Question Below average performance with RTX 3080 and AMD Ryzen 7 5800X

[adsfaasfa]

Hey there, I've noticed that my overall FPS in Cyberpunk is below average compared to other folks with similar setups.
GPU: EVGA GeForce RTX 3080 FTW3 ULTRA GAMING, 10G-P5-3897-KL, 10GB
CPU: AMD Ryzen 7 5800X-
Motherboard: Asus TUF Gaming X570-Plus (Wi-Fi)
RAM: G.SKILL Trident Z Neo Series 32GB (2 x 16GB) 288-Pin PC RAM DDR4 3600 (PC4 28800) Intel XMP 2.0 Desktop Memory Model F4-3600C14D-32GTZN
PSU: 850W 80 Plus Gold

This is my benchmark userscore https://www.userbenchmark.com/UserRun/55679518

I play on a ultra wide monitor (Alienware AW3420DW 34" UW-QHD (3440x1440) 120Hz Curved IPS) and I get on around 40-60 FPS with mixed Ultra settings (I have volumetric fog resolution set to low and volumetric clouds quality to off). I look at youtube videos of other people with similar setups running the game way above 60fps on Ultra + Ray Tracing. So what gives? I overclocked my CPU with precision boost overdrive and have DOCP enabled. I can barely hit 60 in areas where there isn't much to load but as soon as I go deeper into the city, my frames drop to 45-50. This is with DLSS set to Balance too!

 

[drivinfast247]

What are GPU temps under load? Did you use DDU to remove old drivers before installing new version?

GPU installed in top PCIE slot?

RAM installed in slots a2-b2?

Latest BIOS and chipset driver?

 

[adsfaasfa]

What are GPU temps under load? Did you use DDU to remove old drivers before installing new version?

GPU installed in top PCIE slot?

RAM installed in slots a2-b2?

Latest BIOS and chipset driver?

BIOS has the latest stable release/update.
Both my GPU and CPU hover around 55C-60C
My GPU is indeed in the top PCIE slot.
My ram is installed on the 2nd and 4th slot as they should be.
Ive looked up solutions and guides on this matter, tweaked with Nvidia's control panel settings, tried it with gsync off and a restart. The only two things I havent done is a fresh Windows 11 install (which Im trying not to do unless its my only option) or a clean graphic driver install by using the DDU method.

 

[logainofhades]

Your SSD is nearly full, which is doing you no favors. Also double check you have XMP enabled.

 

[logainofhades]

Their board's VRM's are fine.

View: https://www.youtube.com/watch?v=_7PkZwY9PWM&t=584s

 

[Karadjgne]

Are you comparing the fps of other users who are using 1440p (2560x1440) QHD to your results at 3440x1440 UWQHD? Because that's a sizable hit to fps with the added pixel count.

 

[5900x]

I don't know but from the things I have seen, reaching 60fps in Cyberpunk with your hardware and all the bling at 4K is an achievement. Anything above 60 is just the cherry on the cake. Your hardware is fine as far as I'm concerned. Personally I would play at 2K with 100fps than 4K at 60.

 

[DavidM012]

It's your mobo: - Ok it's fine but not Perfect. There is a List of Mobo's in the Range on the G-Skill qvl,

The review

Page 1 of the review:

The VRMs are covered by decent-size heatsinks which did the job at stock, Precision Boost Overdrive (PBO) and while overclocking. Asus states a 12+2 VRM setup, but the board has a ‘true’ rating of 4x3 + 2, since the controller, the ASP 1106GGQW is a 6-phase unit in 4+2 mode. Power is fed to the VRMs through a required 8-pin EPS connector, plus an optional 4-pin which will provide more than enough power for the Ryzen 3 CPU lineup. I’m not sure how overclocking would go on the beefier chips, especially the 3950X, but the TUF board held up just fine with a 3700X.

Page 4 of the review:

Overclocking on this board yielded 4.224 GHz using 1.343V on the core. While the clock speed is a bit higher than what was achieved on the X570 Aorus Elite, this was due to spread spectrum and the floating BCLK as we used the same 42.25 multiplier. Anything above this voltage left us with temperatures above 90C, too hot for comfort. Regardless, anything above this multiplier at this voltage caused an error in stress testing.

It's the vroom vrms

G.SKILL F4 DDR4 3600 C14 2x16GB
901 User benchmarks, average bench 118%
2 of 4 slots used
32GB DIMM DDR4 clocked @ 3600 MHz

How could he get 3600mhz without XMP enabled?

Ryzen 5000 ram guide

And he has the fastest cas 14 memory too.

Should be at the 1:1:1 fclock ratio with 3600mhz and he has the exact same gskill kit for it?

Model F4-3600C14D-32GTZN gskill /what userbenchmark says it is

What the actual A ?

On gskill product page there is also F4-3600C14D-32GTZNA but I don't see 32GTZN on it any more. Dunno what it means though.

So if everything's supposed to be optimal, yet the memory is severely underperforming according to userbenchmark it sounds like as soon as it hits that 3600mhz optimum clock with infinity fabric, it's actually... not?

Drives me barmy that the author of the 5000 series ram guide listed the exact part number of the c15 4000 memory but not the exact part number of the c14 3600 memory. There's an N and an NA. dunno what it does. Later model? Have to ask gskill. If it's the exact same thing with the exact same timings why's it got an A?

When you say similar - are they using the 4000mhz c15 memory overclock?

DDR4-3200 should unquestionably be the starting point for Ryzen 5000. We recorded a 5.8% performance difference between DDR4-2133 and DDR4-3200, the native frequency supported on Zen 3. Our test results confirmed that DDR4-3600 is still the sweet spot for this generation, though. It delivered 7.2% more performance than DDR4-2133. Compared to DDR4-3200, however, DDR4-3600 was only 1.3% faster.

Unsurprisingly, DDR4-4000 is the ultimate goal if your processor and budget allow for it. The performance gap between DDR4-2133 and DDR4-4000 stretched as high as 7.7%. Compared to DDR4-3200 and DDR4-3600, DDR4-4000 provided small uplifts in the range of 1.8% and 0.5%, respectively.

'In a gaming environment, DDR4-4000 and DDR4-3600 offered 4.2% and 5.6% better frame rates, respectively, than DDR4-2133. '
As always, memory frequency only impacts specific titles where the processor does all the heavy lifting and the game engine responds well to improved memory performance. For example, DDR4-4000 netted up to 6.1% better frame rates than DDR4-2133 in Shadow of the Tomb Raider and up to 19.7% in Far Cry New Dawn.

• Memory ranks often don't contribute to better gaming performance on Ryzen 5000, but memory frequency does in an environment where your graphics card isn't the bottleneck and the title taxes the processor heavily.
• Even on Zen 3, four memory ranks are still the ideal configuration for getting the best overall performance out of your Ryzen 5000 processor, particularly in applications. With today's capacity, this basically translates to having at least 32GB of memory in your system, whether it be two dual-rank or four single-rank memory modules. Most of the time, the first option is always easier on the pocket.

So you're chasing a few % - Maybe they have 4x8gb single rank instead of 2x16gb or maybe they have a different mobo , or both or maybe even a 4000mhz overclock: Is the Cyberpunk game engine one that responds to a higher memory frequency? Who would know that?

Oh more barmy:

The qvl, bottom of page 15 of 19

G.SKILL F4-3600C14D-16GTZN 2x 8GB SS Samsung 14-15-15-35 1.35 Ryzen™ 5000-Series CPU 3600

The 8gb modules are single sided so the 16gb modules must be Dual rank or Double Sided (but not on the qvl exactly 100% precisely) and the product specification page doesn't say if they are in fact, single sided or double sided dimms: Like 8x2 samsung or if they used like larger hynix chips on the actual 16gb dimm so they would be single sided. Have to ask Gskill.

G.SKILL F4-3600C16Q-64GTZNC 4x 16GB DS SK Hynix Well the 64gb kit 16x 4 hynix dimms are cas 16 and double sided.

Could also be 'silicon lottery' between sets of dimms basically you're looking for perfection where all the odd % differences are stacked in your favor.

Well it's totally weird that the Precise Exact information on that simply Isn't There. Well maybe all that means is, science is unable to attain perfection.

 

[logainofhades]

It is not the motherboard, or its VRM's. Their board is more than adequate for a 5800x. Also their userbenchmark is showing their CPU is performing better than expected. The CPU, nor the motherboard, are at fault here. While UB is terrible for comparisons between different CPU's, it can be useful for troubleshooting what component may be giving you trouble. In this case the parts performing worse than expected are the SSD and the ram.

 

[DavidM012]

Yeah but the memory should be optimal at 3600mhz with the 1:1:1 ratio to the inifinity fabric: How did he get 3600mhz without XMP enabled, and the dimms are in the correct sockets?

G.SKILL F4 DDR4 3600 C14 2x16GB
901 User benchmarks, average bench 118%
2 of 4 slots used
32GB DIMM DDR4 clocked @ 3600 MHz

And it's cas14 memory - that if anything should be faster?

 

[logainofhades]

The ram is DDR4 3600, but could be running at a sub optimal timing set, if the ram speed was set manually, instead of enabling XMP.

 

[adsfaasfa]

The ram is DDR4 3600, but could be running at a sub optimal timing set, if the ram speed was set manually, instead of enabling XMP.

I went in my BOIS to set DOCP (XMP). My ratio is indeed 1:1:1 (

https://imgur.com/oLyo4Dj

View: https://imgur.com/oLyo4Dj
)
As for if I should have my CPU overclocked, I only recently overclocked it with a precision boost overdrive method. It did give me a boost in frames.
Also their resolution is 3440 x 1440 and they seem to be using 32GB 3600 CL 16 Ram (

View: https://www.youtube.com/watch?v=1tQtssa3NqI
)
Its not just Cyberpunk btw, I've looked through other gameplays with similar specs and they're achieving higher frames (e.g. Hunt Showdown).
My SSD is getting a bit full but I feel like it won't make a difference since I had it less than %50 full before going down this rabbit hole.
Ill be doing a clean install on my graphics drivers tomorrow when I have the time. If that doesn't work then I believe its justs my OS (Windows 11)
I did clone my OS at one point to my current C: (Samsung 980 PRO 2TB). So I feel like that may be the issue if after doing a clean graphic driver install doesn't do anything.

 

[Karadjgne]

The Secondary and Tertiary timings are a tad loose. That's Samsung B-die ram, so tRFC should be 300, tRDTDSCL 4, tWRWRSCL 4, tFAW 16, tRCDWR can probably be reduced to 10-12. tRRDL should work at 6, maybe even 4. tWTRS/L can probably be reduced to 4/10 or 4/8. tCWL might handle 14. tRDRSD/DD and tWRWRSD/DD can likely handle 4 and 6 etc.

Buildzoid (Actually Hardcore Overclocking) has multiple vids on Samsung B-die timings, what they do/don't do, what's important and what's not.

But being. Ryzen, you've most likely left some performance on the table with Auto timings.

As with anything else from User Benchmark, you are up against the same ram, as in a model number, but that's also judged by what other ppl are doing with that model, so ppl who tweak their pc's gain an advantage in the rankings vs ppl who do not.

I love how Ubm states you are running at 21st % (at 3600 on 3600 ram) so therefore you should follow the link to learn how to enable XMP.