AMD Ryzen 7 5800X 3D Cache Eight Core 4.5GHz (Socket AM4) Processor - Retail - Go Go Go xD

[true_gamer]

1440p is where you see the biggest difference, especially the min CPU FPS. 400pts difference between each Clock/Timings. 3733 C14 is the fastest here.

Not much difference between 3200 C14 and 3600 C16 in the other tests, but there is an uplift with 3733 C14, especially the CPU mins.

1080p is a strange one, especially with the 4090.
Max CPU FPS is margain of error at this res, as Windows could start a task at any given moment during the benchmark, and results can differ due to the main load on the CPU, so any extra process would have a negative impact on the results.

What would have been more accurate, but very time consuming, is running 3 test of each, and take the mean of each one. That would be 27 tests to run...

 

[Troezar]

I run my RAM at XMP 3600 C14 1T - seems like there is no point messing with it.

 

[Guest2]

1440p is where you see the biggest difference, especially the min CPU FPS. 400pts difference between each Clock/Timings. 3733 C14 is the fastest here.

Not much difference between 3200 C14 and 3600 C16 in the other tests, but there is an uplift with 3733 C14, especially the CPU mins.

1080p is a strange one, especially with the 4090.
Max CPU FPS is margain of error at this res, as Windows could start a task at any given moment during the benchmark, and results can differ due to the main load on the CPU, so any extra process would have a negative impact on the results.

What would have been more accurate, but very time consuming, is running 3 test of each, and take the mean of each one. That would be 27 tests to run...

Interesting. Also yes, very time consuming, don't do that!

I know my RAM is good for 3800Mhz, C14 as had it running this on my 5800X.
The 5800X3D doesnt like 1900 FCLK though - WHEA errors, sound issues and black screens in gaming & benching.

Would you recommend me trying the same C14 settings, but 3733Mhz 1867 FCLK with the 5800X3D? I game at 3440 x 1440 (so somewhere between 1440p and 4k)

Currently my RAM is at XMP 3600, 1800 FCLK, C16. I either leave it at XMP, run it at C14 or up the FCLK also and run C14. Idea is to raise 1% lows but might not be worth it if results are only margin of error

 

[true_gamer]

I run my RAM at XMP 3600 C14 1T - seems like there is no point messing with it.

I would say, you would be looking at performance between the 3600 C16 and 3733 C14 of the 1440p tests.
I say this as the workload is more shared by both GPU and CPU, meaning that if a Windows process decided to run during a benchmark, the CPU is not maxed out, and the results would not be so negatively impacted compared to 1080p where the CPU is hitting full load on many cores.

Interesting. Also yes, very time consuming, don't do that!

I know my RAM is good for 3800Mhz, C14 as had it running this on my 5800X.
The 5800X3D doesnt like 1900 FCLK though - WHEA errors, sound issues and black screens in gaming & benching.

Would you recommend me trying the same C14 settings, but 3733Mhz 1867 FCLK with the 5800X3D? I game at 3440 x 1440 (so somewhere between 1440p and 4k)

Currently my RAM is at XMP 3600, 1800 FCLK, C16. I either leave it at XMP, run it at C14 or up the FCLK also and run C14. Idea is to raise 1% lows but might not be worth it if results are only margin of error

There was no real difference when I tested 3800 C14 vs 3733 C14 as I had to loosen the sub timings for 3800 C14.
At 4000 C16, there was a small bump in performance, although there was no WHEA errors, I had little niggles that showed there was more to the stability than WHEA errors with a 2000MHz IF.
So I still think 3733 C14 is the sweetspot with 100% stability. (I've been running these timings for the past 2yrs on my 5800X and now the 5800X3D).

You can give the timings a go, testing both 3600 C16 and 3733 C14. I do have some good quailty B-Die, so run some stability tests like OCCT Mem test etc.

Most importantly, as with my Motherboard, when the latest Bios came out, I have to manually set my SOC, CLDO, VDDP, VDDG IOG/CCD as my system would error over 3200MHz on stock Auto voltages. (Common problem).

 

[Guest2]

There was no real difference when I tested 3800 C14 vs 3733 C14 as I had to loosen the sub timings for 3800 C14.
At 4000 C16, there was a small bump in performance, although there was no WHEA errors, I had little niggles that showed there was more to the stability than WHEA errors with a 2000MHz IF.
So I still think 3733 C14 is the sweetspot with 100% stability. (I've been running these timings for the past 2yrs on my 5800X and now the 5800X3D).

You can give the timings a go, testing both 3600 C16 and 3733 C14. I do have some good quailty B-Die, so run some stability tests like OCCT Mem test etc.

Most importantly, as with my Motherboard, when the latest Bios came out, I have to manually set my SOC, CLDO, VDDP, VDDG IOG/CCD as my system would error over 3200MHz on stock Auto voltages. (Common problem).

I'm running the 8 pack dark pro 16-16-16-38, this - https://www.overclockers.co.uk/team...00c16-3600mhz-dual-channel-kit-my-002-8p.html

On the 1.2.0.3b bios it was stable like this for 2 years, passing Karhu, TestMem5 extreme anta and RunMemTest-Pro-v2.5



I found leaving SOC, CLDO, VDDP, VDDG IOG/CC on auto worked the best and was most stable. Just with VDIMM set to 1.49V

I recently upgraded the BIOS to AGESA ComboAm4v2PI 1.2.0.7, mainly for 5800X3D so it could be the BIOS at fault rather than the CPU not liking 1900 FCLK. Either way, the RAM and board (MSI Meg Unify X570) are fine but the new BIOS could be setting auto voltages differently or other settings which affect FLCK... or it could just be the CPU which I suspect it is

 

[true_gamer]

I'm running the 8 pack dark pro 16-16-16-38, this - https://www.overclockers.co.uk/team...00c16-3600mhz-dual-channel-kit-my-002-8p.html

On the 1.2.0.3b bios it was stable like this for 2 years, passing Karhu, TestMem5 extreme anta and RunMemTest-Pro-v2.5



I found leaving SOC, CLDO, VDDP, VDDG IOG/CC on auto worked the best and was most stable. Just with VDIMM set to 1.49V

I recently upgraded the BIOS to AGESA ComboAm4v2PI 1.2.0.7, mainly for 5800X3D so it could be the BIOS at fault rather than the CPU not liking 1900 FCLK. Either way, the RAM and board (MSI Meg Unify X570) are fine but the new BIOS could be setting auto voltages differently or other settings which affect FLCK... or it could just be the CPU which I suspect it is

I would manually input those voltages, and reduce them, same with the VDIMMs, 1.45V is the max I would go for a 24/7 overclock. More voltage equals more heat, equals less stability. So unless you have active cooling on your DIMMS, I would look to refine these settings, as once they become heat soaked, stability could worsen.

My 5800X had a weird issue, that it didn't like 1900FLCK, but was OK with 2000FLCK. My 5800X3D has no issues upto 2000FLCK.

 

[Guest2]

I would manually input those voltages, and reduce them, same with the VDIMMs, 1.45V is the max I would go for a 24/7 overclock. More voltage equals more heat, equals less stability. So unless you have active cooling on your DIMMS, I would look to refine these settings, as once they become heat soaked, stability could worsen.

My 5800X had a weird issue, that it didn't like 1900FLCK, but was OK with 2000FLCK. My 5800X3D has no issues upto 2000FLCK.

When I tried 1.45V it was not stable. It was a fine line between unstable at 1.48 and unstable at 1.50!
Similar with lowered manually entered voltages on SOC and other but auto was fine.

Interesting, I might see if it boots at 2000FCLK / 4000Mhz DDR4. My 5800X didnt even POST at 2000 FCLK

Edit - tried 2000fclk, 4000ddr with xmp auto settings. Memory overclock failure on boot.
Went back into bios and set auto on ram.
2000 fclk i could get into windows But noticed memory was at 2400 so ram settings had not taken.
Rebooted but then wouldnt boot and needed cmos clear.
My guess is 2000 flck with memory at 4000mhz loosened to maybe c18 may work but is it worth it…probably not

 

[true_gamer]

When I tried 1.45V it was not stable. It was a fine line between unstable at 1.48 and unstable at 1.50!
Similar with lowered manually entered voltages on SOC and other but auto was fine.

Interesting, I might see if it boots at 2000FCLK / 4000Mhz DDR4. My 5800X didnt even POST at 2000 FCLK

Edit - tried 2000fclk, 4000ddr with xmp auto settings. Memory overclock failure on boot.
Went back into bios and set auto on ram.
2000 fclk i could get into windows But noticed memory was at 2400 so ram settings had not taken.
Rebooted but then wouldnt boot and needed cmos clear.
My guess is 2000 flck with memory at 4000mhz loosened to maybe c18 may work but is it worth it…probably not

I understand, you would need to loosen timings, and enable GDM for 1900FLCK to run at 1.45v
Same thing for 2000FLCK.
You will also need to manaully set the SOC, CLDO, VDDP, VDDG IOG/CC voltages manually, then try 1900FLCK and so on?
If no luck then you are limited by the memory controller.

Another options, is to copy like for like my 3733 timings and voltages and see if your system posts and is stable just to make sure it isn't your DIMMs quality that's holding you back.

 

[Guest2]

I understand, you would need to loosen timings, and enable GDM for 1900FLCK to run at 1.45v
Same thing for 2000FLCK.
You will also need to manaully set the SOC, CLDO, VDDP, VDDG IOG/CC voltages manually, then try 1900FLCK and so on?
If no luck then you are limited by the memory controller.

Another options, is to copy like for like my 3733 timings and voltages and see if your system posts and is stable just to make sure it isn't your DIMMs quality that's holding you back.

Think ive tried 14,13,13,13,28, 42 in the past and it wasnt stable so copying your 3733 timings and voltages probably wouldnt work

Tried my old settings at 3733Mhz RAM, 1867 FCLK and so far all good. I tended to find if it passed the 30 min mark in TestMem 5 with the extreme config it was good. Will see how I get on with this
Also running Kombo Strike mode 3, so lower CPU temps and boost all core to 4450

 

[Guest2]

Not sure if this is of any help @michty_me
In short at 3440 x 1440 it does push the minimum FPS slightly (by 7fps) but there's only a few frames in it otherwise.

Other games may vary


DDR3733, 1867 IF, C14 (above settings)


DDR3600, 1800IF, C16 (XMP)


Display settings (same for both)

 

[true_gamer]

Pretty much confirmed what I concluded. 3733 C14 is the sweetspot.

 

[michty_me]

Not sure if this is of any help @michty_me
In short at 3440 x 1440 it does push the minimum FPS slightly (by 7fps) but there's only a few frames in it otherwise.

Other games may vary


DDR3733, 1867 IF, C14 (above settings)


DDR3600, 1800IF, C16 (XMP)


Display settings (same for both)

Thanks! Looks good! Will you keep it at 3733?

 

[Guest2]

Thanks! Looks good! Will you keep it at 3733?

Undecided.
XMP felt really good in Warzone 2.

I rebooted and loaded 3733 which I had saved, had another game and got some weird stuttering. 99% sure it was just the server I was in.

Would be good to test some other game benchmarks to see if the minimums are bumpted similar to SOTTR.

Anyone have ideas on good benchmarks that show CPU lows?

Edit - I’ll try Assetto Corsa and Cyberpunk

AC was inconclusive, must be a rubbish benchmark!
3733 test 1 - min 24
3733 test 2 - min 59
3733 test 3 - min 47

3600 test 1 - min 21
3600 test 2 - min 26


Cyberpunk - Yep, they are the right way around. Lower minimum on 3733 C14. Exactly the same settings
Must be mainly GPU when it holds a set lock or dips / raises

3733 C14


3600 C16

 

[Paul_cz]

So I was watching this Hardware Unboxed test:
https://www.youtube.com/watch?v=2HqE03SpdOs

And at 5:40 of the video, I saw some impressive 5800X3D number there in Forza Horizon 5. I assume they use integrated benchmark.
In 1080p at high settings, they get 271 fps average.

So I have 5800X3D as well, albeit with different GPU (3080Ti, they have Radeon 6950).

And I got, on same settings, 189fps. Almost hundred less!

I tried 720 with DLSS on ultra performance (so real res like 240p - to completely eliminate GPU bottleneck) and got 197 fps - still 74fps less than Hardware Unboxed.

Even if they are testing some older version of Forza or whatever, I find this really weird.

Can someone with 3D and Forza please test this, and let me know if my framerate (in both 1080p and that 720p with DLSS) is within the norm? And if it is, can someone explain how can HBU can have so much higher framerate?

Thanks

 

[Troezar]

So I was watching this Hardware Unboxed test:
https://www.youtube.com/watch?v=2HqE03SpdOs

And at 5:40 of the video, I saw some impressive 5800X3D number there in Forza Horizon 5. I assume they use integrated benchmark.
In 1080p at high settings, they get 271 fps average.

So I have 5800X3D as well, albeit with different GPU (3080Ti, they have Radeon 6950).

And I got, on same settings, 189fps. Almost hundred less!

I tried 720 with DLSS on ultra performance (so real res like 240p - to completely eliminate GPU bottleneck) and got 197 fps - still 74fps less than Hardware Unboxed.

Even if they are testing some older version of Forza or whatever, I find this really weird.

Can someone with 3D and Forza please test this, and let me know if my framerate (in both 1080p and that 720p with DLSS) is within the norm? And if it is, can someone explain how can HBU can have so much higher framerate?

Thanks

All AMD platform and the 6950 being up with a 3090Ti making the difference?

 

[Guest2]

All AMD platform and the 6950 being up with a 3090Ti making the difference?

It it posssible to get the benchmark only (free) for Forza horizon 5? Then i could try it with my 3090 ti

 

[Paul_cz]

All AMD platform and the 6950 being up with a 3090Ti making the difference?

90fps difference? That would be rather insane wouldn't it? And as I said - I tested even in 720p with DLSS to completely eliminate GPU as a possible bottleneck. Still could not crack 200 fps, let alone 271. Can you test it? Just to confirm my result is normal.

It it posssible to get the benchmark only (free) for Forza horizon 5? Then i could try it with my 3090 ti

It is not available free, but you can get Forza 5 via gamepass for one dollar.

 

[Guest2]

It is not available free, but you can get Forza 5 via gamepass for one dollar.

Do you have it on Xbox pass or something like Steam? I don't have any of the xbox stuff enabled on my machine as mainly play Modern Warfare 2 / Warzone 2 and Dirt Rally 2.0

 

[Paul_cz]

Do you have it on Xbox pass or something like Steam? I don't have any of the xbox stuff enabled on my machine as mainly play Modern Warfare 2 / Warzone 2 and Dirt Rally 2.0

Yeah gamepass requires the xbox app that can be downloaded from MS Store.

 

[Guest2]

What do you score in F1 2021?

Hardware unboxed, 6950XT, F1 2021, 1080p, High TAA. 316 low, 481 average
Me, 3090 Ti, F1 2021, 1080p, High TAA. 403 low, 478 average. So a lot better