*UPDATE 2: AMD RX VEGA Spotted in TechPowerUp GPU Database (Supposed to launch April 28)

[zMeul]

1 minute ago, Humbug said:

What they have avoided is raw performance. No slides comparing it to Nvidia GPUs in popular games or benchmarks etc. That may be a bad sign.

au contraire mon ami

 

every launch where AMD could not stop talking turned out to have hidden elements

this is why I'm thinking Vega will actually be very disruptive, but ... kinda' late to the party

[Delicieuxz]

1 minute ago, Humbug said:

What they have avoided is raw performance. No slides comparing it to Nvidia GPUs in popular games or benchmarks etc. That may be a bad sign.

Maybe AMD just don't want to invite a cock-blocking maneuver from Nvidia ahead of Vega's launch - although Nvidia's already done that with the GTX 1080 Ti.

[juri-han]

2 hours ago, cj09beira said:

VEGAS clock should be around 1500 so i call it just speculation

I love how people just ignore this and got straight for the benchmark and  immediately post about how "disappointed" they are......

 

lets say that is vega and it's underclocked for some reason, wouldn't  it at stock 1500-1600 clock speed pass the ti?

[alphastrata]

hmmm... fingers crossed!

[Phate.exe]

2 hours ago, Cartoons Plural said:

All that being said this isn't vega. So irrelevant. 

Aren't the Vega compute cards supposed to be 12.5 teraflop?  It wouldn't make a ton of sense for the consumer gaming GPU version to somehow have less performance.  Unless it's small Vega, although even then I'd assume they would want something between RX 580 and GTX 1080 performance (to compete with the 1070).

[ApolloFury]

5 hours ago, YoloSwag said:

Looking at relative performance, it looks like it is just slightly under the GTX 1080 which, IMO, is sad. But the thing is, AMD's high end cards especially the ones with HBM have been performing better in higher resolutions than 1080p, so this could be a mixed relative performance chart. No one should even get current high end cards just for 1080p anyway. However, FPP is under the GTX 1080 as it is as per the database just 9.83 GFLOPS.

It says on that very page: "Radeon RX Vega performance estimated based on architecture, shader count and clocks."

They probably took the performance of the R9 Fury X and extrapolate from it.

[ivan134]

14 minutes ago, ApolloFury said:

It says on that very page: "Radeon RX Vega performance estimated based on architecture, shader count and clocks."

They probably took the performance of the R9 Fury X and extrapolate from it.

The estimate is a pretty dumb one too. This isn't a final indicator of anything, but let's use the RX 470/570 as an example. It has the same number of SPs as the R9 380x, but it's about 47% faster. A 4096 SP GCN 4+ GPU at 1284 MHz with the same IPC improvements would sit between the 1080 and 1080 ti at 1080p. If it clocks up to 1500, it would trade blows with the 1080 ti.

 

This is the chart I used for the math and I assumed the 1050 ti was the same speed as a 380x.

[alamox]

wth kitguru ? the TPU database  info for unreleased products are just kind of place holders, that gets updated when the product is released.

beside do you really believe AMD would release Vega without any event of conference ? dream on, the up coming dates where we might have vega release is computex 28-30th May , or E3 13-15th June.

also there is no way Vega is slower than a 1080, the drawback from is far greater than just releasing a higher end card than the 480, and they had a year knowing the 1080's performance, AMD will target at least 10% over the 1080, and just by the number of CUs, vega should be exactly twice the performance of a 480, which puts it around 20% over the 1080 and about 10-15% under the Ti.

[MoonSpot]

All I want for Christmas is 1440 ultrawide 60fps freesync max details.  Can't say I'd be surprised if the perf on techpowerup is as far as it gets. 

 

Quote

This graphics card is not released yet.
Data on this page may change in the future.

There are those who take the stance that when Pandora opened the jar/box unleashing plagues, pestilence, disease, pain and death upon mankind.  That when she closed the jar, only keeping hope trapped with in it, she did so believing that giving man hope or believing in hope would have been too cruel.

 

Something to think about anyways.

[samcool55]

Sooo if it's the same hour as ryzen, it should be in a few hours.

Well, i guess we'll see

[Prysin]

@Humbug

So if this leak is anything to go by. well....

[YoloSwag]

4 hours ago, ApolloFury said:

It says on that very page: "Radeon RX Vega performance estimated based on architecture, shader count and clocks."

They probably took the performance of the R9 Fury X and extrapolate from it.

DANG! that was not there went I made this post. They just added it lately due to controversy probably.

 

Does anyone here still have a cached sample or a screen of the original one?

 

UPDATE:

I added a cached version of the page which you can view in the original post.

Edited April 29, 2017 by YoloSwag
Updated the post.

[sof006]

Hype! This will be interesting. Hopefully it gives Nvidia a run for their money.

[YoloSwag]

8 minutes ago, sof006 said:

Hype! This will be interesting. Hopefully it gives Nvidia a run for their money.

As others have pointed out the reported clock speeds are low (1200MHz vs 1500MHz from speculation)

But if it was just 5% slow compared to a GTX 1080 with the 1200MHz clock, how well might it perform to a GTX 1080Ti or Titan Xp if it had 1500MHz.

 

I'm hyped and would by this if that were the case as long as they stick to $500. But if they won't, I guess a discounted GTX 1080 or a GTX 1070 will do.

[LukaH]

ryzen leaks were exciting. vega leaks are just sad and on repeat.

[Humbug]

1 hour ago, YoloSwag said:

 

I'm hyped and would by this if that were the case as long as they stick to $500. But if they won't, I guess a discounted GTX 1080 or a GTX 1070 will do.

if it has titan Xp performance as AMD's don woligroski said then it won't be $500.

[Humbug]

7 minutes ago, LukaH said:

ryzen leaks were exciting. vega leaks are just sad and on repeat.

Vega leaks are almost non-existant. Cause the board partners like Sapphire, MSI etc seem to be trust worthy.

 

Ryzen leaks were almost a daily occurrence because AMD had to send out CPU samples for the motherboard guys to test their stuff. And those guys could not resist LOL...

 

[cj09beira]

7 hours ago, juri-han said:

I love how people just ignore this and got straight for the benchmark and  immediately post about how "disappointed" they are......

 

lets say that is vega and it's underclocked for some reason, wouldn't  it at stock 1500-1600 clock speed pass the ti?

it very well could

[cj09beira]

3 hours ago, MoonSpot said:

All I want for Christmas is 1440 ultrawide 60fps freesync max details.  Can't say I'd be surprised if the perf on techpowerup is as far as it gets. 

 

There are those who take the stance that when Pandora opened the jar/box unleashing plagues, pestilence, disease, pain and death upon mankind.  That when she closed the jar, only keeping hope trapped with in it, she did so believing that giving man hope or believing in hope would have been too cruel.

 

Something to think about anyways.

vega is at least 1500 mhz because the Mi25 is 25 tflops of fp16 and to achieve that: 1525mhz * 4 (instructions per clock) * 4096 compute_units = 24.98Tflops (fp16),

from 1200 to 1500 is still 25% increase in clock speeds, knowing how the pro market always has lower clock speeds it could very well be even higher, how much we just don't know

[YoloSwag]

1 hour ago, Humbug said:

if it has titan Xp performance as AMD's don woligroski said then it won't be $500.

But I remember the R9 290X ($550) beating the Maxwell GTX Titan ($1000-$1200) That's why they had to release a GTX 780 Ti. Sure it's not $500 but what is a difference of $50 worth?

 

EDIT: Correction, it's Titan KEPLER not Maxwell.

Edited April 29, 2017 by YoloSwag
More info

[cj09beira]

6 minutes ago, YoloSwag said:

But I remember the R9 290X ($550) beating the Maxwell GTX Titan ($1000-$1200) That's why they had to release a GTX 780 Ti. Sure it's not $500 but what is a difference of $50 worth?

well it should be cheaper to produce than the fury x, because its 596mm^2 vs 520ish mm^2, it only has two hbm2 chips, which also mean that the interposer is also smaller, but it should be more expensive than a 290x @ 438mm^2. so we have a price range of 550-600 dollars 

[YoloSwag]

1 minute ago, cj09beira said:

well it should be cheaper to produce than the fury x, because its 596mm^2 vs 520ish mm^2, it only has two hbm2 chips, which also mean that the interposer is also smaller, but it should be more expensive than a 290x @ 438mm^2. so we have a price range of 550-600 dollars 

This I like. Especially if it really turns out to be a Titan slayer.

 

Out of topic: Are you really a Floatplane Pilot or is it a tittle here in LTT? There seems to be lots of Floatplane Pilots here.

[cj09beira]

Just now, YoloSwag said:

This I like. Especially if it really turns out to be a Titan slayer.

 

Out of topic: Are you really a Floatplane Pilot or is it a tittle here in LTT? There seems to be lots of Floatplane Pilots here.

yes i am 

[Agosto]

Oh come on AMD, still only 64 ROPs for all those stream processors? pls stahp

[Fetzie]

3 hours ago, Prysin said:

@Humbug

So if this leak is anything to go by. well....

It isn't. The clock speeds are too low and they reference the wrong chipset.