Hardware & Technical GPU Upgrade: RX Vega 64 to 5700XT

[Robert Maynard]

Starting point: AMD 1800X; RX Vega 64 (reference, stock clocks); Asus PRIME X370-PRO; 16GB DDR4 3600 CL19 GSkill Sniper X urban Kit of 2; 256GB NVMe

Elite Dangerous:

Unigine Superposition:
RX Vega 64 (why it is reported as a Vega 56 I don't know)

5700XT

World of Tanks Benchmark:
RX Vega 64

5700XT

Forza Horizon 4:
RX Vega 64

5700XT

Summary:

 

[Patrick_68000]

Impatient to see the giant leap.

 

[GuruNot]

WoT Encore doesnt seem as big a jump as I expected, 5700XT is only ~11% faster.

 

[Chris Simon]

WoT Encore doesnt seem as big a jump as I expected, 5700XT is only ~11% faster.

11% faster (In the worst scenario) than Vega at half the price and half the power? I'd call it a win.

 

[Robert Maynard]

11% faster (In the worst scenario) than Vega at half the price and half the power? I'd call it a win.

In my case it's a £389 card replacing a £449 card - I got the RX Vega 64 on day one at MSRP (just).

 

[GuruNot]

11% faster (In the worst scenario) than Vega at half the price and half the power? I'd call it a win.

It is a win, but some benchmarks show the increase as over 20% ..... That is all ... You clearly cant believe everything you read on the Internet

 

[Robert Maynard]

E: D performance is all over the place - sometimes over 130FPS in the station, dropping to less than 30FPS - for no apparent reason. Hovering around 32FPS at Dav's Hope (exact same position and viewpoint from the RX Vega 64 test) rather than 83FPS.

Bug report time.
User error, as you were....

 

[GuruNot]

E: D performance is all over the place - sometimes over 130FPS in the station, dropping to less than 30FPS - for no apparent reason.

Bug report time.

Early GPU drivers ?

 

[Chris Simon]

It is a win, but some benchmarks show the increase as over 20% ..... That is all ... You clearly cant believe everything you read on the Internet

Well, that's a given, of course.

In other news:

Source: https://www.youtube.com/watch?v=D3AoenEBM2E

This is interesting. Standard AMD backet and memory/VRM layout means other coolers fit on the card.
I now wonder whether I should reconsider my decision to wait for the Sapphire to make a proper cooler for the XT and instead buy a reference model now and gut one of my old GPUs to try and mod it or slap a watercooler on it.

E: D performance is all over the place - sometimes over 130FPS in the station, dropping to less than 30FPS - for no apparent reason. Hovering around 32FPS at Dav's Hope (exact same position and viewpoint from the RX Vega 64 test) rather than 83FPS.

Bug report time.

Gamer's Nexus and Hardware unboxed did complain that the drivers suck (as usual at launch) so I's just wait a couple of days till they fix it.
Overclocking doesn't work either, according to what they say.

In my case it's a £379 card replacing a £449 card - I got the RX Vega 64 on day one at MSRP (just).

Well, I bought my Vega for almost 600 quid.

 

[Robert Maynard]

Scratch that last - I R numpty. Things don't work too well when you leave Forza running in the background when running other benchmarks.

.... more testing to be done (and I think that all but the Forza benchmark on the 5700XT are complete pants due to the aforementioned "Forza running in the background" issue).

 

[Chris Simon]

Ahahah.

Well, technically, Forza running in the background should not affect GPU scores (too much) as it only takes up RAM and some CPU power but isn't rendered.
Still, though... Professional move! Good job.

 

[Robert Maynard]

Ahahah.

Well, technically, Forza running in the background should not affect GPU scores (too much) as it only takes up RAM and some CPU power but isn't rendered.
Still, though... Professional move! Good job.

I did start to wonder what was going on when the GPU utilisation was 98% with no (apparent) benchmark running - it seems that Forza might render the image behind the benchmark output at full power even when in the background. Temps were a bit toasty too.

 

[Robert Maynard]

Testing finished for this evening. Elite Dangerous testing complete; summary table added.

 

[Chris Simon]

So it looks like that with my new CPU and 5700XT (when it happens), I'm probably looking at at least 30-40% improvement overall.
Very cool.

Thank you for your time to do this, Robert. (I know you did it for yourself, too, but still. )

 

[Robert Maynard]

Thank you for your time to do this, Robert. (I know you did it for yourself, too, but still. )

Part II will add the difference in CPU to the mix (but probably only tested on the 5700XT).

 

[Bob McBobblehead]

I'm waiting until August to get the Saphire Radeon RX 5700 XT Nitro+

 

[Chris Simon]

I'm waiting until August to get the Saphire Radeon RX 5700 XT Nitro+

Ha! You read my mind!
I actually have my recent (before Vega) RX580 Nitro+ special edition (blue) displayed on my shelf, because it's the most beautiful GC I've ever owned.
If they do a special edition of the 5700XT, I'm gonna bleed money.

 

[znôrt]

so if i get this right the 5700xt would be roughly equivalent to the 1080ti at half the price? that would be good news ...

 

[Bob McBobblehead]

Part II will add the difference in CPU to the mix (but probably only tested on the 5700XT).

Do you have a VR headset? If so how well does the 5700 XT perform in VR?

 

[Robert Maynard]

so if i get this right the 5700xt would be roughly equivalent to the 1080ti at half the price? that would be good news ...

This review might give a better basis for comparison:

Source: https://www.youtube.com/watch?v=rz47WqRDDK4

Do you have a VR headset? If so how well does the 5700 XT perform in VR?

I don't, sorry.