Vega/Navi Rumors (Updated)

[PhonakV30]

but man 1600 Mhz ?! I want it ......

 

[CatMerc]

Fooled me for the first few slides while reading the article.

In my defense, I was half reading, half playing Mass Effect Andromeda, and 100% sleep deprived (it's 4AM!).

 

[Arachnotronic]

Fake or real ? There is no stopping of the Hype Train now.

Even if this weren't an obvious April Fools prank, there are a lot of other clues.

For one thing: "AOTS" is not stylized properly -- they'd probably write out "Ashes of the Singularity" but if they didn't, it'd be AoTS, not AOTS. Sniper 4 is also problematic -- isn't it Sniper Elite 4? Then of course you have Deus Ex -- they wouldn't leave out characters indicating that it is Deus Ex: Mankind Divided.

Further, in the "thanks to High Bandwidth Cache" bit, it is unlikely that a professional marketing team would use the phrase "thanks to," and even if it did, "thanks" would have been capitalized.

And, finally, we now that the GPUs won't be called "RAGE," we know it's Radeon RX Vega, so that should have killed its credibility right there.

This is actually a really poorly done April Fool's joke.

 

[tential]

Even if this weren't an obvious April Fools prank, there are a lot of other clues.

For one thing: "AOTS" is not stylized properly -- they'd probably write out "Ashes of the Singularity" but if they didn't, it'd be AoTS, not AOTS. Sniper 4 is also problematic -- isn't it Sniper Elite 4? Then of course you have Deus Ex -- they wouldn't leave out characters indicating that it is Deus Ex: Mankind Divided.

Further, in the "thanks to High Bandwidth Cache" bit, it is unlikely that a professional marketing team would use the phrase "thanks to," and even if it did, "thanks" would have been capitalized.

And, finally, we now that the GPUs won't be called "RAGE," we know it's Radeon RX Vega, so that should have killed its credibility right there.

This is actually a really poorly done April Fool's joke.

I'm happy April 1st is on a Saturday. I can just game all day and turn off forums. May watch some Dota 2 tournaments or something.
Whole internet is annoying on April 1st.

 

[JDG1980]

NDA lift on a Saturday? Bullshit.

 

[zinfamous]

NDA lift on a Saturday? Bullshit.

NDA lift would have been known at least 1 week in advance anyway.

...OK, let's just assume this is all fake and be happy that it's fake.

let's.....

 

[Bacon1]

Looks legit

 

[lobz]

Even if this weren't an obvious April Fools prank, there are a lot of other clues.

For one thing: "AOTS" is not stylized properly -- they'd probably write out "Ashes of the Singularity" but if they didn't, it'd be AoTS, not AOTS. Sniper 4 is also problematic -- isn't it Sniper Elite 4? Then of course you have Deus Ex -- they wouldn't leave out characters indicating that it is Deus Ex: Mankind Divided.

Further, in the "thanks to High Bandwidth Cache" bit, it is unlikely that a professional marketing team would use the phrase "thanks to," and even if it did, "thanks" would have been capitalized.

And, finally, we now that the GPUs won't be called "RAGE," we know it's Radeon RX Vega, so that should have killed its credibility right there.

This is actually a really poorly done April Fool's joke.

you must kill at parties

 

[Mopetar]

but man 1600 Mhz ?! I want it ......

Wasn't the thinking that it would have to hit around 1.5 GHz based on the TFlop numbers AMD gave?

That was assuming it kept the same ~4K shaders as the Fury though.

 

[richierich1212]

NDA lift on a Saturday? Bullshit.

Yeah that's how I had already figured out it was BS from the start haha.

 

[96Firebird]

Looks like a lot of people didn't read the endnotes...

 

[zinfamous]

Looks like a lot of people didn't read the endnotes...

hey, at least they sent us out with some Grateful Dead tunes (via FotM>Scarlet lyrics)


Yeah, someone was effing high.

 

[Magee_MC]

hey, at least they sent us out with some Grateful Dead tunes (via FotM>Scarlet lyrics)


Yeah, someone was effing high.

I think that you meant Scarlet>Fire.

 

[zinfamous]

I think that you meant Scarlet>Fire.

well, the lyrics in that AF joke actually went Fire>something>Scarlet, which is why I posted that (just scroll to the bottom). Yes, odd...but it actually happened in a few shows too, you know. I guess it always depends on the batch...

 

[Glo.]

https://forum.beyond3d.com/posts/1973875/

I don't know if anyone has spotted this post.

Sebbbi, on Beyond3D forum:

Just wanted to clarify that I meant AMD GCN2 (consoles) vs Nvidia's latest (Maxwell/Pascal). AMD PC GPUs have also improved since GCN2.

Improvements for general performance:
- GCN3 introduced delta color compression. Including ability to sample/load compressed textures without decompress step.
- GCN3 improved geometry tessellation performance
- GCN4 improved geometry performance in general (including fast strips, primitive discard, etc).
- GCN4 improved delta color compression.
- GCN4 added instruction prefetch (reduces pipeline latency, again helps with geom bottleneck).
- GCN4 improved async compute scheduling (GPU side)

GCN5 (Vega) adds these general performance improvements:
- L2 cache includes L2 ROP cache (L1 ROP caches under L2). Don't need to flush caches between pixel shader passes.
- Tiled rasterizer. Reduces overdraw, bandwidth and makes ROPs more efficient in general.
- Improved geometry pipeline (including proper load balancing, up to 2x higher peak throughput)
- General purpose memory paging system

(I didn't list features that don't bring performance improvements without programmer intervention)

All of these improvements mean that GCN5 should run general purpose pixel/vertex shader code much better than GCN2. GCN5 has most of the same tricks that are seen in modern Nvidia GPUs. There are nice compute improvements as well, but they need special programmer support (DPP, SDWA, FP16). We will see the real impact of these improvements when DX12 SM 6.0 becomes available. Doom is already using these features with Vulkan, resulting in nice gains.

 

[Ajay]

Well, this also confirms for me that AMD is still building on GCN. Some were saying earlier that it was a new uarch. This dude has some darn good access to AMD GPU info.

 

[Krteq]

Well, AMD itself talks about Vega as a GCN5 in this presentation - GPUOpen - GDC2017-Advanced-Shader-Programming-On-GCN

So it means GCN ISA is maintained and compatibility shouldn't be an issue.

 

[CatMerc]

Spoiler: Large image

Anyone knows if HBM2 PHY's are bigger than HBM1 PHY's?

This kinda puts into perspective just how much space HBM saves.

 

[CatMerc]

Spoiler: Large image

Took it upon myself to figure it out.
GP102 using HBM2 could be a 430mm^2 die, to put things into perspective.

540mm^2 Vega better be damn good.

 

[Malogeek]

Vega is 540mm?

 

[CatMerc]

Vega is 540mm?

Estimates from Raja holding the die. They range from 500 to 560 mm^2.

 

[piesquared]

Unless the two little dies were Vega and the big die was something else.

 

[Bacon1]

 

[Elfear]

Took it upon myself to figure it out.
GP102 using HBM2 could be a 430mm^2 die, to put things into perspective.

540mm^2 Vega better be damn good.

Thanks!

HBM offers some impressive density improvements over GDDR5.

 

[Malogeek]

Estimates from Raja holding the die. They range from 500 to 560 mm^2.

Jeez, I hope it's within my budget.