5090 game performance on par with 4090.

[Clarence Boddicker]

In 3DMark my 5090 is getting either average or above scores compared to other 5090’s, but while gaming it seems to be about the same as a 4090, maybe even slightly worse. This is at both 4k and 1440p.

When I first installed it, I was getting really poor benchmark results and lots of crashes. I then changed the pcie Gen setting in the bios from auto to 4 and this solved the problem, benchmarks up to where they should be and no more crashes.

Something just doesn’t seem right when gaming though. When comparing to videos on YouTube with the exact same game settings, I’m a good 20-40% behind other 5090’s. I’ve got a 14700k while they are mostly using 9800x3d’s but that shouldn’t cause such a difference in fps though should it?

I haven’t tested on may games yet, mainly Spider-Man 2, Ratchet and Clank and war zone but it’s not performing on par in any of those games. Spiderman 2, native at max settings+RT, I’m in the 40’s even dipping down into the high 30’s.

Just wondering if anyone’s got any ideas, any bios/windows settings etc that I may have overlooked? Seems strange that 3DMark scores are fine but game fps is so much worse than others?

Edit: Forgot to mention that Ive got a b760 mobo and ram is ddr5 5200, not the best I know but shouldnt hold back performance that much should it?

 

[Tetras]

I’m a good 20-40% behind other 5090’s. I’ve got a 14700k while they are mostly using 9800x3d’s but that shouldn’t cause such a difference in fps though should it?

It is hard to say, because it would depend on the game and your settings, but I'd expect the overall average to be lower than that, especially at 4K.

Edit: Forgot to mention that Ive got a b760 mobo and ram is ddr5 5200, not the best I know but shouldnt hold back performance that much should it?

Again: yes, it could, you have a GPU that is at the bleeding edge, so pulling less performance from the 14700K with slow RAM won't help. When you use max settings and RT that tends to load the CPU up too (seem to remember Spiderman being one of those that does that) and the percentage difference between hardware gets greatly exaggerated in those kind of extreme bottlenecks.

 

[Clarence Boddicker]

Again: yes, it could, you have a GPU that is at the bleeding edge, so pulling less performance from the 14700K with slow RAM won't help. When you use max settings and RT that tends to load the CPU up too (seem to remember Spiderman being one of those that does that) and the percentage difference between hardware gets greatly exaggerated in those kind of extreme bottlenecks.

I was hoping to keep this mobo and cpu for another year or two but if its causing this amount of performance loss then Im going to have to rethink that.

Do you it’s worth getting some 6400 MHz RAM first and see if that makes a difference?

 

[Rroff]

5200MT is pretty slow for the 14700K but not going to cause 20-40% lower performance. Might want to check the 14700K is boosting properly and not thermal throttling, etc.

 

[Tetras]

but if its causing this amount of performance loss then Im going to have to rethink that.

Comparing your PC to third party videos and benchmarks isn't just gaming though, I mean, is the performance of your PC not good enough to enjoy your games? I'd expect a 14700K to get most of the performance of a 5090 at 4K, but even the 9800X3D arguably bottlenecks a 5090 at times.

Do you it’s worth getting some 6400 MHz RAM first and see if that makes a difference?

I don't know about anyone else, but the problem I have in making suggestions is that:
1. All the reviews of the 5090 were with a 9800X3D.
2. The RAM benchmarks that show a difference are inevitably at 1080p.

With a 5090 you're right at the edge and going to be pushing bottlenecks where 99% of gaming builds are not going to, so you could be in a situation where slow RAM impacts just 1 game, at 1 set of settings and no other game. Drivers (and even Windows updates/feature versions) could potentially make a big difference too, especially if you're playing games at max settings at sub 100 fps.

In the absence of reliable comparisons, I'd suggest:
- Check hardware temps and power limits to make sure there's no throttling.
- Monitor the utilisation % in your games when you're experiencing problems, see if you can learn anything from that.
- Underclock the CPU (e.g. turn off the turbo) and see if that makes any difference. If it does not, you don't have a CPU bottleneck.
- Lower the settings (e.g. turn off RT, lower the details, lower the resolution) and if that makes a big difference you're probably GPU bottlenecked). Note that 4K gaming is still very tough in the latest games, even for a 5090.
- Try overclocking your current memory and see if it helps, 5200 is very slow and I'd expect almost all 5200 memory to be comfortably overclockable. Do read some proper guides first though, don't just go in the BIOS and select 6000.

 

[Clarence Boddicker]

5200MT is pretty slow for the 14700K but not going to cause 20-40% lower performance. Might want to check the 14700K is boosting properly and not thermal throttling, etc.

Im pretty sure the 14700k is fine. Ive got a good stable undervolt on it and it scores over 36000 in cinebench with no thermal throttling.

 

[Bidley]

When the 9800X3D reviews came out, some reviewers were finding the 4090 had increased performance, which indicates a degree of CPU bottleneck even at the top tier. Similar was said of the 5090 in reviews.

That being said, CPU bottlenecks aren't linear anymore and depend on the game and specific scenario in the game.

I wouldn't expect a 14700k or slightly slower RAM to lose you 20-40%. That's significant.

I'd be looking at drivers. I recently updated to the latest Nvidia driver and got about 5% increased performance. A couple of days later, I randomly turned my PC on and found I was getting another 2/3% more performance. I hadn't even installed or uninstalled anything, or changed any Afterburner settings. It was entirely repeatable.

I would try clean reinstalling drivers, or double-DDUing and seeing what that does.

 

[Clarence Boddicker]

I'd be looking at drivers. I recently updated to the latest Nvidia driver and got about 5% increased performance. A couple of days later, I randomly turned my PC on and found I was getting another 2/3% more performance. I hadn't even installed or uninstalled anything, or changed any Afterburner settings. It was entirely repeatable.

I would try clean reinstalling drivers, or double-DDUing and seeing what that does.

Ive actually already DDU’d and its made no difference. Strangely Ive also got a small increase from the new drivers but that is only in 3DMark benchmarks. I get 14822 in Steel Nomad now, up from about 14300. It just seems to be games that I get a lack of performance.

I forgot to mention in the OP that I only got 150-200 fps in Mech Warriors 5: Mercenaries, thats with RT off, a 4090 on youtube paired with a 5800x3d gets 300-400. Literally double my fps.

 

[Tetras]

I forgot to mention in the OP that I only got 150-200 fps in Mech Warriors 5: Mercenaries, thats with RT off, a 4090 on youtube paired with a 5800x3d gets 300-400. Literally double my fps.

Are the settings and the resolution used exactly the same as yours? E.g. same resolution, no upscaling, frame generation, the details the same? Was your GPU showing decent utilisation and no throttling?

I don't know this game at all, but was the place also the same? E.g. you weren't in a super busy area and they were in an open desert?

Note that the drivers of the 5090 aren't exactly... polished and since some 32-bit stuff (like PhysX) is depreciated, it can end up with much lower performance than expected in some games.

I get 14822 in Steel Nomad now, up from about 14300.

The score sounds about right, which suggests you don't have a major hardware issue.

 

[Clarence Boddicker]

Are the settings and the resolution used exactly the same as yours? E.g. same resolution, no upscaling, frame generation, the details the same? Was your GPU showing decent utilisation and no throttling?

I don't know this game at all, but was the place also the same? E.g. you weren't in a super busy area and they were in an open desert?

Yep, settings and resolution exactly the same, I cant remember my gpu utilization but it was a different part of the game I was at and was actually less busy than the area in the video.

Ive been looking at comparisons between 5200 and 6400 RAM and it does seem that it can make a considerable difference in some cases. I think Im going to upgrade it.

May I ask you what speed you would go for? My mobo supports up to 7600 but from what I’ve read 6000-6400 seems to be the sweet spot.

What do you think about 2x16GB 6400 cl32?

 

[Grim5]

In 3DMark my 5090 is getting either average or above scores compared to other 5090’s, but while gaming it seems to be about the same as a 4090, maybe even slightly worse. This is at both 4k and 1440p.

When I first installed it, I was getting really poor benchmark results and lots of crashes. I then changed the pcie Gen setting in the bios from auto to 4 and this solved the problem, benchmarks up to where they should be and no more crashes.

Something just doesn’t seem right when gaming though. When comparing to videos on YouTube with the exact same game settings, I’m a good 20-40% behind other 5090’s. I’ve got a 14700k while they are mostly using 9800x3d’s but that shouldn’t cause such a difference in fps though should it?

I haven’t tested on may games yet, mainly Spider-Man 2, Ratchet and Clank and war zone but it’s not performing on par in any of those games. Spiderman 2, native at max settings+RT, I’m in the 40’s even dipping down into the high 30’s.

Just wondering if anyone’s got any ideas, any bios/windows settings etc that I may have overlooked? Seems strange that 3DMark scores are fine but game fps is so much worse than others?

Edit: Forgot to mention that Ive got a b760 mobo and ram is ddr5 5200, not the best I know but shouldnt hold back performance that much should it?

5090 is just not much of an upgrade over the 4090 and secondly, to actually access any performance advantage of the 5090 requires the best cpu/ram on the market and you don't have that so you've wasted your money only buying the 5090

 

[Tetras]

Ive been looking at comparisons between 5200 and 6400 RAM and it does seem that it can make a considerable difference in some cases. I think Im going to upgrade it.

I'd strongly recommend you try overclocking what you have first, especially if you're junking it anyway. It will confirm for you if it makes any meaningful difference in the exact scenarios you play.

May I ask you what speed you would go for? My mobo supports up to 7600 but from what I’ve read 6000-6400 seems to be the sweet spot.

I'm not the person to help you with that, but from what I can recall watching buildzoid's videos, 13th-14th gen CPUs support up to 7200 fairly comfortably, but above that it gets tricky and I wouldn't want to pay the premium prices anyhow. 6000-6400 tends to be best value, but those recommendations are mainly for AMD systems because it loses performance when you have to change the memory divider from 1:1.

Yep, settings and resolution exactly the same, I cant remember my gpu utilization but it was a different part of the game I was at and was actually less busy than the area in the video.

I see, well, I still wouldn't put much stock in one game and one video, it could be anything really..., a driver issue with that specific game, something that needs optimisation, or even a fake video.

I'd want something that you could compare with a more reliable source, like TechPowerUp, or someone who does lots of videos like Daniel Owen.

 

[Varys]

Just guessing but I expect X3D and ddr5 5200 is your answer, probably bottlenecked.

"Im pretty sure the 14700k is fine. Ive got a good stable undervolt on it and it scores over 36000 in cinebench with no thermal throttling." - I score 46000 in cinebench so thats nearly 28% difference there. Also benchmarks often are often not seen by the computer as games (so bench performance can be higher)

Don't forget, if you are using DLSS or frame gen, even though you have a bleeding edge graphics card, the game is rendered at a lower resolution, a lot more work is put on the CPU and Ram to process the upscaling and additional frames.

Try swapping your ram first (I'd suggest getting some that is also EXPO compatible in case you later decide to go the AMD route), then if no joy might be time for new mobo and cpu

Just had a quick look and the 9800x3d is about 20-30% better in gaming than the 14700k at 4k (low ram speed aside), although the Intel is better for productivity. If you need a prod chip go for the 9950x3d

https://tech4gamers.com/ryzen-7-9800x3d-vs-core-i7-14700k/

Added some 1080p benchmarks (before anyone starts crying), which also includes 9950x3d and productivity benches:-

https://gamersnexus.net/cpus/amd-ry...00x3d-285k-9950x-more#9950x3d-game-benchmarks

 

[Clarence Boddicker]

Just had a quick look and the 9800x3d is about 20-30% better in gaming than the 14700k at 4k (low ram speed aside), although the Intel is better for productivity. If you need a prod chip go for the 9950x3d

https://tech4gamers.com/ryzen-7-9800x3d-vs-core-i7-14700k/

Ok so the 20-40% difference Im seeing is about right then. I think like you said, I’ll swap to EXPO compatible RAM for now and see what difference that makes.

 

[stellasuppa]

have you checked gpu z to see if your missing any ROPs ?

 

[Clarence Boddicker]

have you checked gpu z to see if your missing any ROPs ?

I have, all there. I think it must be a combination of cpu and slow ram causing it.

 

[Varys]

Ok so the 20-40% difference Im seeing is about right then. I think like you said, I’ll swap to EXPO compatible RAM for now and see what difference that makes.

Try Gskill 6000Mhz cl30. The Expo kit is also compatible with Intel chips but you might have to manually tweak the timings, but its probably the CPU accounting for most of the difference

 

[SuperElectric]

Ive actually already DDU’d and its made no difference. Strangely Ive also got a small increase from the new drivers but that is only in 3DMark benchmarks. I get 14822 in Steel Nomad now, up from about 14300. It just seems to be games that I get a lack of performance.

I forgot to mention in the OP that I only got 150-200 fps in Mech Warriors 5: Mercenaries, thats with RT off, a 4090 on youtube paired with a 5800x3d gets 300-400. Literally double my fps.

What do you get in Spiderman 2 and Ratchet & Clank? Be specific with your settings.

EDIT: Sorry, missed that you said 40's in Spiderman 2 with native max RT. Yeah that sucks.

These guys get 72.7fps with a 13600K with DDR5-5600 in Spiderman 2, Max Ray Tracing. I would guess that something is wrong with your CPU, maybe the degradation problem that Intel had over the last two years. I would try a lot of CPU benchmarks first, if it falls more than 5% below what a 14700K is supposed to get, then the CPU might be faulty.

 

[Clarence Boddicker]

What do you get in Spiderman 2 and Ratchet & Clank? Be specific with your settings.

EDIT: Sorry, missed that you said 40's in Spiderman 2 with native max RT. Yeah that sucks.

These guys get 72.7fps with a 13600K with DDR5-5600 in Spiderman 2, Max Ray Tracing. I would guess that something is wrong with your CPU, maybe the degradation problem that Intel had over the last two years. I would try a lot of CPU benchmarks first, if it falls more than 5% below what a 14700K is supposed to get, then the CPU might be faulty.

I really don’t think there’s anything wrong with the cpu, I’ve got a good stable undervolt on its and never had any signs degradation, crashes, bsod etc. It scores 36000 on cinebench. That said there’s obviously something very wrong if my fps is getting spanked by a 13600k and 4090.

 

[SuperElectric]

I really don’t think there’s anything wrong with the cpu, I’ve got a good stable undervolt on its and never had any signs degradation, crashes, bsod etc. It scores 36000 on cinebench. That said there’s obviously something very wrong if my fps is getting spanked by a 13600k and 4090.

Interesting conundrum, both CPU and GPU synthetic benchmarks are correct, but games are not?

If it was me, next step is to watch what the CPU and GPU are doing in HW Monitor while the game is on. Are they both i) hitting the correct clocks, ii) using the correct power?

Then I would reinstall windows, and if that doesn't work, I'd just fire up the parts cannon until the problem went away