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

[Felix123BU]

View attachment 210858

View attachment 210859

Can you please re-upload keeping in mind the threads topic? Please see first page:

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)

Thank you!

 

[CGi-Quality]

Can you please re-upload keeping in mind the threads topic? Please see first page:

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)

Thank you!

Weird, I thought it was fixed (I saw the error myself and re-uploaded before you replied). It's fine though, I'll re-up shortly.

Edit: Done. Got the old one mixed up in the upload, I'm guessing.

That said → lower result than I excepted with that proc, but hey, it is what it is.

 

[Cowboystrekk]

Weird, I thought it was fixed (I saw the error myself and re-uploaded before you replied). It's fine though, I'll re-up shortly.

Edit: Done. Got the old one mixed up in the upload, I'm guessing.

That said → lower result than I excepted with that proc, but hey, it is what it is.

Have you tweaked ram or OCed CPU? You have a lot of potential

 

[CGi-Quality]

Have you tweaked ram or OCed CPU? You have a lot of potential

RAM is running at its 2400MHz XMP, but with the proc, I figured I wouldn't need to. I OC'ed my Haswell-E (5960X) proc about 7 years ago, but haven't dabbled in OC since.

I am considering upgrading to 4000MHz RAM, though. 2400 may not cut it.

 

[Felix123BU]

RAM is running at its 2400MHz XMP, but with the proc, I figured I wouldn't need to. I OC'ed my Haswell-E (5960X) proc about 7 years ago, but haven't dabbled in OC since.

I am considering upgrading to 4000MHz RAM, though. 2400 may not cut it.

Yup high-speed low latency RAM helps a lot, low latency maybe even more than speed

 

[CGi-Quality]

Yup high-speed low latency RAM helps a lot, low latency maybe even more than speed

I also found a batch that's C16/3600. May go with it!

 

[Felix123BU]

I also found a batch that's C16/3600. May go with it!

C16/3600 is good, C14/3600 is great, but its also damn expensive compared to C16/3600, and the performance difference is not that great if you don't want to squeeze the absolute best out of your kit

 

[CGi-Quality]

C16/3600 is good, C14/3600 is great, but its also damn expensive compared to C16/3600, and the performance difference is not that great if you don't want to squeeze the absolute best out of your kit

I assume though that C16/3600 should easily smoke C14/2400 and it will show? In the past, I was once lead to believe that RAM speeds really didn't make THAT big a difference. I'd be happy to hear that this was totally false.

 

[Cowboystrekk]

Yup high-speed low latency RAM helps a lot, low latency maybe even more than speed

I bet you can get 20% better performance with better ram. Try finding a Samsung B-die kit. Ripjaws 3200cl14 is quite cheap and you should be able to OC those to 4000cl16 in most cases

 

[Ja.KooLit]

 

[damric]

Ryzen 5 3600@ 4.35GHz
4x8GB DDR4 @3800MT/s 16-19-19-39 CR1 (Hynix)
Vega 64 @1750/1100

 

[Cowboystrekk]

Ryzen 5 3600@ 4.35GHz
4x8GB DDR4 @3800MT/s 16-19-19-39 CR1 (Hynix)
Vega 64 @1750/1100

You can probably get around 170 on CPU game with a few tweaks on the ram, post a zentimings screenshot and we can help you

 

[damric]

Good luck. These are some very cheap RAM.

 

[Cowboystrekk]

Good luck. These are some very cheap RAM.

Have you tested the fast preset at 3800? Try it. If they are Hynix DJR they are some of the best chips out there

If fast preset don't work even at 1.45V try tras 35, tRC 60, tFAW 20, trrds 5, trrdl 8, tRFC 540, tWR 16, tRTP 8, scls 4. If that works try tRC 58, trfc 504, tfaw 16, trrds 4, trrdl 6, tWR 12, tRTP 6.

 

[damric]

Have you tested the fast preset at 3800? Try it. If they are Hynix DJR they are some of the best chips out there

If fast preset don't work even at 1.45V try tras 35, tRC 60, tFAW 20, trrds 5, trrdl 8, tRFC 540, tWR 16, tRTP 8, scls 4. If that works try tRC 58, trfc 504, tfaw 16, trrds 4, trrdl 6, tWR 12, tRTP 6.

I've honestly tried so many settings. I think it's more of a limitation on this memory controller, and it's a weird one. With my Zen and Zen+ CPUs I could run at tighter timings using the calculator with the same sticks, 3466/14 if I recall. This CPU can do almost a 1950 fabric clock, but it seems to hate timings tightened, which is weird to me. Especially tRFC it hates tightening. Regardless, I'll give your suggested timings a shot tomorrow when I get a chance. My wife's rig has a shitty series wired motherboard and CJRs and can do 4x16GB way easier on her 5800x than I can even do 4x8GB with my parallel topology board.

 

[Mbellantoni]

Hello all, i was browsing the internet looking for a good way to test my manual ram OC in realtime on my 5600x and came across this forum. Here are my results. My ram profile is 100 percent stable in a sense it doesnt throw any critical errors. However, a 2000flck is rough to get stable and it will throw whea 19's( interconnect bus error) which are corrected errors ONLY under extreme heavy loads like p95 large fft test. I cannot replicate them under any game, benchmark or memory based stress test besides p95. Large fft so im rolling with it. There is a very narrow margin with soc and cldo voltages that minimize these errors in p.95 to near none that took more time than i care to admit to dial in lol. Too much voltage or too little resulted in more whea 19's. I dont run prime much unless im off doing another oc so im happy with it and check for them daily and seem to have found luck with my voltages.

Anyways this is on a 5600x at 4.7ghz (pbo) with custom ppt values and negative curve. This 4000cl14 profile made a difference of 27 frames and brought my gpu productivity (gpu bound) 15% over the stock 3600cl16 gskill profile (f4-3600c16-8gtzn)

 

[Felix123BU]

Hello all, i was browsing the internet looking for a good way to test my manual ram OC in realtime on my 5600x and came across this forum. Here are my results. My ram profile is 100 percent stable in a sense it doesnt throw any critical errors. However, a 2000flck is rough to get stable and it will throw whea 19's( interconnect bus error) which are corrected errors ONLY under extreme heavy loads like p95 large fft test. I cannot replicate them under any game, benchmark or memory based stress test besides p95. Large fft so im rolling with it. There is a very narrow margin with soc and cldo voltages that minimize these errors in p.95 to near none that took more time than i care to admit to dial in lol. Too much voltage or too little resulted in more whea 19's. I dont run prime much unless im off doing another oc so im happy with it and check for them daily and seem to have found luck with my voltages.

Anyways this is on a 5600x at 4.7ghz (pbo) with custom ppt values and negative curve. This 4000cl14 profile made a difference of 27 frames and brought my gpu productivity (gpu bound) 15% over the stock 3600cl16 gskill profile (f4-3600c16-8gtzn)

Very respectable score for a 5700XT This game and this benchmark loves fast tightened ram.

 

[Cowboystrekk]

Hello all, i was browsing the internet looking for a good way to test my manual ram OC in realtime on my 5600x and came across this forum. Here are my results. My ram profile is 100 percent stable in a sense it doesnt throw any critical errors. However, a 2000flck is rough to get stable and it will throw whea 19's( interconnect bus error) which are corrected errors ONLY under extreme heavy loads like p95 large fft test. I cannot replicate them under any game, benchmark or memory based stress test besides p95. Large fft so im rolling with it. There is a very narrow margin with soc and cldo voltages that minimize these errors in p.95 to near none that took more time than i care to admit to dial in lol. Too much voltage or too little resulted in more whea 19's. I dont run prime much unless im off doing another oc so im happy with it and check for them daily and seem to have found luck with my voltages.

Anyways this is on a 5600x at 4.7ghz (pbo) with custom ppt values and negative curve. This 4000cl14 profile made a difference of 27 frames and brought my gpu productivity (gpu bound) 15% over the stock 3600cl16 gskill profile (f4-3600c16-8gtzn)

Good score, very close to mine. I ran CO -30 and +200 PBO when I got my best score. You can probably run trtp at 6 and trp at 15 (requires 2T instead of GDM), that might give you a few frames. When I compare your score against mine you run lower CL by 2, tRC by 7 and tRFC by 44, but 1 higher tRP, 2 higher WR and 3 higher tRTP. What is your aida64-scores?

 

[Splinterdog]

All hardware at default - no OC.
Ryzen 5600X
RX 5700 XT
Mem 3200GHz XMP 2.0

 

[Mbellantoni]

All hardware at default - no OC.
Ryzen 5600X
RX 5700 XT
Mem 3200GHz XMP 2.0
View attachment 212107
View attachment 212108

Thats pretty damn impressive for being stock at those ram speeds. Even my best run (233fps) in non exlusive fullscreen - my max cpu render capped at 697. And is far below yours although the other cpu scores were still a little higher. Very nice!

Good score, very close to mine. I ran CO -30 and +200 PBO when I got my best score. You can probably run trtp at 6 and trp at 15 (requires 2T instead of GDM), that might give you a few frames. When I compare your score against mine you run lower CL by 2, tRC by 7 and tRFC by 44, but 1 higher tRP, 2 higher WR and 3 higher tRTP. What is your aida64-scores?

Im using my daily Oc's here. Unfortunately i think i hit the crap end of the silicon lottery on my cpu as it needs more voltage then some of the samples ive seen out there. My first core wont take any sort of negative curve at all or it will drop threads in prime. 0 -19 -14 -20 -20 -20 at 4.7ghz, scalar 4x, ppt values are 125-75-105 as i found better boost behavior keeping the ppt values closer to stock rather than leaving them uncapped.

Gpu is 2095 boost @ 1115mv and 1810mhz on the memory side.

As far as the ram goes i used the dram calc at 3600 fast preset then changed the primarys around and lowered the trc and trfc quite a bit. As well as changing the soc and cldo voltages around to "stabilize" the memory controller. Proc odt's and all that are still the same as i found no reason to change them after stability testing. Although in my limited testing swapping these around can have a extremely minor effect on bandwidth and latency. There seems to be a sweet spot

I may see what i can extract out of it further in terms of subtimings but i feel like i may be at the bleeding edge at this point. Im running these at 1.56.5v which is right in line with 4000cl14 xmp's (1.5v) with much tighter subtimings than those you buy off the shelf from g skill and such.

Pretty sure my saving grace is my fan cooling my dram. I used a amd wraith cooler fan i had laying around and custom made a bracket for my case and mounted the fan onto it using double sided hanging tape. It sits directly over the ram and it looks pretty damn clean tbh lol. works like a dream. My dram doesnt pass 35c under heavy gaming loads or memory stress test. Its probably the only reason i can run this profile stable. Before my dram was getting up into the mid to high 40's and samsung bdie doesnt like that when your pushing it. Although i feel my ram is pretty voltage tolerant and from my testing theres only a 1-2C difference in temps under load from my xmp profile.

As far as aida my trail has expired and im too cheap right now to buy it. But judging my last run (3800cl14) with looser trc and trfc i would estimate im at about 59-60,000mb/s read and around 53.5NS of latency

 

[Splinterdog]

Thats pretty damn impressive for being stock at those ram speeds. Even my best run (233fps) in non exlusive fullscreen - my max cpu render capped at 697. And is far below yours although the other cpu scores were still a little higher. Very nice!

Im using my daily Oc's here. Unfortunately i think i hit the crap end of the silicon lottery on my cpu as it needs more voltage then some of the samples ive seen out there. My first core wont take any sort of negative curve at all or it will drop threads in prime. 0 -19 -14 -20 -20 -20 at 4.7ghz, scalar 4x, ppt values are 125-75-105 as i found better boost behavior keeping the ppt values closer to stock rather than leaving them uncapped.

Gpu is 2095 boost @ 1115mv and 1810mhz on the memory side.

As far as the ram goes i used the dram calc at 3600 fast preset then changed the primarys around and lowered the trc and trfc quite a bit. As well as changing the soc and cldo voltages around to "stabilize" the memory controller. Proc odt's and all that are still the same as i found no reason to change them after stability testing. Although in my limited testing swapping these around can have a extremely minor effect on bandwidth and latency. There seems to be a sweet spot

I may see what i can extract out of it further in terms of subtimings but i feel like i may be at the bleeding edge at this point. Im running these at 1.56.5v which is right in line with 4000cl14 xmp's (1.5v) with much tighter subtimings than those you buy off the shelf from g skill and such.

Pretty sure my saving grace is my fan cooling my dram. I used a amd wraith cooler fan i had laying around and custom made a bracket for my case and mounted the fan onto it using double sided hanging tape. It sits directly over the ram and it looks pretty damn clean tbh lol. works like a dream. My dram doesnt pass 35c under heavy gaming loads or memory stress test. Its probably the only reason i can run this profile stable. Before my dram was getting up into the mid to high 40's and samsung bdie doesnt like that when your pushing it. Although i feel my ram is pretty voltage tolerant and from my testing theres only a 1-2C difference in temps under load from my xmp profile.

As far as aida my trail has expired and im too cheap right now to buy it. But judging my last run (3800cl14) with looser trc and trfc i would estimate im at about 59-60,000mb/s read and around 53.5NS of latency

Looks like we have the same hardware, but I have a tad more memory. Maybe that's the difference?
I haven't yet put my toes in the overclocking game in a serious way. Much to learn.

 

[Cheese_On_tsaot]

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.

 

[Mbellantoni]

Looks like we have the same hardware, but I have a tad more memory. Maybe that's the difference?
I haven't yet put my toes in the overclocking game in a serious way. Much to learn.

I honestly couldnt tell you lol. Ive ran it multiple times to confirm my scores were consistant. Maybe you had a freak run? I would check again 2 or 3 more times to make sure its consistant.

Otherwise i wouldnt expect much of a difference between our gpu's especially being as cpu bound as we are in this test. My gpu is oced and undervolted with temps in mind. Performance gains are probably marginal as far as gpu' s go 1-2 fps over stock but thermals are low (57-62C depending on the game)

If i recall correctly a gamers nexus video said something about running dual rank dimms (16x2) or 4 dimms total results in a performance increase on 5000 series architecture. If this is the case then im pretty blown away on how much of a difference it actually makes.

As far as the cpu overclock goes ive heard ryzen5000 is pretty well optimized for gaming out of the box. Multicore leaves some to desired but single core doesnt leave much for improvement even with pbo on. Ive tried different speeds on this particular benchmark such as 4725-4850 with pbo and the improvements were marginal at best and the voltages were outrageous even at 4725 compared to 4700 there was a difference of 3volts just at the title screen alone.

This is all theory of course. Who knows maybe you just have a bad ass cpu + mobo + ram that works together really well.

 

[Cheese_On_tsaot]

I honestly couldnt tell you lol. Ive ran it multiple times to confirm my scores were consistant. Maybe you had a freak run? I would check again 2 or 3 more times to make sure its consistant.

Otherwise i wouldnt expect much of a difference between our gpu's especially being as cpu bound as we are in this test. My gpu is oced and undervolted with temps in mind. Performance gains are probably marginal as far as gpu' s go 1-2 fps over stock but thermals are low (57-62C depending on the game)

If i recall correctly a gamers nexus video said something about running dual rank dimms (16x2) or 4 dimms total results in a performance increase on 5000 series architecture. If this is the case then im pretty blown away on how much of a difference it actually makes.

As far as the cpu overclock goes ive heard ryzen5000 is pretty well optimized for gaming out of the box. Multicore leaves some to desired but single core doesnt leave much for improvement even with pbo on. Ive tried different speeds on this particular benchmark such as 4725-4850 with pbo and the improvements were marginal at best and the voltages were outrageous even at 4725 compared to 4700 there was a difference of 3volts just at the idle screen alone.

This is all theory of course. Who knows maybe you just have a bad ass cpu + mobo + ram that works together really well.

I have single rank dimms and they can't OC at all regardless of being B die...

 

[Mbellantoni]

I have single rank dimms and they can't OC at all regardless of being B die...

Got a zentimings screenshot?