Ryzen 7 5800X / RX 7900 XTX Bottlenecked!

[Larry]

I need help in identifying what's bottlenecking my system.

I'm scratching my head here wondering how hardware benchmarks are getting 200+ FPS in games at 1440p on the 7900XTX while I'm struggling to get 150 at 1080p with similar specs and settings.

For one, the likes of Hardware Unboxed are getting an average FPS of 260(DX11) @1440p in the likes of The Witcher 3 meanwhile I'm getting 120-150FPS max at 1080p. This is similar in Metro Exodus.

https://youtu.be/dnxXT2sx8nA?t=355 The above video shows they're using ultra quality settings in DX11 and getting these kind of frames. It's blowing my mind.

Hardware Unboxed setup:

RX 7900 XTX
Ryzen 7 5800X3D(Stock?)
32GB 3200Mhz CL14 DDR4 RAM
MSI MPG X570 Carbon Max Wifi

My Specs:

RX 7900 XTX
Ryzen 7 5800X ~ Stock
32GB Dual Channel Dual Rank 3000Mhz CL14 DDR4 RAM
MSI MPG X570 Gaming Plus
Corsair 850W Platinum rated PSU

What's bottlenecking me here? I don't understand. The 5800X3D is the better CPU in terms of gaming but not 100+% better. The RAM also couldn't account for over 100% performance difference could it?

I will note that checking into this yesterday I found I did not have XMP/A-OMP(Whatever the AMD equivalent is called) enabled, turning this on resulted in my FPS increasing by 60FPS(From 60-70fps in built up areas to 110~) as my RAM was only clocking at 2400Mhz as opposed to the 3000Mhz it should have been clocked at.

I've also double checked to make sure my GPU is plugged into PCIe x16 slot and not the slower PCIe slot. I also double checked my RAM is in the correct slots(Apparently it is; A2 + B2).

This is driving me nuts!

[Cathulhu]

For a simple first check, could you do a simple benchmark from:
https://www.userbenchmark.com/
And then link the results? That gives us a few pointers of where to look.

Is AMD Resizable BAR activated in the BIOS?

[🦜Cloud Boy🦜]

Running 2400Mhz RAM will surely reduce performance of your system quite a lot.

The 5800X3D can be lot faster than other CPUs if the game is cache intensive.

YouTubers such as Hardware-Unboxed are professionals, they use a totally clean PC with fresh OS installed, nothing running in the background. So theirs results can be different from an average PC user. Also, they use other available methods to increase the fps, such as Resizable-bars.

[UserNotFound]

Userbenchmark is useless, it'd skew results against AMD, both CPU and GPU. Perhaps Firestrike Extreme would be a better benchmark tool...it's been said the RTX 4090 would be bottlenecked by even the fastest Intel CPU (excluding the world record of 8-9GHz?). Wheteher the RX 7900 XTX would be likewise bottlenecked, I ain't sure as it's not as fast as the RTX 4090.

[Cathulhu]

Again, it gives POINTERS. An idea where to look at.

[GOD RAYS ON ULTRA™]

What's there to look at? The OPer isn't even running the same settings. The OPer is limiting their CPU with 1080p resolution on a GPU that was designed for much higher.

For 1080p I would've went for a different GPU. The 7900xtx is a waste of money here.

[UserNotFound]

It begs the question, waas the RX 7900 XTX installed in a completely new build, hence new OS and such. Or, was there previously another card in that rig before the upgrade. If so, did OP run DDU first to remove all traces of previous driver before installing the card?

I've never encountered a performance issue because I'd always do the following when I install a new card (previous card can be AMD or nVidia, doesn't matter as I'd do the same)
1. Download latest driver, ensure it's in a easily accessible drive in your system
2. Download DDU, unplug LAN/Ethernet on your PC from (to prevent access to net)
3. Run DDU, set it to run in Safe Mode, you may have to restart DDU after this
4. Run DDU again and do as prompted, including restarting in Safe Mode and uninstalling driver
5. Upon reboot, install latest driver, and reboot if prompted
6. Reconnect to net and you should be ready to go

Install driver by simply uninstalling previous driver isn't gonna rid your PC of all traces of previous driver (which may be in your PC registry), running DDU would wipe it from your PC ensuring a 'clean' install of newer driver.

From my experience, AMD GPU's are a little more fickle about driver installation, a clean install like this ensure peak performance.

[Illusion of Progress]

If you're not running basically EXACTLY the same system, then the results they give aren't something you should compare to. Variables have to be controlled. Now, sure, sometimes you can expect within a certain range with "close enough" hardware, or sometimes a given thing you're comparing/testing might only rely on a part or two, and those are the same, but even then you might not get exactly the same results (I found this out when my 3700X was getting a score basically exactly between a 3600X and 3700X in Cinebench years ago, which is mostly just CPU from what I understand).

The 5800X and 5800X3D are not at all comparable. The 5800X3D really can be up to wildly faster, or sometimes no faster (maybe even SLIGHTLY slower).

RAM can do a lot, especially on Ryzen. The Infinity Fabric is basically a FSB-esque thing that ties a lot of things together insofar as communicating goes, and by default, this will run 1:1 with RAM (with real RAM speed, not the effective DDR speed). Zen 2 (Ryzen 3000 series) and up basically like 3,600 Mhz RAM as they generally never fail to run a 1,800 MHz IF clock 9really it's 3,733 MHz and 1,866 MHz but that's a more niche RAM speed). Sometimes a bit higher, but it's luck beyond that point. 2,400 MHz is a 1,200 MHz IF clock. That's rather slow on top of the RAM being slower. You really want 3,600 MHz for Zen 2 or above, but 3,200 MHz is alright. With 3,000 MHz, eh, probably not worth replacing over unless you're also interested in going up in capacity anyway.

Then there's things like Windows updates, drivers, BIOS settings, settings in games, etc., etc., etc.

[Larry]

Hi everyone,

Thanks for your replies and input.

The main issue I'm having with this is how I can narrow the gap. Where's the likely bottleneck here?

I know there's a bottleneck but before I start dishing out cash on a fix I'd like a bit of insight into what it likely is. Is it likely the RAM? Will I need to buy 3600Mhz CL16 RAM? Going from 2400Mhz to OC 3000Mhz had a drastic impact on my performance(At least in Metro and The Witcher 3) but will 3600Mhz bring me much closer to the 260fps these reviewers are experiencing?

To answer some posts here:

- I did a clean install of the new GPU(Using DDU, rebooting, complete 100% fresh install).

- A 7900 XTX is wasted on 1080p but I won't be upgrading to a 1440p 144Hz monitor until I address this performance issue.

- I'm using the same settings as Hardware Unboxed except at a lower resolution which should result in even more FPS I would think.

- I do expect some variables in performance but not a massive hit as I'm experiencing here. It's like I'm losing out on 80% of the performance of my hardware. I think we can all agree that a 5800X vs a 5800X3D wouldn't amount to losing out on 100 or more fps, if anything it would be 10-20fps.

There's got to be something else going on here to make such a drastic difference with my machine. I was leaning toward the RAM but for the sake of 200Mhz(When my RAM is OC'ed) I'm wondering if there's something else like RAM configuration(A2-B2 however MSI advise of this dual memory slot allocation).

[emoticorpse]

If you do that DLSS thing where you upscale it to simulate 1440p or whatever what are the fps?

[Illusion of Progress]

I couldn't answer you on how much performance faster RAM will bring, or if it's worth it.

The CPU being a 5800X3D is a wild card though. It's between "basically nothing" or "a way lot more" compared to the 5800X. It's honestly just better to pretend it's a wildly varying Zen 3+ and not consider it a slightly better Zen 3 because it is NOT comparable to the normal Zen 3 CPUs at all. It's too variable. In SOME cases it competes right with or even slightly above the Ryzen 7000 series. Of course it's not as good as them on average, but you can't just go "it should only be 10 FPS more". Maybe it's more, maybe it's slightly less (worst case scenario, it is Zen 3 performance or even a hair less). You'd need to find 5800X3D reviews (which hopefully include the 5800X, if not the 5700X is close enough) for an exact game to be sure of the difference, and even then THOSE tests are "uncontrolled" compared to another tests results, because of those things called uncontrolled variables. Looking at differences in other games and guesstimating it based on an average isn't good enough, not with this particular CPU especially.

In general though, trying to go "I have the same part A and am using the same settings except this but get results more different than I expected" is ultimately going to be down to two things.

Your hardware is flawed and under-performing as a result (typically rather unlikely).

The things that aren't the same, the variables, are the cause.

Those are... the only two explanations really, no?

Proper testing methodology for comparison is having ALL of the variables "controlled". That means, accounted for. If not, then those things are likely where difference arises from.

Different BIOS, different RAM (frequency, timings rank), different game updates, different GPU drivers, different Windows environments/updates, it goes on and on and on, and all these sorts of things are all called "uncontrolled" unless you ensure you have both systems on hand and can verify every little thing. Sure, you can look a fair bit of this up and review sites try and disclose what they can... but the principle remains.

[pasa]

Do you use software to display cpu/gpu usage % and temps? Is your cpu at 100% and reasonably cool?

Also, is the low fps specific to a single game or applies evently to several?

[Skoops]

Publicado originalmente por Hardʬare Hero:

What's there to look at? The OPer isn't even running the same settings. The OPer is limiting their CPU with 1080p resolution on a GPU that was designed for much higher.

For 1080p I would've went for a different GPU. The 7900xtx is a waste of money here.

This^^^

You are bottlencked by your CPU at 1080p that's why you're getting less frames than 1440p.

Google " CPU bottle neck @1080p " and you'll find several in depth explanations.

[emoticorpse]

Publicado originalmente por Skoops:

Publicado originalmente por Hardʬare Hero:

What's there to look at? The OPer isn't even running the same settings. The OPer is limiting their CPU with 1080p resolution on a GPU that was designed for much higher.

For 1080p I would've went for a different GPU. The 7900xtx is a waste of money here.

This^^^

You are bottlencked by your CPU at 1080p that's why you're getting less frames than 1440p.

Google " CPU bottle neck @1080p " and you'll find several in depth explanations.

I'm not saying OP's pc is underperforming or overperforming but I really don't get the logic here. I do understand that "bottlenecking" eats away at efficiency/performance but the way it's presented by Hardware Hero (and you?) I am interpreting the claim of a GPU not being able to perform at 1080p because it wasn't designed to almost like saying a racing car can't do 55 mph because it was designed for 200+ mph or something like that.

I mean, absolutely regardless of the situation, hardware should be able to handle 1080p better than 1440p right?. Am I wrong in simplifying it this much?. Isn't he supposed to get more fps in 1080p than 1440p in every case scenario? Or am I mistaken here?

I wouldn't mind being educated since I really haven't dug too deep into the specifics of "bottlenecking".

[The Rt Hon. Sir_Sampie]

Publicado originalmente por emoticorpse:

Publicado originalmente por Skoops:

This^^^

You are bottlencked by your CPU at 1080p that's why you're getting less frames than 1440p.

Google " CPU bottle neck @1080p " and you'll find several in depth explanations.

I'm not saying OP's pc is underperforming or overperforming but I really don't get the logic here. I do understand that "bottlenecking" eats away at efficiency/performance but the way it's presented by Hardware Hero (and you?) I am interpreting the claim of a GPU not being able to perform at 1080p because it wasn't designed to almost like saying a racing car can't do 55 mph because it was designed for 200+ mph or something like that.

I mean, absolutely regardless of the situation, hardware should be able to handle 1080p better than 1440p right?. Am I wrong in simplifying it this much?. Isn't he supposed to get more fps in 1080p than 1440p in every case scenario? Or am I mistaken here?

I wouldn't mind being educated since I really haven't dug too deep into the specifics of "bottlenecking".

I've never heard of a situation where a card would perform worse at a lower resolution, at the very most I would expect to see the same or similar performance if the CPU bottleneck was that severe.

To OP's issue, I'm not sure what's going on. Your best bet would be to do some tinkering of your own to narrow down the issue (if it's even an issue at all). For example, testing other games and comparing the frame rates you get, checking other reviews and benchmarks to see if they are also wildly different, maybe even be as drastic as reinstalling windows to see if that has any effect (albeit, that's a inconvenient one). You may even try using some faster memory to see how tangible the performance gains are.

Hope you figure out the problem!

[emoticorpse]

Publicado originalmente por Sir_Sampie:

Publicado originalmente por emoticorpse:

I'm not saying OP's pc is underperforming or overperforming but I really don't get the logic here. I do understand that "bottlenecking" eats away at efficiency/performance but the way it's presented by Hardware Hero (and you?) I am interpreting the claim of a GPU not being able to perform at 1080p because it wasn't designed to almost like saying a racing car can't do 55 mph because it was designed for 200+ mph or something like that.

I mean, absolutely regardless of the situation, hardware should be able to handle 1080p better than 1440p right?. Am I wrong in simplifying it this much?. Isn't he supposed to get more fps in 1080p than 1440p in every case scenario? Or am I mistaken here?

I wouldn't mind being educated since I really haven't dug too deep into the specifics of "bottlenecking".

I've never heard of a situation where a card would perform worse at a lower resolution, at the very most I would expect to see the same or similar performance if the CPU bottleneck was that severe.

To OP's issue, I'm not sure what's going on. Your best bet would be to do some tinkering of your own to narrow down the issue (if it's even an issue at all). For example, testing other games and comparing the frame rates you get, checking other reviews and benchmarks to see if they are also wildly different, maybe even be as drastic as reinstalling windows to see if that has any effect (albeit, that's a inconvenient one). You may even try using some faster memory to see how tangible the performance gains are.

Hope you figure out the problem!

yeah, I know but post #12 seems to be defending lower fps at lower resolution. Not sure if I'm misenterpreting it or not.