• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Radeon RX 7900 GRE - Review Thread

winjer

Member















relative-performance-2560-1440.png


picture26.webp
 
Last edited:

Pagusas

Elden Member
Seemes easy enough for nvidia to counter, drop the 4070 super to $600. and the 4070 TI Super to $700. Spend an extra $50 for a more premium product make sense.
 
Last edited:

Bry0

Member
hPFZR1b.jpg


Much bigger difference in TPU tests between 7800XT and 7900GRE, maybe card is very power limited and some versions have that increased?
6% vs 9% isn’t that big. I’m sure it is probably power limited though. It seems like upping that gets you 5-8% more performance pretty easily.
 

Gaiff

SBI’s Resident Gaslighter
Great stuff but once again, the segment that's in dire need of competitive products is the $200-300 segment.

Also, the results don't add up. Richard from DF has the 7900 GRE beating the 4070 Super by a whopping 27% in Cyberpunk at 1440p Ultra but Hardware Unboxed has them tied at 1440p High settings. Rich also says the 7900 GRE is 20% faster than the 7800 XT in Cyberpunk but Hardware Unboxed has it a mere 2% ahead and Daniel Owen only 7% ahead at 1440p. That's without ray tracing.

Richard from DF came away very impressed and the 7900 GRE consistently boasted huge leads over the 7800, 4070, and sometimes even the 4070S in rasterization. HU on the other hand were disappointed.

Yeah, the results of that card are all over the place.
 

Kenpachii

Member
Great stuff but once again, the segment that's in dire need of competitive products is the $200-300 segment.

Also, the results don't add up. Richard from DF has the 7900 GRE beating the 4070 Super by a whopping 27% in Cyberpunk at 1440p Ultra but Hardware Unboxed has them tied at 1440p High settings. Rich also says the 7900 GRE is 20% faster than the 7800 XT in Cyberpunk but Hardware Unboxed has it a mere 2% ahead and Daniel Owen only 7% ahead at 1440p. That's without ray tracing.

Richard from DF came away very impressed and the 7900 GRE consistently boasted huge leads over the 7800, 4070, and sometimes even the 4070S in rasterization. HU on the other hand were disappointed.

Yeah, the results of that card are all over the place.

Can these benchers start to go back to there roots and actually benchmark games propperly instead of this selective bullshit benching.

U pick games that are the most graphical demanding, enable everything it has and see the performance.

This AMD sugarcoating is getting out of hand.

Call me up when they actually use pathtracing cyberpunk at 1440p max settings with dlss + framegen vs amd version.

Rast performance hasn't been relevant for a while now.

Benching at high settings in cyberpunk with a 500+ euro GPU is laughable.
 
Last edited:

Bojji

Member
Can these benchers start to go back to there roots and actually benchmark games propperly instead of this selective bullshit benching.

U pick games that are the most graphical demanding, enable everything it has and see the performance.

This AMD sugarcoating is getting out of hand.

Call me up when they actually use pathtracing cyberpunk at 1440p max settings with dlss + framegen vs amd version.

Rast performance hasn't been relevant for a while now.

Benching at high settings in cyberpunk with a 500+ euro GPU is laughable.

Yeah it should be ultra for every game but tests with RT and without rt should be there as well.

Daniel Owen has best performance delta between 7800xt - 7900gre test.

DF on the other hand has many RT heavy tests, I don't know...
 

winjer

Member
Great stuff but once again, the segment that's in dire need of competitive products is the $200-300 segment.

Also, the results don't add up. Richard from DF has the 7900 GRE beating the 4070 Super by a whopping 27% in Cyberpunk at 1440p Ultra but Hardware Unboxed has them tied at 1440p High settings. Rich also says the 7900 GRE is 20% faster than the 7800 XT in Cyberpunk but Hardware Unboxed has it a mere 2% ahead and Daniel Owen only 7% ahead at 1440p. That's without ray tracing.

Richard from DF came away very impressed and the 7900 GRE consistently boasted huge leads over the 7800, 4070, and sometimes even the 4070S in rasterization. HU on the other hand were disappointed.

Yeah, the results of that card are all over the place.

In rasterization, the 7900 GRE does significantly better than in rasterization, when compared to the 7800XT.
I think that the 7800Xt only has a 128kb vector register file. While the 7900 GRE has 192kb. On RDNA3, this can have a significant impact for RT loads.

HU separates their RT and rasterization tests. And that graph that was posted, is only for rasterization.
DF mixes RT with rasterization. But also has more games with RT. So it will average out to a bigger difference.
 
Last edited:

Gaiff

SBI’s Resident Gaslighter
In rasterization, the 7900 GRE does significantly better than in rasterization, when compared to the 7800XT.
I think that the 7800Xt only has a 128kb vector register file. While the 7900 GRE has 192kb. On RDNA3, this can have a significant impact for RT loads.

HU separates their RT and rasterization tests. And that graph that was posted, is only for rasterization.
DF mixes RT with rasterization. But also has more games with RT. So it will average out to a bigger difference.
That’s the thing though, not talking about averages but individual tests such as the Cyberpunk one. One guys sees a 27% increase compared to the 4070S but the other only sees 2%? There’s something wrong here.

Besides that, Rich was highly positive. Steve from HU was just meh.
 

buenoblue

Member
That’s the thing though, not talking about averages but individual tests such as the Cyberpunk one. One guys sees a 27% increase compared to the 4070S but the other only sees 2%? There’s something wrong here.

Besides that, Rich was highly positive. Steve from HU was just meh.
I mean I've never seen Steve not be Meh to be honest
 

SolidQ

Member
One guys sees a 27% increase compared to the 4070S but the other only sees 2%? There’s something wrong here.
Raster perfomance on RDNA3 in cyberpunk is very strong compared to NV
ad3cdc885ee8a401138d4178f5a4dd46.jpg


Also here from another source. you can see RDNA2 vs RDNA3
1d41845a083bd2dd697a839cee137556.png


Seems something wrong with HuB results
Daniel
db344ff092087928d118cae78ade72ab.jpg
I2Hard
bc438ca1b6b97acc269d1681ec16348e.jpg
 
Last edited:

winjer

Member
That’s the thing though, not talking about averages but individual tests such as the Cyberpunk one. One guys sees a 27% increase compared to the 4070S but the other only sees 2%? There’s something wrong here.

Besides that, Rich was highly positive. Steve from HU was just meh.

With RT in CP2077, HU reports a 10% difference between the 7800Xt and the 7900GRE.
Techpowerup, reports a 16% difference. And Computerbase reports 8%. PCGH reports 10%.
Seems like Rich might have screwed up his tests again.
 

Bojji

Member
With RT in CP2077, HU reports a 10% difference between the 7800Xt and the 7900GRE.
Techpowerup, reports a 16% difference. And Computerbase reports 8%. PCGH reports 10%.
Seems like Rich might have screwed up his tests again.

He is testing with gameplay loop internal CDPR tech demo vs walking in the game or official in game benchmark others are testing.

His test should be more demanding.
 

winjer

Member
This GPU does have the compute power to do well, but it's starved for memory bandwidth. And AMD limiting the memory OC to just 18.5 gbps, from a base of 18 gbps, is a terrible decision.
Maybe someone will be able to mod a bios or something that removes this cap.
 

Gaiff

SBI’s Resident Gaslighter
With RT in CP2077, HU reports a 10% difference between the 7800Xt and the 7900GRE.
Techpowerup, reports a 16% difference. And Computerbase reports 8%. PCGH reports 10%.
Seems like Rich might have screwed up his tests again.
That's without RT.

I'll break down what I saw.

This is HU's results for Phantom Liberty at 1440p/High WITHOUT RT.

CP2077_1440p-p.webp


The 4070S and 7900 GRE are even and the GRE is just 3% faster than the 7800 XT. Rich on the other hand has the 7900 GRE leading the 4070S by 27% and the 7800 XT by 20%.

Daniel Owen's bench at the same settings as Rich (albeit in a different section) has the 7900 GRE 7% ahead of the 7800 XT, not a massive 20%.

b4RCAtP.png

Raster perfomance on RDNA3 in cyberpunk is very strong compared to NV
ad3cdc885ee8a401138d4178f5a4dd46.jpg


Also here from another source. you can see RDNA2 vs RDNA3
1d41845a083bd2dd697a839cee137556.png


Seems something wrong with HuB results
Daniel
db344ff092087928d118cae78ade72ab.jpg
I2Hard
bc438ca1b6b97acc269d1681ec16348e.jpg
These results actually line up with what Rich has but not HU. GN only has those benches in RT so not really useful. I'm talking rasterization.
 
Last edited:

winjer

Member
That's without RT.

I'll break down what I saw.

This is HU's results for Phantom Liberty at 1440p/High WITHOUT RT.

CP2077_1440p-p.webp


The 4070S and 7900 GRE are even and the GRE is just 3% faster than the 7800 XT. Rich on the other hand has the 7900 GRE leading the 4070S by 27% and the 7800 XT by 20%.

Daniel Owen's bench at the same settings as Rich (albeit in a different section) has the 7900 GRE 7% ahead of the 7800 XT, not a massive 20%.

b4RCAtP.png


These results actually line up with what Rich has but not HU. GN only has those benches in RT so not really useful. I'm talking rasterization.

Like Bojji Bojji said, it's probably because of how and where they ran the benchmark.
This happens frequently between different outlets, because most test in different ways, in different places, with different hardware , etc.
 
Last edited:

Gaiff

SBI’s Resident Gaslighter
Like Bojji Bojji said, it's probably because of how and where they ran the benchmark.
This happens frequently between different outlets, because most test in different ways, in different places, with different hardware , etc.
I've seen differences but they are generally pretty understandable or within margin of error. Say, one outlet can get a 5% differential and the other one can get 10-11, or even 15. I've never seen 2% vs 27%. One is irrelevant and the other is basically a different class of performance.
 

winjer

Member
I've seen differences but they are generally pretty understandable or within margin of error. Say, one outlet can get a 5% differential and the other one can get 10-11, or even 15. I've never seen 2% vs 27%. One is irrelevant and the other is basically a different class of performance.

Looking at several reviews, seems like that DF is the one odd out.

I have added the Kitguru and PCWorld reviews.
PCWorld shows similar results to DF. And So does Level1tech. But only compared to the 4070 and 4070S.
 
Last edited:

Bojji

Member

IpHtIKQBwffV6hg4.jpg

Massive difference with OC, most cards get like 5-8%.

Not a bad GPU at all.
 

Crayon

Member
Wow 15% more on that oc. I usually say amd cards are good deals after a year of price drop but this actually has a decent good price right out the gate. I think the 7600 16gb is also an interesting deal. That's $330 and should get you through this console gen and into the next cross gen period.
 
Top Bottom