Rx 5700 XT Underpeforming?

[UBatcha]

I've felt like my PC has been underperforming but I thought it might be driver issues or something else but now I have the time I decided to benchmark it properly using the standard programs like 3D mark, Unigine etc and I know the results are lower than expected compared to other people.
This was confirmed by my Userbenchmark score - https://www.userbenchmark.com/UserRun/27284032

I can give more information if needed but I have updated to the latest BIOS and as far as I know all drivers seem up to date

RX 5700XT is the Powercolor Red Dragon and has not been overclocked (Running latest drivers so 20.4.2 I think)

I've run Superposition a few times and I always get just under or over 5,000 (From 4,955 to the best being 5005)

[SavageNeo]

do not trust userbenchmark.

for example run unigine heaven benchamrk and compare scores to other 5700 xt's

[Mister Woof]

43 minutes ago, UBatcha said:

I've felt like my PC has been underperforming but I thought it might be driver issues or something else but now I have the time I decided to benchmark it properly using the standard programs like 3D mark, Unigine etc and I know the results are lower than expected compared to other people.
This was confirmed by my Userbenchmark score - https://www.userbenchmark.com/UserRun/27284032

I can give more information if needed but I have updated to the latest BIOS and as far as I know all drivers seem up to date

RX 5700XT is the Powercolor Red Dragon and has not been overclocked (Running latest drivers so 20.4.2 I think)

I've run Superposition a few times and I always get just under or over 5,000 (From 4,955 to the best being 5005)

Edit: i also have used userbenchmark and it showed underperforming 5700xt. it's stupid. my 5700xt performs better based on comaprison with other reviewer's game benchmarks.

 

https://www.userbenchmark.com/UserRun/27285953

 

17th percentile lol, what a crack of shit. I undervolted mine and set it to a lower frequency so the card doesn't turn into a jet engine furnace.

 

Edit: 2nd edit - wrong photos. this is correct now for my undervolt settings.

 

Here's my 5700XT for comparison, in a free benchmark (Shadow of the Tomb Raider trial). The latter is 100% GPU bound, so that should remove any processor differences.

 

1080p

 

 

1440p

 

 

 

[Mister Woof]

and for reference, i loaded up a completely unusable GPU profile (2200mhz core, +50% power limit, 1850mhz GDDR, will throttle in actual real games that stress it)

 

https://www.userbenchmark.com/UserRun/27286195

 

it's bullshit. people do it just to skew metrics.

 

1080p w/ridiculous profile:

 

1440p ridiculous profile

 

Minimal difference in real world performance, but userbenchmark says 17th percentile for the first one and 90th percentile for the latter. Firmly believe people just game it for some kind of bragging rights (i mean, it's what I would do, LOL)

 

I'll take my undervolted setting, thanks.

 

edit: i got totally mixed up on my screenshot folder, so pardon my many edits.

[xAcid9]

Stop using Userbench garbage. What is your score in Timespy/Firestrike?

[Xkillerpn]

1 dont forget that gpu needs also a good cpu.

 

Even a ryzen 1600 wil be a bottleneck for that at 1080p

[Gohardgrandpa]

26 minutes ago, Xkillerpn said:

1 dont forget that gpu needs also a good cpu.

 

Even a ryzen 1600 wil be a bottleneck for that at 1080p

The user bench they linked showed a ryzen 5 3600. You shouldn’t have issues with that setup and it’s almost the same as mine and I have no complaints with my system. 
@UBatcha

Im at work now but I’ll run userbench on my pc when I get home and post it here so you can see what I get score wise. 

[UBatcha]

So update, I did some further testing and these are the results I came out with (a few of these results improved slightly with a bios update for example Cinebench is now 3224): 

 

I'm starting to think it's less to do with my GPU and more to do with my Ryzen 3600 CPU underperforming for example, I just ran Cinebench R20 and It didn't even boost to 4.2ghz:

[Mister Woof]

14 minutes ago, UBatcha said:

So update, I did some further testing and these are the results I came out with (a few of these results improved slightly with a bios update for example Cinebench is now 3224): 

 

I'm starting to think it's less to do with my GPU and more to do with my Ryzen 3600 CPU underperforming for example, I just ran Cinebench R20 and It didn't even boost to 4.2ghz:

all core on the 3600 is slightly under 4ghz; ~3600 score is the reported performance for stock 3600 PBO off.

 

If you have some background tasks running while you benchmark, then 32xx score doesn't sound too fishy.

[UBatcha]

40 minutes ago, Mister Woof said:

all core on the 3600 is slightly under 4ghz; ~3600 score is the reported performance for stock 3600 PBO off.

 

If you have some background tasks running while you benchmark, then 32xx score doesn't sound too fishy.

I understand that however isn't at least one core supposed to boost to 4.2Ghz and if not then do I have grounds to get an RMA - Current situation means I'll wait to get it RMA'd but I want to make sure I'm getting the full performance of what I paid for

Also I didn't have much running in the background, standard windows stuff thats all

[Mister Woof]

5 minutes ago, UBatcha said:

I understand that however isn't at least one core supposed to boost to 4.2Ghz and if not then do I have grounds to get an RMA - Current situation means I'll wait to get it RMA'd but I want to make sure I'm getting the full performance of what I paid for

Also I didn't have much running in the background, standard windows stuff thats all

you'll probably never see single core advertised boost in any real world performance. maybe close in apps that only use 2 or less cores, because OS always utilizes something in the background. You do not have grounds for RMA. They always say "up to", which is not a guarantee. Intel does the same thing.

[BTGbullseye]

On 4/28/2020 at 9:25 AM, UBatcha said:

I've felt like my PC has been underperforming but I thought it might be driver issues or something else but now I have the time I decided to benchmark it properly using the standard programs like 3D mark, Unigine etc and I know the results are lower than expected compared to other people.
This was confirmed by my Userbenchmark score - https://www.userbenchmark.com/UserRun/27284032

As a reference, here is my system (similar to yours) that is currently #3 on the GPU Superposition top performers list here on the LTT forums... https://www.userbenchmark.com/UserRun/27369723 (note, this benchmark was run with a reference model 5700 XT at stock settings, not my bleeding-edge overclock)

 

Your system is 100% not underperforming according to that comparison.

 

Almost all the higher performing cards are really just unstable overclocks that people are running exclusively for the test, much like some people do for LN2 cooling world records. (or are using different drivers/older tests that produced higher results)

On 4/28/2020 at 9:28 AM, SavageNeo said:

do not trust userbenchmark.

It is useful for general purpose testing to see what parts of a system may be underperforming. It's not perfect, no, but it isn't bad to use when you have the knowledge of how to properly compare the numbers.