AMD Vega 10 3DMark Fire Strike Results Surface

[Chaos]

No? Vega 10 is the arch name, it's supposed to be at least on par with 1080ti (again, supposed), why do i have the feeling you got it all wrong?

Back in early 2016, everything pointed towards there being a bigger Vega 11 GPU and smaller Vega 10 GPU. Then some time later Vega 10 became bigger GPU, Vega 11 smaller one. After even more time, Vega 11 completely disappeared and only Vega 10 was talked about.

Also I don't think "Vega 10" is the uarch name. Vega 10 is a code name for GPU (like Hawaii, Tonga, Fiji, Polaris 10/11 etc.).

Unless AMD comes up with the new name for the massively refined GCN that Vega will be based upon I guess name will be GCN5 (Polaris cards being based upon GCN4).

 

[oxidized]

Back in early 2016, everything pointed towards there being a bigger Vega 11 GPU and smaller Vega 10 GPU. Then some time later Vega 10 became bigger GPU, Vega 11 smaller one. After even more time, Vega 11 completely disappeared and only Vega 10 was talked about.

Also I don't think "Vega 10" is the uarch name. Vega 10 is a code name for GPU (like Hawaii, Tonga, Fiji etc.).

Unless AMD comes up with the new name for the massively refined GCN that Vega will be based upon I guess name will be GCN5 (Polaris cards being based upon GCN4).

Vega 11 was just a smaller chip, a bit like 104/106 for nvidia, if vega 10 is like Gx102, vega 11 is like Gx104/106 and newer polaris is either the 104 or the 106, it's a really weird thing, but Vega 10 was always supposed to be the biggest one, which can be cut in different chips too, like GP102 was, with 1080ti and Titans, so i guess that if AMD wants to launch 3 different SKUs they'll all be based on Vega 10, where the fastest chip is the full chip, or almost full, and the others are just cuts. Anyway we'll see

 

[bug]

Vega will come in several different SKUs from high end to low end. It looks to me like the info found is from a low end Vega and not the Vega shown earlier in the year.

It looks to me like wishful thinking. We all want fast cards, because we all want to game at 4k.
But the simple fact is we have no idea which Vega was benched here (with the only lead hinting at the fastest model, but let's disregard that for one more week).

 

[NBH]

I think AMD are going to have a hard time if 2 of the 3 Vega cards are competing with the 1070 and 1080. Those cards are nearly 1 year old and I'm sure Nvidia will have no issues reducing the price of them to compete with Vega should it be released at a cheaper price. After an extra year, Vega has to have better performance or better efficiency/power usage otherwise it is just technology which has already been around for a year but with Freesync compatibility.

The card that could compete with the 1080Ti is more interesting though as Nvidia won't want to budge on price as it is so new and any competition here is much more current.

 

[aj28]

I've seen a couple of sites refer to the MI25 as passively-cooled when talking about Vega rumors, and that's fine because it's technically true, but it's important to realize that all major components in a rack server are passively cooled, including the CPUs. This is because they use shrouds to channel air from high-speed intake fans directly over those components. So, the implication that MI25 could perform as a passively-cooled card in the capacity that we generally think of a graphics card, installed in an ATX workstation chassis, is a bit misleading.

That said, I certainly hope to see Vega succeed and provide much-needed competition to nVidia on the enthusiast end of the market. Unless MI25 is an extremely highly-binned part, I'd expect the flagship consumer graphics chip to achieve similar performance, which I believe correlates to a core clock closer to 1,500 MHz.

 

[Vayra86]

Im pretty sure if you dont have a cpu pushing that gpu then the points are lower. However given im wrong i still seem to find ost 1070s in the range of 18000 points unless they are in the range of 2100mhz core clock

Almost every GP104 SKU clocks 2050 or better, and the 1070 generally gets over 2100.

I had one that did 2176 ingame and stabilized at 2149mhz @ 70 C.

Another thing to realize: we have the (ever so slightly, but still, 4-5% is more than 0) faster 11GBps 1080's now too.

 

[medi01]

Almost every GP104 SKU clocks 2050 or bette

And elephants can fly.
But we are discussing Vega here.

 

[Vayra86]

And elephants can fly.
But we are discussing Vega here.

Then read the whole thread and specifically what I was replying to before you post something that doesn't discuss Vega yourself.

 

[bug]

And elephants can fly.
But we are discussing Vega here.

Funny, I thought we were speculating.

 

[medi01]

1.6GHz Clock could archive amazing performance.

Fury X runs at 1050, 1600 is a 52% higher clock.
Let's say they got 10% improvement from architecture, that would be 68% improvement over Fury X.

For reference:
1080 is 37% faster than Fury (1080p)
1080Ti is 68% faster than Fury (1080p)
https://www.techpowerup.com/reviews/MSI/GTX_1080_Ti_Gaming_X/30.html

 

[efikkan]

Another day, another set of Vega results see the light of it.
…
This is known to be Vega because the device ID, 687F:C1, was spotted on AMD's own hands while running that Doom demo in 4K. The device clocks seem to be in line with previous leaks: a 1200 MHz core clock and 8GB of video memory running at 700 MHz memory clocks…

Another day, another article about the same source… Isn't this the third time you've covered the same "leak"? I'm aware of the information vacuum, but is it necessary to recycle the same thing?

Still, we know 687F:C1 to be the device ID of the first engineering sample of Vega 10, demonstrated multiple times by AMD in December-January. If AMD continue their device naming scheme, we can expect the final Vega 10 products to be named 687F.1, 687F.2, 687F.3, …

Back in early 2016, everything pointed towards there being a bigger Vega 11 GPU and smaller Vega 10 GPU. Then some time later Vega 10 became bigger GPU, Vega 11 smaller one. After even more time, Vega 11 completely disappeared and only Vega 10 was talked about.

No, that's just a misquotation. Vega 10 is the large one.

 

[Vayra86]

Fury X runs at 1050, 1600 is a 52% higher clock.
Let's say they got 10% improvement from architecture, that would be 68% improvement over Fury X.

For reference:
1080 is 37% faster than Fury (1080p)
1080Ti is 68% faster than Fury (1080p)
https://www.techpowerup.com/reviews/MSI/GTX_1080_Ti_Gaming_X/30.html

Ideally, they would get the top SKU a bit higher than the 1080ti, seeing as they are releasing very late this time around and AMD really needs to capture that performance crown for the brand image, even if only just for a couple of months.

One can dream...

 

[ADHDGAMING]

I seriously doubt these numbers are anywhere near final IF even the Big Vega.

 

[springs113]

Another day, another article about the same source… Isn't this the third time you've covered the same "leak"? I'm aware of the information vacuum, but is it necessary to recycle the same thing?

Still, we know 687F:C1 to be the device ID of the first engineering sample of Vega 10, demonstrated multiple times by AMD in December-January. If AMD continue their device naming scheme, we can expect the final Vega 10 products to be named 687F.1, 687F.2, 687F.3, …


No, that's just a misquotation. Vega 10 is the large one.

I seem to remember only 1 YouTuber ever stating what Vega 11 was after new horizon and he pointed it out to be bigger than 10. Hint he's not a nvidia guy. But anyways I'm pretty sure we'll all be happy in less than a months time or so.

 

[TheGuruStud]

The internet has been falling for fake benchmarks since the beginning of benchmarks.

 

[PerfectWave]

"For reference:
1080 is 37% faster than Fury (1080p)
1080Ti is 68% faster than Fury (1080p)"

seriously we all know that furyx was not for 1080p at least comapre them at qwhd or 4k LOL

 

[efikkan]

The internet has been falling for fake benchmarks since the beginning of benchmarks.

As you can see e.g. here, there clearly exist benchmarks of Vega 10. If you add it to a comparison it will reveal the device name. But these are benchmarks of engineering samples, so it's not representative of the final product.

 

[PerfectWave]

https://videocardz.com/69277/amd-radeon-rx-vega-rumors-part-1

 

[TheGuruStud]

As you can see e.g. here, there clearly exist benchmarks of Vega 10. If you add it to a comparison it will reveal the device name. But these are benchmarks of engineering samples, so it's not representative of the final product.

You can spoof that crap. And it just so happens that the scores are what a Fury X would get...

Trolololol in video card land.

 

[Dave65]

Trick of the eye, what are we generally looking at here..Grain of salt taken

 

[bug]

"For reference:
1080 is 37% faster than Fury (1080p)
1080Ti is 68% faster than Fury (1080p)"

seriously we all know that furyx was not for 1080p at least comapre them at qwhd or 4k LOL

Apart from AMD's own marketing team*, nobody believes Fury was built for 4k. Because it can't actually do 4k@60fps without lowering settings.

*and they're paid to believe it

 

[TheGuruStud]

"For reference:
1080 is 37% faster than Fury (1080p)
1080Ti is 68% faster than Fury (1080p)"

seriously we all know that furyx was not for 1080p at least comapre them at qwhd or 4k LOL

Apart from AMD's own marketing team*, nobody believes Fury was built for 4k. Because it can't actually do 4k@60fps without lowering settings.

*and they're paid to believe it

I thought 1080ti was only 17% faster than vanilla at 1080P... The real gains were at 4K (showcasing the terrible scaling of pascal).

 

[Vayra86]

I thought 1080ti was only 17% faster than vanilla at 1080P... The real gains were at 4K (showcasing the terrible scaling of pascal).

Don't think its terrible scaling so much, rather its 1080ti running into CPU bottlenecking at 1080p. Even in synthetics you can see this.

 

[TheGuruStud]

Don't think its terrible scaling so much, rather its 1080ti running into CPU bottlenecking at 1080p. Even in synthetics you can see this.

Driver bottleneck if so.

 

[Vayra86]

Driver bottleneck if so.

Possibly, I remember the same being said about Fury's 1080p perf compared to its scaling at 4K. You could just as easily flip the argument and say it scales well at high res

Regardless, it has little to do with architecture and everything with pushing over 120-150fps.

Just to point that about synthetics out for you; here's the CPU footprint of 3DMark Firestrike - in the pure Graphics tests you can see one core at 90+% already; the first burst of max load you see is physics so duh, but the second one is Combined, which makes up part of the graphics score. I saw 75% GPU usage on a 1080. In addition, relatively, I see 25k graphics scores compared to the 21k of a 1070 OC'd. That's not the +30% I see in ingame FPS...