RX Vega

[AMDPRO]

I am starting this thread to gather as much information as is possible about the upcoming RX vega release.  I hope you guys are as hyped for it as I am!

[Prokart2000]

5 minutes ago, AMDPRO said:

I am starting this thread to gather as much information as is possible about the upcoming RX vega release.  I hope you guys are as hyped for it as I am!

If its good i am going to upgrade to it

[XboxCarsForza]

It's kind of disappointing considering how good ryzen is!

[AMDPRO]

From the thing we already know, its supposedly will!

[AMDPRO]

What is disappointing about it?

Just now, XboxCarsForza said:

It's kind of disappointing considering how good ryzen is!

 

[App4that]

Yep, dat hyped.

 

[DocSwag]

13 minutes ago, AMDPRO said:

I am starting this thread to gather as much information as is possible about the upcoming RX vega release.  I hope you guys are as hyped for it as I am!

You know that those two pics were faked by WCCF for April Fools, right?

[Hunter-97-G]

1600 Mhz with 4096 SM's when the 480's/580's having 2308 struggle a ton to reach 1500 through overclocking? No way.

[DocSwag]

1 minute ago, Hunter-97-G said:

1600 Mhz with 4096 SM's when the 480's/580's having 2308 struggle a ton to reach 1500 through overclocking? No way.

It's not true, read my post:

10 minutes ago, DocSwag said:

You know that those two pics were faked by WCCF for April Fools, right?

 

[othertomperson]

Wooo HBM because memory bandwidth is always what limits my framerates... has anyone ever been limited by vram bandwidth? That didn't even matter on a 970.

[Hunter-97-G]

Just now, DocSwag said:

It's not true, read my post:

 

I know it's not true, but it's always fun to talk about it. My optimistic prediction is 1400Mhz, which would make for 11.4 tflops. That would make it a significant improvement over the 1080 (15% or so), though it would lag like another 15% behind the Ti. The top model that is, which I guess will be sold for 550-600. A cut down VEGA for around 400-450 would be great to see, too.

[DocSwag]

Just now, Hunter-97-G said:

I know it's not true, but it's always fun to talk about it. My optimistic prediction is 1400Mhz, which would make for 11.4 tflops. That would make it a significant improvement over the 1080 (15% or so), though it would lag like another 15% behind the Ti. The top model that is, which I guess will be sold for 550-600. A cut down VEGA for around 400-450 would be great to see, too.

I agree, most likely the flagship Vega will be between the 1080 and 1080 ti, with the cut down version being around the 1080 ish.

[othertomperson]

1 minute ago, Hunter-97-G said:

I know it's not true, but it's always fun to talk about it. My optimistic prediction is 1400Mhz, which would make for 11.4 tflops. That would make it a significant improvement over the 1080 (15% or so), though it would lag like another 15% behind the Ti. The top model that is, which I guess will be sold for 550-600. A cut down VEGA for around 400-450 would be great to see, too.

This is the* problem with teraflops. More cores and lower frequency is much less likely to be utilised than fewer cores at higher frequency, while being apparently equivalent.

 

...ok it's not the problem with teraflops. The real problem is that as a measure it's, uh, bollocks.

[DocSwag]

Just now, othertomperson said:

This is the problem with teraflops. More cores and lower frequency is much less likely to be utilised than fewer cores at higher frequency, while being apparently equivalent.

Not to mention that there are other things that affect performance per core as well.

[Hunter-97-G]

Just now, DocSwag said:

I agree, most likely the flagship Vega will be between the 1080 and 1080 ti, with the cut down version being around the 1080 ish.

Looks like a R9 Fury- Fury X situation all over again.

[Hunter-97-G]

1 minute ago, othertomperson said:

This is the problem with teraflops. More cores and lower frequency is much less likely to be utilised than fewer cores at higher frequency, while being apparently equivalent.

Yes, the architecture of AMD even with the advatage of 14nm FinFET is still a bit less efficient than what Nvidia has on Pascal, which is pretty much Maxwell on 16nm with more CUDAs and dope clockspeeds.

[DocSwag]

Just now, Hunter-97-G said:

Yes, the architecture of AMD even with the advatage of 14nm FinFET is still a bit less efficient than what Nvidia has on Pascal, which is pretty much Maxwell on 16nm with more CUDAs and dope clockspeeds.

Honestly I think 14nm Samsung/GloFo is worse than 16nm TSMC, as evidenced by the lower clock speeds of the 1050 and 1050 ti, though Nvidia is still a little ahead of AMD when it comes to power efficiency, as evidenced by the 1050 vs 460.

[Hunter-97-G]

7 minutes ago, DocSwag said:

Honestly I think 14nm Samsung/GloFo is worse than 16nm TSMC, as evidenced by the lower clock speeds of the 1050 and 1050 ti, though Nvidia is still a little ahead of AMD when it comes to power efficiency, as evidenced by the 1050 vs 460.

 

We'll see about VEGA's efficiency, if the top card is between the 1080 and the Ti 220W would make sense, and like 200 for the cut down version, but who knows, probably they'll run 250W or even 275 like the 390 series and the Furies.

[AMDPRO]

5 hours ago, Hunter-97-G said:

1600 Mhz with 4096 SM's when the 480's/580's having 2308 struggle a ton to reach 1500 through overclocking? No way.

It's a different gpu architecture so it's possible

[AMDPRO]

5 hours ago, DocSwag said:

You know that those two pics were faked by WCCF for April Fools, right?

I dought that as they have released trailers in china that appear very similar and those came after April 1st

[DocSwag]

17 minutes ago, AMDPRO said:

I dought that as they have released trailers in china that appear very similar and those came after April 1st

That wouldn't make sense that they appear IN THE EXACT SAME FORM afterwards and they were legit. It's impossible for wccf to have predicted EXACTLY what the slides would've looked like, every pixel accurately.

 

That's not to mention that the Rage X2 Fury has the pcie slot out of place lol.

 

Plus the card in the picture has no fan or vents, which makes even less sense.

 

These pics weren't shown before april 1st, so if these exact pics were shown afterwards they were copies of these which are FAKE.

[DocSwag]

22 minutes ago, AMDPRO said:

It's a different gpu architecture so it's possible

GCN just doesn't clock that high. That's just how it is.

 

Even Pascal doesn't have 1600 MHz stock on Samsung 14nm, but rather 1400 or so MHz. It's not gonna happen.

[AMDPRO]

10 hours ago, DocSwag said:

GCN just doesn't clock that high. That's just how it is.

 

Even Pascal doesn't have 1600 MHz stock on Samsung 14nm, but rather 1400 or so MHz. It's not gonna happen.

The 1080ti has a base clock at 1569 which makes 1600 pheasable 

[DocSwag]

12 minutes ago, AMDPRO said:

The 1080ti has a base clock at 1569 which makes 1600 pheasable 

And it's on TSMC 16nm.

 

There's a reason why the 1050 and 1050 ti don't clock as high as a 1080 ti. They're made at Samsung instead of TSMC.

[roylapoutre]

It's fun to see all those speculations. Everyone knows Vega is coming, but AMD keeps doing shit with their marketing. What did we have so far ? "Poor Volta" in the first trailer ? The naming "RX Vega" ? Tee-shirts ? And Doom at 4K with a tad more fps than a 1080 with Vulkan ? Come on... I'd be itching to get a 1080/1080 Ti if I were in the market for a GPU right now, AMD is teasing Vega for too long with too little useful specs. 

 

But yeah, 1600MHz seems hard to reach, but who knows, maybe with all these secrets AMD is keeping, we'll have something completely breathtaking ! Titan XpPpP perf for 1070 price ! WOOOOOOOOOOOOOW