gets the CPU benchmark thread he wants created?

Ok, then, don’t worry I’ll just see myself out.

I was not really to please him, he could have pleased himself with a thread, but more out of curiosity about how certain CPUs reflect in the benchmark, a thing not clearly visible in a higher resolution scenario, especially since this game is quite CPU dependent. And that guy did not want this thread, he wanted THAT thread

I can generally agree that game CPU benchmarks are not the best use of a game, but they have a place and people who like to test these things.

 

[toilet pepper]

Try trrds 4, trrdl 6, tfaw 16, trc 48, twr 12, trtp 6. If that works you can try trp 15, tras 31, trc 47 and trfc 282.


Dunno about mem vtt. Are rgb on ram active? If it is and you dont care about it, try disabling it as it increases temp on ram, this might make it possible to run higher voltage.

Thanks for the tip. Here's what I got with your suggestion. I'm heavily thermal limited as the rig is in an ITX case.

here's my timings on the same rig with a ryzen 3600. I can't do it with my 5800x.

 

[Taraquin]

Thanks for the tip. Here's what I got with your suggestion. I'm heavily thermal limited as the rig is in an ITX case.

View attachment 198248
View attachment 198247


here's my timings on the same rig with a ryzen 3600. I can't do it with my 5800x.

View attachment 198249

3% more fps Further suggestions: twtrl 12 or 10, trrdl 4, twrrd 1. If you can run voltage higher on the ram you can tighten further. With 1.45V you can try 15-15-15-14-30 44 trc and 264 trfc. That should give you a good boost. You might be able to run cl14, but that depends on how good binned your ram is.

If you have the same ram on the 3600 try trp 15, tras 33, trc 48, trfc 288, trrds 4, trrdl 6, tfaw 16, twr 12, trtp 6, that mught work at 1.4V or a bit higher.

 

[steevebacon]

 

[Det0x]

 

[Space Lynx]

View attachment 204512

300 fps... mighty impressive...

 

[Det0x]

300 fps... mighty impressive...

No need to stop there..

 

[Felix123BU]

No need to stop there..
View attachment 205978
View attachment 205979

350, go big or go home

bet that 14-14-14 memory helps a lot mine cant go lower than 16 at 3800mhz 1:1:1

 

[CGi-Quality]