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

[Taraquin]

Yup, same for me, @2000 I get whea interconnect bus errors whatever voltage I would push through it, whatever timings, tried up to 1.65, did not help. Might be the CPU, might be the RAM for me.

My sticks are weird, they get unstable at high voltages, anything above 1.45v will spit out errors, a setting that is 100% stable at 1.38v will be unstable at 1.5v, regardless of any other consideration.
But it does need only 1.38 for 3800 CL16, at least that is good

This is what I can do max, and this I use on a daily basis. Not to bad for a meeh ram kit which is rated for max 3333 CL16-17-16 @1.35v

View attachment 212540View attachment 212541


"Best I have got so far is 16-16-15 and 282 trfc" - at 2000 FCLK? Cool Is it stable?

Yeah, ran 25 TM5s, I gave no wheas. No crashes since May (when I got agesa 1.2.0.1, now on 1.2.0.3A). Too bad my ram us a shitty bin. 1t is out of the question and I seem to need 0.05V extra at same timings compared to others.

 

[Mbellantoni]

Yup, same for me, @2000 I get whea interconnect bus errors whatever voltage I would push through it, whatever timings, tried up to 1.65, did not help. Might be the CPU, might be the RAM for me.

My sticks are weird, they get unstable at high voltages, anything above 1.45v will spit out errors, a setting that is 100% stable at 1.38v will be unstable at 1.5v, regardless of any other consideration.
But it does need only 1.38 for 3800 CL16, at least that is good

This is what I can do max, and this I use on a daily basis. Not to bad for a meeh ram kit which is rated for max 3333 CL16-17-16 @1.35v

View attachment 212540View attachment 212541


"Best I have got so far is 16-16-15 and 282 trfc" - at 2000 FCLK? Cool Is it stable?

Its deffinently your flck causing the whea 19's. All we can really do is hope an agesa update eliminates them. That is weird about your ram not being able to push further voltages. Seems like you have a nice healthy profile there though

Yeah, ran 25 TM5s, I gave no wheas. No crashes since May (when I got agesa 1.2.0.1, now on 1.2.0.3A). Too bad my ram us a shitty bin. 1t is out of the question and I seem to need 0.05V extra at same timings compared to others.

I cant even run gdm off at all on my sticks. It needs to be on. 1t or 2t results in memory so unstable that my bios can even crash. I dont even try anymore due to the risk of corrupting my OS. Im pretty stable now at 264 trfc at 1.60v but im probably going to drop my voltage down to 1.57.5 today and run a long hcimemtest. I had my trfc at like 234 before and i would get an error regardless of voltages.

 

[Felix123BU]

Its deffinently your flck causing the whea 19's. All we can really do is hope an agesa update eliminates them. That is weird about your ram not being able to push further voltages. Seems like you have a nice healthy profile there though


I cant even run gdm off at all on my sticks. It needs to be on. 1t or 2t results in memory so unstable that my bios can even crash. I dont even try anymore due to the risk of corrupting my OS. Im pretty stable now at 264 trfc at 1.60v but im probably going to drop my voltage down to 1.57.5 today and run a long hcimemtest. I had my trfc at like 234 before and i would get an error regardless of voltages.

Yes, I also think its the FLCK, 2000 is said to be rather rare even for the 5000 series, and I also tried some super relaxed timings, same errors. And the fact that my ram also does not like high voltages is explained in some forums and ram oc guides, some sticks just do not play nice with high voltage. Could also be the mobo, but others have pushed ram much much higher than me on the same model, sooo.....

I cant really complain about performance, its where I want it, the only reason to get faster sticks would be to play and tune them, that would be fun, but spending a couple of hundreds on a really good pair is just a waste of money for me Another reason would be to see how much I can push this benchmark, if I where to look at the GPU averages, I get 542, and CPU renderer 352, I would sort of want to see how much a super fast ram would elevate the CPU game, and subsequently the whole score

As for GDM off, I can run it at 3600mhz max, with some timings a bit looser, but the results are basically the same as GDM on plus higher frequency, so no point for me, as anything higher than 3600 will not work with GDM off for me, would probably need lots higher voltage.

 

[Mbellantoni]

Yea theres not many people running 2000flck error free at the moment. I currently have mine on a stability test. I tightened the timings to 14-16-14-28-38 and the trfc to 252. I dropped the voltage to 1.58.5 and it looks promising so far. I played around with subtimings but im trying to drop voltage and at this point im only gettin very minor benefits with decreased stability for every timing change so i left it alone. This is how im cooling my ram, my buddy is going to design me a 3d printed bracket for a high flow 120mm fan like a noctua. But as it currently sits this old wraith fan keeps my ram at a cool 35c max under heavy loads.

 

[Taraquin]

Could it be easier with ram\less wheas for me due to 2-dimm only motherboard? 4-dimms which almost all MBs have add complexity with signals etc. If ITX-MBs and the very few mATX-boards with only 2 dimms also have fewer wheas that might be a thing? I had a few wheas in May before I tuned voltages and got the latest agesa. Dunno what did it, but I run VDDG CCD and VDDP as low as I can, also keep VDDG IOD and SOC at the lowest I can before performance drops.

 

[Mbellantoni]

Could it be easier with ram\less wheas for me due to 2-dimm only motherboard? 4-dimms which almost all MBs have add complexity with signals etc. If ITX-MBs and the very few mATX-boards with only 2 dimms also have fewer wheas that might be a thing? I had a few wheas in May before I tuned voltages and got the latest agesa. Dunno what did it, but I run VDDG CCD and VDDP as low as I can, also keep VDDG IOD and SOC at the lowest I can before performance drops.

Possibly. I know evgas new mobo is only coming with 2 dimm slots. Im not sure though. I would imagine in general a 2x16gb setup will clock better than an 4x8 because its not providing power to 4 different dimms.

 

[Zyll Goliat]

This is a bit strange....Can someone explain to me how is it possible to have more frames rendered and actually lower avg fps?

 

[Felix123BU]

This is a bit strange....Can someone explain to me how is it possible to have more frames rendered and actually lower avg fps?

Because even if it says Frames Per Second its not really frames per second, I also do not really know how the so called FSP is calculated, but its not true FPS

 

[Mbellantoni]

I went back to the drawing board for the final time. was able to drop the voltage to 1.58v and luckily was able to have someone i consider an expert overclocker take a look at my previous profile. i was able to tighten some timings down and not lose stability at all. as far as game performance goes i think im at the point of diminishing returns though. ive had better runs with slightly worse timings. so its coming down to margin of error between runs right now and think its just based off of the cpus and gpus performance per run rather than the ram timings making an actual difference. it is slightly faster though in any synthetic benchmark like geekbench, membench, geekbench etc.. im pretty much spent at this point lol.

 

[Felix123BU]

@Mbellantoni , funny, I was doing something similar just now

Put the latest bios to see if there where some change, not a lot except I can now run GDM off at 2T stable with basically the same performance as GDM on 1T. GDM off 1T does not give WHEA errors, just apps randomly closing for no good reason

I was thinking also about all the time lost with RAM tuning, so did a comparison vs pure stock, XMP and 3800 CL16 tuned.

Here goes, pure stock Ram 2133 CL15

XMP Ram 3333 CL16

difference vs pure stock +35 FPS

3800 CL16 tuned as far as I could

difference vs pure stock +58 FPS
difference vs XMP +23 FPS


I could probably get a lot more extra fps here with a 5950x or a 10900k plus some 3800+ CL14 tuned ram sticks for, but realistically in the real world and at higher resolutions the differences would be to small to really matter if you have at least a decent ram setup and XMP turned on

 

[Taraquin]

@Mbellantoni , funny, I was doing something similar just now

Put the latest bios to see if there where some change, not a lot except I can now run GDM off at 2T stable with basically the same performance as GDM on 1T. GDM off 1T does not give WHEA errors, just apps randomly closing for no good reason

I was thinking also about all the time lost with RAM tuning, so did a comparison vs pure stock, XMP and 3800 CL16 tuned.

Here goes, pure stock Ram 2133 CL15

View attachment 212730View attachment 212731

XMP Ram 3333 CL16

View attachment 212732View attachment 212733

difference vs pure stock +35 FPS

3800 CL16 tuned as far as I could

View attachment 212734View attachment 212735

difference vs pure stock +58 FPS
difference vs XMP +23 FPS


I could probably get a lot more extra fps here with a 5950x or a 10900k plus some 3800+ CL14 tuned ram sticks for, but realistically in the real world and at higher resolutions the differences would be to small to really matter if you have at least a decent ram setup and XMP turned on

If you could run 4000cl16 with higher voltage like me that could yield up to 4ns in aida and up to 20fps i SOTTR is my guestimate.

 

[Zyll Goliat]

I turned HT off and gain even few more frames from 130 to 134.....

 

[Felix123BU]

If you could run 4000cl16 with higher voltage like me that could yield up to 4ns in aida and up to 20fps i SOTTR is my guestimate.

If it would work, possible, have a hunch that FCLK 2000 is not stable on my CPU, might be wrong though, will probably test with super loose timings to exclude CPU instability

 

[Taraquin]

If it would work, possible, have a hunch that FCLK 2000 is not stable on my CPU, might be wrong though, will probably test with super loose timings to exclude CPU instability

Find the IF-limit first Try everything on auto with 1.35V ram and up to 3866/1933, 3933/1966 etc. If you have agesa 1.2.0.1 or newer there is a much bigger chance of hitting 4000/2000+ vs agesa 1.1.x.x or lower. If it won't boot try soc at 1.14V or a bit higher, vddg iod at 1.06, vddg ccd 0.94, vddp 0.9, set procodt to 28-37 and let drvstr/cad etc stay at auto. Gdm and 1t is also recommended.

At 1.4V 4000cl17-17-16 300 tRFC is 100% stable, and perf is not that much worse vs 4000cl16. But you must use 2t and gdm off for cl17.

 

[Felix123BU]

Find the IF-limit first Try everything on auto with 1.35V ram and up to 3866/1933, 3933/1966 etc. If you have agesa 1.2.0.1 or newer there is a much bigger chance of hitting 4000/2000+ vs agesa 1.1.x.x or lower. If it won't boot try soc at 1.14V or a bit higher, vddg iod at 1.06, vddg ccd 0.94, vddp 0.9, set procodt to 28-37 and let drvstr/cad etc stay at auto. Gdm and 1t is also recommended.

At 1.4V 4000cl17-17-16 300 tRFC is 100% stable, and perf is not that much worse vs 4000cl16. But you must use 2t and gdm off for cl17.

2000 boots and gets into Windows, but produces lots of WHEA errors (Bus interconnect), anyway, will probably test sometime. At one point in time 3600 was max, then a couple of bioses later 3800 was perfect, maybe 4000 is possible

 

[Taraquin]

2000 boots and gets into Windows, but produces lots of WHEA errors (Bus interconnect), anyway, will probably test sometime. At one point in time 3600 was max, then a couple of bioses later 3800 was perfect, maybe 4000 is possible

Try with the voltages I recommended, that can resolve wheas on some setups

 

[Cheese_On_tsaot]

View attachment 212134


Something is clearly off with this benchmark, my RAM is just 200mhz slower than the 5700 XT combo.

Hmmm Windows build difference, mine is 1 iteration older.

New result with Patriot Viper 3733mhz CL17 RAM at XMP profile.

At 3800 16-19-19-38-57 1.45v

 

[Taraquin]

New result with Patriot Viper 3733mhz CL17 RAM at XMP profile.
View attachment 212942

At 3800 16-19-19-38-57 1.45v

View attachment 212949

I`m unable to read the CPU game avg due to low quality image. Remember that we test at 1080p lowest

 

[Cheese_On_tsaot]

I`m unable to read the CPU game avg due to low quality image. Remember that we test at 1080p lowest

Yes but a 2060 is the bottleneck in some of the bench at 1080p so 800x600 is a real CPU load.

 

[Felix123BU]

Yes but a 2060 is the bottleneck in some of the bench at 1080p so 800x600 is a real CPU load.

Could be, but everybody else did the test at 1080p, so in the interest of keeping this thread clean, please be so kind and re-upload the bench at 1080.

Thank you

 

[Cheese_On_tsaot]

Could be, but everybody else did the test at 1080p, so in the interest of keeping this thread clean, please be so kind and re-upload the bench at 1080.

Thank you

CPU Performance​

Rules don't apply when you are going against the point of your own thread by bringing in this illusionary ruling which removes the CPU benchmark and then becomes CPU / GPU.

I turned HT off and gain even few more frames from 130 to 134.....

View attachment 212769

You are GPU bound almost half of the time in the benchmark.

The OP only considered his own hardware when making a CPU bench thread for people with a gamut of different GPU options.

Respect would be realizing that not everyone has a GPU that is fully untapped at 1080p hitting almost 300 FPS.


Respect is respectfully setting out a thread to include those who wish to join in.

As before in this thread though the OP dismissed a submission due to their own ignorance to another user, the user was correct, not the OP.


The people here make the thread.

 

[Felix123BU]

CPU Performance​

Rules don't apply when you are going against the point of your own thread by bringing in this illusionary ruling which removes the CPU benchmark and then becomes CPU / GPU.


You are GPU bound almost half of the time in the benchmark.

The OP only considered his own hardware when making a CPU bench thread for people with a gamut of different GPU options.

Respect would be realizing that not everyone has a GPU that is fully untapped at 1080p hitting almost 300 FPS.


Respect is respectfully setting out a thread to include those who wish to join in.

As before in this thread though the OP dismissed a submission due to their own ignorance to another user, the user was correct, not the OP.


The people here make the thread.

I see, so you, being the only one who did not post a 1080p lowest result, are "the people"

I am again very kindly asking that you, the people, respect all other previous posters in this thread by providing something meaningful that we all can compare against.

You probably know that, as you said, there are multiple possible setups, and some will see a bottleneck at 1080p, other at 720p, other even lower than that if their GPU is to weak.

The point of this thread is to compare 1080p lowest and discuss about how to improve or what holds us back at this specific setting. If each would post whatever settings he feels like, this thread would have no meaning at all.

 

[the54thvoid]

LQ'd the 'argument'.

For reference, we've been here before in the benchmark threads. If an OP states a rule, and the thread runs fine, we stick to that rule. Otherwise, there is no benchmark, only random, unrelated performances. And for clarity, any bickering about why the settings are wrong will get you barred. This isn't a democracy folks; this is the rules:

Fullscreen
Exclusive Fullscreen
DirectX 12
DLSS OFF
Vsync OFF
Resolution 1920 X 1080
Anti-Aliasing OFF

Graphic Settings - Lowest Profile (please leave it at Lowest without any changes for the purpose of this test)

Obey, or leave.

 

[Zyll Goliat]

I mean it's really easy to see IF your GPU is the bottleneck in this benchmark....You can always do the test in lower resolution(880x600) and compare your results...here I did just for fun and actually the results are more or less identical(in my case) as you can see bellow

So HT off give me 4 fps(134) and lowest res 800x600 only 2 fps(132) which is almost margin of error......

 

[Cheese_On_tsaot]

Lets see.

Prior result at truly CPU bound 800x600. 200 fps

Following OP's ruling and the moderator.

When rules are rules but no logic or intelligence is to be found with the rules, we get stupid people domineering the inteligent.

You cannot fix stupid and I fully agree, you can try to tell them and guide them, but they will likely throw it in your face with even more perceived superiority because "Rules"

Here is my result following your rules, my previous ones make me correct and are to the point of the thread which is a CPU benchmark, I am smart, you are not and there is nothing you can do about it other than try to silence me


Have a good day.