• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD VEGA confirmed for 2017 H1

Status
Not open for further replies.
Heat might end up being a consideration for me. My case doesn't have great ventilation and currently my 970 with it's windforce cooler is actually heating up the case. Seems warmer to touch the side of the case with the 1700 AM4 inhere than it was with the 3570k. I've bought a new exhaust fan to something better than stock.

Considerations for new upgrade:

1. FreeSync/Gsync. Next card means next monitor.
2. Heat. If the Vega is much hotter than the 1000 series then that could be important for me as case has poor airflow.
 
TheRealDeal said:
I think people also need to realise these are NCU Shaders and not GCN shaders. What if one NCU shader is faster than one of Nvidia's Cuda cores at there given core clocks. Vega also has more even compared to Titan Xp so it might be a non issue. AMD did says the Shaders were up to twice the through put compared to GCN. With all the other improvements the 4096 might be enough.
Click to expand...

I've definitely seen the 2x NCU figure vs. CU, however it's hard to tell if that is marketing of Vega's half-precision FP (FP16) capability. Sadly, FP16 is not very useful for games without massive optimisation effort on the game/engine.
 
0h_no_not_again said:
I've definitely seen the 2x NCU figure vs. CU, however it's hard to tell if that is marketing of Vega's half-precision FP (FP16) capability. Sadly, FP16 is not very useful for games without massive optimisation effort on the game/engine.
Click to expand...

I just went and read again it's actually the geometry pipeline that's up to 2 x as fast, Anandtech did say the NCU shader is a full on new design with improved IPC so who knows how much they have gained here.
 
Last edited:
TheRealDeal said:
I think people also need to realise these are NCU Shaders and not GCN shaders. What if one NCU shader is faster than one of Nvidia's Cuda cores at there given core clocks. Vega also has more even compared to Titan Xp so it might be a non issue. AMD did says the Shaders were up to twice the through put compared to GCN. With all the other improvements the 4096 might be enough.
Click to expand...
I believe AMD's claims of 2x throughput applied to their new geometry engine, not shaders.

It's looking more like the new compute engine features will probably be more primed for non-gaming aspects, but we'll see. I imagine the expanded functionality of the new compute shaders will require some very specific optimizations to take full use of and that just doesn't seem likely when the vast majority of hardware out there wont be running on Vega.
 
Seanspeed said:
I believe AMD's claims of 2x throughput applied to their new geometry engine, not shaders.

It's looking more like the new compute engine features will probably be more primed for non-gaming aspects, but we'll see. I imagine the expanded functionality of the new compute shaders will require some very specific optimizations to take full use of and that just doesn't seem likely when the vast majority of hardware out there wont be running on Vega.
Click to expand...

Yea it was the Geometry Pipeline (that's what i get for not reading again before typing :D:D). Corrected myself above after another read on Anand. The new Shaders are a full on new Design with Improved IPC but no one knows how much. Hopefully all the improvements in the Vega design lead up to one fast new architecture and card.
 
Last edited:
TheRealDeal said:
Yea it was the Geometry Pipeline (that's what i get for not reading again before typing :D:D). Corrected myself above after another read on Anand. The new Shaders are a full on new Design with Improved IPC but no ones knows how much. Hopefully all the improvements in the Vega design lead up to one fast new architecture and card.
Click to expand...
I just hope AMD isn't doing another 'getting ahead of themselves' sort of innovation where current standards and practices just wont incentivize developers to make best use of Vega's new capabilities just yet.

They need something that performs to a high potential out-the-box. None of this 'it'll come good in 2 years' sort of stuff. Nvidia benefit hugely everytime AMD's cards dont represent well on benches in reviews.
 
Last edited:
Seanspeed said:
I feel like you're deliberately misunderstanding what I'm saying.

I said 'in a given range'. How does Fury X prove me wrong? It's exactly what I'm talking about. 4000 core GPU and the big chip of the lineup. Same with the 7970.

It's not about who makes bigger chips - Nvidia or AMD. Like, you're bringing that comparison up out of nowhere and it's not at all relevant to anything I'm saying. I'm not saying Vega will be 600mm². I'm saying it will be the 'big chip' of AMD's current lineup.
Click to expand...

You said AMD wouldn't or couldn't make a 6000 core GPU because they don't make GPU's as big as nVidia's, i'm showing you hysterically they have and there is no reason why they wouldn't again, ergo a 6000 core AMD card has every chance of existing.

0h_no_not_again said:

Something is up with the Vega NCU's:
All reports are saying that Vega is a 4096 NCU (CU) device, however the 14nm Vega die we have seen is bigger than expected at 475mm^2 to 525mm^2
https://www.reddit.com/r/Amd/comments/5m6uj3/vega_die_size_475mm2_xpost_rhardware/
https://www.computerbase.de/2017-01/amd-vega-preview/

Where did those extra transistors go?!?
Click to expand...

Considering the 28nm 4000 core Fury-X was only 50 to 100mm2^ larger than that i would suggest there are more than 4000 cores in it.

On the same 14nm process the RX 480 2300 core is 230mm2^, not all of the die is taken up by cores.
 
Doom112 said:
Pricing of RX 580 in my country compare to GTX 1070.

RX 580 pricing in my country compare to GTX 1070
SAPPHIRE PULSE RX580 8GB RM1799
SAPPHIRE PULSE RX580 4GB RM1439
SAPPHIRE PULSE RX570 4GB RM1269
SAPPHIRE PULSE RX550 4GB RM569
SAPPHIRE PULSE RX550 2GB RM479

GTX MSI 1070 gaming X 8GB RM 1799.

In shot, it is DOA.
Click to expand...

1070 gaming x. A quick look suggest they are generally RM2229+ (£400+).
 
humbug said:
You said AMD wouldn't or couldn't make a 6000 core GPU because they don't make GPU's as big as nVidia's, i'm showing you hysterically they have and there is no reason why they wouldn't again, ergo a 6000 core AMD card has every chance of existing.

Considering the 28nm 4000 core Fury-X was only 50 to 100mm2^ larger than that i would suggest there are more than 4000 cores in it.

On the same 14nm process the RX 480 2300 core is 230mm2^, not all of the die is taken up by cores.
Click to expand...

It is the leaks, all of which have said 4096 NCUs, that have me bamboozled! Agreed that the compute units are not the only driver for die size, but several Billion transistors is a lot to lose :)

I've gotta say, I think you're right with >4096 NCUs being the Vega die we saw, or there are some really REALLY dramatic changes to the arch.
 
Status
Not open for further replies.
Back
Top Bottom