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

[Tanzmusikus]

It also depends on the RAM topology of the board whether you can achieve high frequencies with four RAM modules.
Your ASUS X570 TUF Plus (Wifi) has a DaisyChain topology, so you can achieve higher frequencies with two modules.
With four modules it would be less, but you might end up with similar transfer rates.

 

[Mbellantoni]

It also depends on the RAM topology of the board whether you can achieve high frequencies with four RAM modules.
Your ASUS X570 TUF Plus (Wifi) has a DaisyChain topology, so you can achieve higher frequencies with two modules.
With four modules it would be less, but you might end up with similar transfer rates.

You are correct. Slapped the new kit in along with my trident z neos and it wont even post at 4000 even with high voltage and the sloppiest of timings. The mem controller simply cant take it or its my mobo itself.

Anyways ill be finishing up 3800 14-15-14-21 soon on hci memtest . its looking promising and then ill run a few sotr benchmarks and post the results

dropped in the ripjaws with the neos for 32 gigs total. played around with some voltages/ cad bus and odt's and got this stable. the top run being 3800cl14 @ 32 gigs and the bottom being 4000cl14 at 16gigs. the bottom result was more of a freak run and it was just the highest score i have ever gotten but its not really consistent (usually 230-233fps) its the screenshot i had so i posted it. otherwise they are both in line with each other. the dual rank bonus from having all the dimm slots filled negating the 200mhz and 100 fclock bonus from running only 2 sticks and vise versa. the timings are pretty much the same besides a change of twrrd from 1 to 3 for stability. i know i have the screenshot somewhere but i did a previous run at 3800cl14 @ 16gb same timings and it scored 226. so running dual rank is for sure an upgrade speed for speed and timing for timing. so my conclusion is running a dual ranked ram config is equal in performance a x2 single ranked config with approx. 200mhz and 100flck with the same timings in this benchmark even though the 4000cl14 profile is superior in bandwidth by about 3,500mb/s per aida and 2ns per aida

heres another 4000cl14

 

[Taraquin]

You are correct. Slapped the new kit in along with my trident z neos and it wont even post at 4000 even with high voltage and the sloppiest of timings. The mem controller simply cant take it or its my mobo itself.

Anyways ill be finishing up 3800 14-15-14-21 soon on hci memtest . its looking promising and then ill run a few sotr benchmarks and post the results

dropped in the ripjaws with the neos for 32 gigs total. played around with some voltages/ cad bus and odt's and got this stable. the top run being 3800cl14 @ 32 gigs and the bottom being 4000cl14 at 16gigs. the bottom result was more of a freak run and it was just the highest score i have ever gotten but its not really consistent (usually 230-233fps) its the screenshot i had so i posted it. otherwise they are both in line with each other. the dual rank bonus from having all the dimm slots filled negating the 200mhz and 100 fclock bonus from running only 2 sticks and vise versa. the timings are pretty much the same besides a change of twrrd from 1 to 3 for stability. i know i have the screenshot somewhere but i did a previous run at 3800cl14 @ 16gb same timings and it scored 226. so running dual rank is for sure an upgrade speed for speed and timing for timing. so my conclusion is running a dual ranked ram config is equal in performance a x2 single ranked config with approx. 200mhz and 100flck with the same timings in this benchmark even though the 4000cl14 profile is superior in bandwidth by about 3,500mb/s per aida and 2ns per aida
View attachment 216246View attachment 216247View attachment 216248

heres another 4000cl14
View attachment 216252

Running all core 4.7 or curve optimizer? If your binning is good you might be able to run 4.85 single and 4.7 all core with - 30 on CO and ppt around 90-100W.

 

[Mbellantoni]

Running all core 4.7 or curve optimizer? If your binning is good you might be able to run 4.85 single and 4.7 all core with - 30 on CO and ppt around 90-100W.

Curve optimizer. 0 , - 19 , -14 , - 18, -19, -19 my chip is BAD. Like...hilariously bad at all core. Pbo is my saving grace. Lol

 

[Taraquin]

Curve optimizer. 0 , - 19 , -14 , - 18, -19, -19 my chip is BAD. Like...hilariously bad at all core. Pbo is my saving grace. Lol

My chips is golden, but ram is really bad, your seems very good, wanna trade ram with me?

 

[Felix123BU]

My chips is golden, but ram is really bad, your seems very good, wanna trade ram with me?

In the same boat buddy, great CPU, crappy RAM, oh well, we can have it all...not without spending tons of cash

 

[Mbellantoni]

My chips is golden, but ram is really bad, your seems very good, wanna trade ram with me?

Lol i think im gonna end up selling this kit i just bought and try to re cooperate my money. I can not get this kit stable and its driving me nuts. I dont think its the ram either i think my mobo is showing its limits with 4 dimms. Gonna try to sell these. Buy a set of ripjaws 3600c14 16gbx2 sticks and then sell off my neos lol

 

[Mbellantoni]

Amd's new driver enabled SAM on 5000 series. Just wanted to show. Cpu render takes a pretty large hit and yet, performance is improved. Staying true to the thread ill post the 1080 results. But there are even larger gains at 1440p. Pretty interesting. The top run is regular. Bottom has SAM enabled

 

[Taraquin]

Amd's new driver enabled SAM on 5000 series. Just wanted to show. Cpu render takes a pretty large hit and yet, performance is improved. Staying true to the thread ill post the 1080 results. But there are even larger gains at 1440p. Pretty interesting. The top run is regular. Bottom has SAM enabled
View attachment 217011View attachment 217012

Interesting. On my 5600X, enabling r-bar reduces CPU game avg by 5-10fps, seems like running AMD GPU doesn`t do that. Cool that the 5000-series gets an uplift. In retrospect, the 5700 and 5700XT gave very good value for money, especially after FSR and if you got the at close to msrp in Q3 2019. Performance seems close to 2070S, only downside for me is lack of DLSS and that`s why I sold my 5700XT and bought 3060ti. How much was the fps-improvement at 1440p?

 

[Mbellantoni]

Interesting. On my 5600X, enabling r-bar reduces CPU game avg by 5-10fps, seems like running AMD GPU doesn`t do that. Cool that the 5000-series gets an uplift. In retrospect, the 5700 and 5700XT gave very good value for money, especially after FSR and if you got the at close to msrp in Q3 2019. Performance seems close to 2070S, only downside for me is lack of DLSS and that`s why I sold my 5700XT and bought 3060ti. How much was the fps-improvement at 1440p?

Getting better fps than 2070s these days in alot of games . Amd is really stepping their game up

 

[Felix123BU]

View attachment 217013View attachment 217014

Getting better fps than 2070s these days in alot of games . Amd is really stepping their game up

Its interesting that you get a FPS increase with SAM, I get a minute 1-3 FPS reduction in this game with SAM enabled, tested a couple of time over different drivers out of curiosity.

 

[harm9963]

Can you share zenTimings settings?


iod voltage already at 1.05, I try to add more?

 

[phanbuey]

that's a crazy 5600x tune. wow.

 

[Mbellantoni]

Are those stable with gdm off at 1t? Im pretty convinced its near impossible to truly get that stable on ryzen. Especially with a clkdrstr of 24ohms @harm9963

Its interesting that you get a FPS increase with SAM, I get a minute 1-3 FPS reduction in this game with SAM enabled, tested a couple of time over different drivers out of curiosity.

Have you tried the latest revision? They also just released new chipset drivers the other day as well. May be worth a shot

 

[Felix123BU]

Are those stable with gdm off at 1t? Im pretty convinced its near impossible to truly get that stable on ryzen. Especially with a clkdrstr of 24ohms @harm9963


Have you tried the latest revision? They also just released new chipset drivers the other day as well. May be worth a shot

Will surely do when AMD releases the next WHQL Gpu driver Chipset ones all always up to date. But if I get bored enough might give it a shot with the latest graphics driver aswell

Was doing a bit of testing with my ram stick these days, the tuning that I did mainly for the purpose of this tests thread was "stable" for the last 3 months in everything, no crash, no weird behavior, nothing, until this week I got some free time and decided to replay Horizon Zero Dawn, that was crashing regularly every 10-15 minutes, never did before this week.
I basically went crazy because I was not sure what was causing it, but in the end narrowed it down to...memory instability Changed Trfc from 265 to 294 and no more crashes after 5 hours and counting. Seems 265 at the set voltage was a bit optimistic. Its the first time a tested ram oc gave me any crash

 

[Mbellantoni]

Will surely do when AMD releases the next WHQL Gpu driver Chipset ones all always up to date. But if I get bored enough might give it a shot with the latest graphics driver aswell

Was doing a bit of testing with my ram stick these days, the tuning that I did mainly for the purpose of this tests thread was "stable" for the last 3 months in everything, no crash, no weird behavior, nothing, until this week I got some free time and decided to replay Horizon Zero Dawn, that was crashing regularly every 10-15 minutes, never did before this week.
I basically went crazy because I was not sure what was causing it, but in the end narrowed it down to...memory instability Changed Trfc from 265 to 294 and no more crashes after 5 hours and counting. Seems 265 at the set voltage was a bit optimistic. Its the first time a tested ram oc gave me any crash

Trfc can be tricky. Ive had test pass 6hrs straight and some fail within 30 minutes with the same trfc. Its one of the trickier timings when it comes to stability for sure

 

[AVATARAT]

Ryzen 5 5600x+PBO+CO Per Core
2x8GB DDR4@4066MHz 16-17-14-28-2T
PowerColor RX 6700 XT 12GB @2720MHz / Mem 2150MHz(17200)

 

[jamse]

Haven't found any timings I could lower without actually losing performance. Anything I try to lower further will just make it run worse in all benchmarks for some reason

 

[Felix123BU]

View attachment 218394
View attachment 218395

Haven't found any timings I could lower without actually losing performance. Anything I try to lower further will just make it run worse in all benchmarks for some reason

That's a nice score for the hardware combo

 

[Taraquin]

View attachment 218394
View attachment 218395

Haven't found any timings I could lower without actually losing performance. Anything I try to lower further will just make it run worse in all benchmarks for some reason

Most of your timings are good. The only ones that should improve performance is setting tras to 24, trc to 42/trfc 252 or trc/trfc 44/264, trrds to 4, tfaw to 16, twr 14 or 12. Setting scls to 4 might improve stability without losing much if any performance.

You could also try curve optimizer that might give you a 3-4% boost in some games/apps.

 

[Felix123BU]

Most of your timings are good. The only ones that should improve performance is setting tras to 24, trc to 42/trfc 252 or trc/trfc 44/264, trrds to 4, tfaw to 16, twr 14 or 12. Setting scls to 4 might improve stability without losing much if any performance.

You could also try curve optimizer that might give you a 3-4% boost in some games/apps.

For me the biggest boost I got in this test and others too was an all-core OC, slightly better than the PBO+curve optimizer, and even though PBO got me to 5.05 Ghz, a 4.7-4.8 Ghz all-core had better FPS every time. The differences are small anyway, memory tuning has by far the biggest impact.

 

[Taraquin]

For me the biggest boost I got in this test and others too was an all-core OC, slightly better than the PBO+curve optimizer, and even though PBO got me to 5.05 Ghz, a 4.7-4.8 Ghz all-core had better FPS every time. The differences are small anyway, memory tuning has by far the biggest impact.

Yeah, I get 234fps at best with +200 pbo(4.6 avg allcore), but my all core 4.8 got 248fps. I have activated resize bar and that steals a few CPU fps though, 4.8 was without r-bar.

So I re-ran the test. The following is 4.8 all core with r-bar. This lost me 8 fps avg on CPU-game compared to the result without r-bar, 4000cl15 tuned.

Next up is Curve optimizer and pbo+200 and ram at 4000cl15 tuned. CO lost 6 fps avg compared to all core 4.8.

Last is the same as above but a slower 4000cl16 with slightly looser timings and almost 0.1V lower vdimm. For unknown reasons it beat the 4000cl15 with tighter timings

Wrap-up: Enabling r-bar reduces CPU-performance by 3%, allcore 4.8 is about 2% faster than CO+200 pbo, but uses about 10-15W more and CPU runs 5-10C warmer. CO might match 4.8 allcore if I diable the 76W limit.

 

[Felix123BU]

Yeah, I get 234fps at best with +200 pbo(4.6 avg allcore), but my all core 4.8 got 248fps. I have activated resize bar and that steals a few CPU fps though, 4.8 was without r-bar.

So I re-ran the test. The following is 4.8 all core with r-bar. This lost me 8 fps avg on CPU-game compared to the result without r-bar, 4000cl15 tuned.View attachment 218418
Next up is Curve optimizer and pbo+200 and ram at 4000cl15 tuned. CO lost 6 fps avg compared to all core 4.8.
View attachment 218419
Last is the same as above but a slower 4000cl16 with slightly looser timings and almost 0.1V lower vdimm. For unknown reasons it beat the 4000cl15 with tighter timings
View attachment 218420

Wrap-up: Enabling r-bar reduces CPU-performance by 3%, allcore 4.8 is about 2% faster than CO+200 pbo, but uses about 10-15W more and CPU runs 5-10C warmer. CO might match 4.8 allcore if I diable the 76W limit.

I also noticed that r-bar slightly reduces CPU performance in this game on AMD Cpu's Not sure about r-bar and Intel Cpu's.

 

[Det0x]

tRFC 228 put up a really good fight with me, but in the end i have managed to knock it back in line

Had multiple runs where testmem simply would stop running without giving any errors, with timer still running, often at cycles between 15 and 24
What fixed it in the end was higher voltages.. +20mv to VDDP, CCD and IOD fixed testmem from stop running between cycles

I'm pretty sure this will be my finalised profile for awhile now..

Lots of things going on in this screen, so i will also write what it show:

25 cycles testmem 1usmus cfg
3000% Karhu ramtest
30min OCCT Large AVX
Aida64 memory benchmark
CPU-Z cpu bench, ST 709 and MT above 14k

And a quick and easy gaming test in SotTR with daily 24/7 settings. (done in windows10)

Seems like i have to wait for Alder lake for someone else to cross the 300 fps line and i can start pushing again..
My old record with everything maxed and not stable was 314 average cpu fps i think, with these new memory stick and settings it should be good for ~320-330 fps when running bench settings
Looking forward what the 5950 xt 3dnow! edition can do in this bench when i get it

For those running windows11:

Windows 11 will hobble gaming performance by default on some prebuilt PCs​

Far Cry New Dawn is the outlier here, which barely shrugs at VBS, with just a 5% reduction in frame rate. But Horizon Zero Dawn drops by some 25%, Metro Exodus by 24%, and Shadow of the Tomb Raider by 28%. Interestingly, the 3DMark Time Spy score only dropped by 10%.

"In our testing with pre-release builds of Windows 11," UL tells us, "a feature called Virtualization-based Security (VBS) causes performance to drop. VBS is enabled by default after a clean install of Windows 11, but not when upgrading from Windows 10. This means the same system can get different benchmark scores depending on how Windows 11 was installed and whether VBS is enabled or not. "

If you are affected:

"HKEY_LOCAL_MACHINE\System\CurrentControlSet\Control\DeviceGuard and add a new DWORD value named EnableVirtualizationBasedSecurity and set its value to 0 "

or

• Run gpedit.msc
• Go to Local Computer Policy > Computer Configuration > Administrative Templates > System > Device Guard
• Double click Turn on Virtualization Based Security
• Select Disabled
• Click OK
• A reboot might be required.

 

[Felix123BU]

tRFC 228 put up a really good fight with me, but in the end i have managed to knock it back in line

Had multiple runs where testmem simply would stop running without giving any errors, with timer still running, often at cycles between 15 and 24
What fixed it in the end was higher voltages.. +20mv to VDDP, CCD and IOD fixed testmem from stop running between cycles

I'm pretty sure this will be my finalised profile for awhile now..
View attachment 219164
Lots of things going on in this screen, so i will also write what it show:

25 cycles testmem 1usmus cfg
3000% Karhu ramtest
30min OCCT Large AVX
Aida64 memory benchmark
CPU-Z cpu bench, ST 709 and MT above 14k

And a quick and easy gaming test in SotTR with daily 24/7 settings. (done in windows10)
View attachment 219165
Seems like i have to wait for Alder lake for someone else to cross the 300 fps line and i can start pushing again..
My old record with everything maxed and not stable was 314 average cpu fps i think, with these new memory stick and settings it should be good for ~320-330 fps when running bench settings
Looking forward what the 5950 xt 3dnow! edition can do in this bench when i get it

For those running windows11:

Windows 11 will hobble gaming performance by default on some prebuilt PCs​

View attachment 219170

Far Cry New Dawn is the outlier here, which barely shrugs at VBS, with just a 5% reduction in frame rate. But Horizon Zero Dawn drops by some 25%, Metro Exodus by 24%, and Shadow of the Tomb Raider by 28%. Interestingly, the 3DMark Time Spy score only dropped by 10%.

"In our testing with pre-release builds of Windows 11," UL tells us, "a feature called Virtualization-based Security (VBS) causes performance to drop. VBS is enabled by default after a clean install of Windows 11, but not when upgrading from Windows 10. This means the same system can get different benchmark scores depending on how Windows 11 was installed and whether VBS is enabled or not. "

If you are affected:

"HKEY_LOCAL_MACHINE\System\CurrentControlSet\Control\DeviceGuard and add a new DWORD value named EnableVirtualizationBasedSecurity and set its value to 0 "

or

• Run gpedit.msc
• Go to Local Computer Policy > Computer Configuration > Administrative Templates > System > Device Guard
• Double click Turn on Virtualization Based Security
• Select Disabled
• Click OK
• A reboot might be required.

Nice post, also, if you get a Alder Lake CPU, I would be highly interested to see results with it