[SOLVED] 5600x Outperforming 5900X

[TM1172]

Good morning all,

I just installed a Ryzen 9 5900X in my system and ran TimeSpy. As you can probably expect, I forgot to update my BIOS with XMP profiles and all that, so my performance was lackluster, with my previous stock settings 5600X beating the 5900X by 8.9% in overall performance.

So I went back into the BIOS and enabled XMP profiles and also enabled above Gen 4 encoding and Resizable BAR support. Since previously, my 3600mhz RAM was limited to 2400mhz, I figured I'd see a pretty good performance bump.

Now the 5600X only beats the 5900X by 0.6%. What is going on? My GPU is achieving average results with the preset Wattman/Adrenaline undervolt, and I have PBO enabled through Ryzen Master. By any sane standard, this chip should outperform the 5600X out of the box. And the 5600X at stock settings wasn't beating any records, either.

Am I missing some huge performance booster or technique? Is my PSU starving the system somehow? All advice welcomed.

Specs:
Ryzen 9 5900X
AMD Radeon RX 6900 XT (Reference)
Gigabyte Aorus X570 Elite Wifi
NZXT Kraken X72 360mm AIO (CPU)
16GB T-Force Extreem 3600mhz CL14
Corsair RM750x PSU
Crucial P1 1TB NVMe SSD (Boot)
SD Black SN720 1TB NVMe SSD (Storage)
3x 1TB SATA SSD's (Storage)
Windows 10 64bit Home
AMD Adrenaline v21.3.1

 

[jojesa]

For gaming both the Ryzen 5 5600X and the Ryzen 9 5900X perform very similar, since most games won't take advantage of the extra cores and threads (or larger cache) on the 5900X.

The Ryzen 9 5900X is a workstation CPU and it will perform better than the Ryzen 5 5600X on multicore dependent tasks such as content creation.

 

[logainofhades]

Check your clock speeds when running. I would bet the 5600x is boosting higher, as it isn't getting as hot.

 

[TM1172]

For gaming both the Ryzen 5 5600X and the Ryzen 9 5900X perform very similar, since most games won't take advantage of the extra cores and threads (or larger cache) on the 5900X.

The Ryzen 9 5900X is a workstation CPU and it will perform better than the Ryzen 5 5600X on multicore dependent tasks such as content creation.

Understood about the workstation vs. gaming, thank you - just wondering why my results are coming in at approximately 50% of the performance of an average 5900X.

Check your clock speeds when running. I would bet the 5600x is boosting higher, as it isn't getting as hot.

Checking the clock speeds from the results tab, the 5900X is boosting about 250mhz higher than the 5600X. Results shot. Also, the temps on the 5900X are never getting above 75c.

 

[madmatt30]

Time spy is primarily a gpu benchmark, it uses the cpu but not to a massive extent.

Either run the full 3dmark benchmark or something multicore orientated like cinebench and you'll see the 5900x absolutely eclipse the 5600x.

The 5900x never has and never will be a 'gaming' cpu, nothing currently out there is going to use much more than 50% of its actually capacity

 

[TM1172]

Time spy is primarily a gpu benchmark, it uses the cpu but not to a massive extent.

Either run the full 3dmark benchmark or something multicore orientated like cinebench and you'll see the 5900x absolutely eclipse the 5600x.

The 5900x never has and never will be a 'gaming' cpu, nothing currently out there is going to use much more than 50% of its actually capacity

Thanks for the response - understanding that the 5900X is not intended to be a gaming CPU, I'm interested in why my CPU's score is bottom 5th percentile in the test, all other things held constant? Other users with the same CPU-GPU config are achieving much higher performance in this same test.

 

[hotaru.hino]

Thanks for the response - understanding that the 5900X is not intended to be a gaming CPU, I'm interested in why my CPU's score is bottom 5th percentile in the test, all other things held constant? Other users with the same CPU-GPU config are achieving much higher performance in this same test.

Don't use 3DMark leaderboard scores as a measure of how good your system runs. Any user-submitted scoreboard is littered with people who tweak and overclock their systems, shifting the baseline. Instead, look at what professional review sites have gotten as they will primarily report stock configuration scores which is more useful as a baseline.

 

[BogdanH]

First, I agree with @madmatt30 -for measuring CPU (single/multicore) performance, Cinebench is quite reliable and easy to use tool.
When someone says 5600X is better for gaming than 5900X, then that doesn't mean 5900X is slower in gaming. It just means, if for gaming mostly, then the price difference (for 5900X) isn't worth paying because of diminishing return (if it's used for gaming only). However fact is, 5900X is faster than 5600X in all scenarios.

About benchmarking at home... Usually we will never get as high scores as we can see by watching reviews. Reviewers mostly benchmark at controlled (and thus repeatable) situations. For example: clean PC (Windows and drivers only), no gimmics in taskbar, optimal cooling, no unnecessary peripherals attached, etc.
But at home.. if we benchmark today, we will probably get quite different scores as we did two days ago: Windows doing it's thing, some "gaming suite" is trying to check if there's an update, NAS drive is (not) in standby mode, etc. Yes, even small tools like HWiNFO eat few scores.
From all above said.. it's enough to make sure we're "close enough" to scores that we can see in reviews -in this case CPU works normally.

 

[madmatt30]

Thanks for the response - understanding that the 5900X is not intended to be a gaming CPU, I'm interested in why my CPU's score is bottom 5th percentile in the test, all other things held constant? Other users with the same CPU-GPU config are achieving much higher performance in this same test.

Yeah there's something wrong there.
Boost clock looks fine, cpu score isn't, it's way down and should be around 13000.

It looks like your 5900x is acting pretty much as a 5600x, those extra cores and threads don't appear to be being used.

 

[TM1172]

Don't use 3DMark leaderboard scores as a measure of how good your system runs. Any user-submitted scoreboard is littered with people who tweak and overclock their systems, shifting the baseline. Instead, look at what professional review sites have gotten as they will primarily report stock configuration scores which is more useful as a baseline.

Thanks. It looks like it is indeed performing within expectations based off a couple professional reviews.

 

[TM1172]

Yeah there's something wrong there.
Boost clock looks fine, cpu score isn't, it's way down and should be around 13000.

It looks like your 5900x is acting pretty much as a 5600x, those extra cores and threads don't appear to be being used.

I’ll dig a bit over at AMD too, but that’s troubling. Wonder if it’s a windows thing or there’s some BIOS setting that I need to update.

 

[Bassman999]

Here are the results I got with my 5900x, if I did the link right
https://www.3dmark.com/results

I scored 15 302 in Time Spy

AMD Ryzen 9 5900X, AMD Radeon RX 6800 x 1, 32768 MB, 64-bit Windows 10}

www.3dmark.com

I didnt update my bios or overclock, but XMP enabled on ram and I think that's it

This and User benchmark mare the only ones I used since they are free

 

[TM1172]

Here are the results I got with my 5900x, if I did the link right
https://www.3dmark.com/results

I scored 15 302 in Time Spy

AMD Ryzen 9 5900X, AMD Radeon RX 6800 x 1, 32768 MB, 64-bit Windows 10}

www.3dmark.com

I didnt update my bios or overclock, but XMP enabled on ram and I think that's it

This and User benchmark mare the only ones I used since they are free

Thanks man. Yeah, I should be seeing CPU scores similar to yours. MadMatt gave me a brainwave that I’m going to check out after lunch - it may be a simple fix, and if so, I’ll report back.

 

[Juular]

Are you benchmarking both CPUs on the same motherboard with the same UEFI settings ? If so there's no other explaination other than 3DMark Time Spy really doesn't benefit from more cores or higher boost clocks of 5900X. As PSU doesn't starve the system in a sense that it doesn't have enough power or smth, there's no such thing. Maybe what's going is that 5900X runs a bit hotter and therefore barely boost higher than 5600X but with the 360mm AIO i find that unlikely. Can you give some screenshots of AMD Ryzen Master for both CPUs while running benchmark in the background ? As with, hidden performance booster, look for Curve Optimizer setting in the BIOS under AMD Overclocking options, it's a relatively thing and can make some difference here.

 

[TM1172]

Yeah there's something wrong there.
Boost clock looks fine, cpu score isn't, it's way down and should be around 13000.

It looks like your 5900x is acting pretty much as a 5600x, those extra cores and threads don't appear to be being used.

Ok so it turns out, the CPU works a lot better when you don’t artificially limit the number of cores by 50% by, say, using “game mode” in Ryzen Master. The reason it was acting like a 5600X is because it was accidentally emulating a 5600X. With Creator mode active, it leverages all 12 cores and I scored about 12,000 on the CPU test in TimeSpy. Still room for improvement but much much closer to expectations. Thank you for the response!

 

[Bassman999]

Thanks man. Yeah, I should be seeing CPU scores similar to yours. MadMatt gave me a brainwave that I’m going to check out after lunch - it may be a simple fix, and if so, I’ll report back.

You should score higher than me with your GPU and possibly more from Bios update

 

[TM1172]

You should score higher than me with your GPU

Yeah, turns out Ryzen Master was handicapping my performance - it was in “Game Mode” and limited the CPU to only 6 of 12 cores. With all 12 cores active, I got a 17,378 overall with 18,888 GPU.

 

[Bassman999]

Yeah, turns out Ryzen Master was handicapping my performance - it was in “Game Mode” and limited the CPU to only 6 of 12 cores. With all 12 cores active, I got a 17,378 overall with 18,888 GPU.

Nice! Now Im jealous lol
I have a 3080 coming tomorrow and Ill see what it does for me.
Having issues with the 6800 and PowerColor after 2 weeks and 3 emails doesnt reply to my RMA at all.

Userbenchmark says below average and it glitches and freezes the PC sometimes and crashes on in game benchmark before it loads.

I hope your RX 6900XT isnt a PowerColor model.
Plus they say I cant adjust he fan speed faster in the tuning section unless I want to forfeit my warranty.

 

[madmatt30]

Yeah, turns out Ryzen Master was handicapping my performance - it was in “Game Mode” and limited the CPU to only 6 of 12 cores. With all 12 cores active, I got a 17,378 overall with 18,888 GPU.

Thats good to know, I had no idea that ryzen master automatically did that in game mode!

 

[Bassman999]

Thats good to know, I had no idea that ryzen master automatically did that in game mode!

Neither did I

 

[jtk2515]

Thanks for the response - understanding that the 5900X is not intended to be a gaming CPU, I'm interested in why my CPU's score is bottom 5th percentile in the test, all other things held constant? Other users with the same CPU-GPU config are achieving much higher performance in this same test.

if your Chasing synthetic scores going from 16gb to 32gb Quad rank helped my CPU score by 3-7%. If you are comparing yourself make sure to remember that.

 

[alexbirdie]

if your Chasing synthetic scores going from 16gb to 32gb Quad rank helped my CPU score by 3-7%. If you are comparing yourself make sure to remember that.

Sorry to say, but memory is either dual or single rank. you wrote "quad rank". That does not exist.
If you mean quad-channel, this only is possible on threadripper-platforms, not on AM4-Ryzen-platform ( or possible on some XEON-intel-platforms,too). The new threadripper-pro even supports 8 channels.

 

[Juular]

If you mean quad-channel, this only is possible on threadripper-platforms

And yet, Zen 3 CPUs benefit from quad-stick kits, not exactly sure why, there's a video on that from GN.

 

[Nemesia]

4x8GB single rank = 2x16GB dual rank. The performance gain is when you use 2x8GB single rank vs 4x8GB single rank.

GN recommended 2x16GB dual rank and it's easier to get the timings etc when you're using 2 sticks against 4 sticks.

 

[Bassman999]

And yet, Zen 3 CPUs benefit from quad-stick kits, not exactly sure why, there's a video on that from GN.

So my Mini ITX B550I with? 2 slots is handicapper from he start?