• 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.
At this point I expect Vega to be a smidge under the performance of a 1070 tbh.

Na I think it will definitely give the 1080 a run for its money (if the leaked specs are true). I think it will struggle to match or beat the Pascal Titan though.
 
Ha, got a sneaky feeling the full fat Vega will smoke the Titan XP by a fair amount (;

I am putting some pennies aside...

This is what I am saying. At worst it will match it.

But I think people are confused because they are looking at Polaris and using that as comparison. Lol.


Na I think it will definitely give the 1080 a run for its money (if the leaked specs are true). I think it will struggle to match or beat the Pascal Titan though.

Waste of time looking at those tbh. So many lies for click bait about the 480 if you recall not so long ago.
 
At stock, the 1080 is delivering in the region of 180% performance of the RX480 with 135% the diesize, ~140% the clockspeed and ~100% power. (Techpowerups RX480 launch day review)

So approx 33% more performance per mm2, for which they clearly rely being able to achieve higher clockspeeds with their arch/physical design to get the lead they do.

However this is slightly skewed by the fact that being a smaller die relative to the GP104, Polaris 10 will have a greater proportion of its die dedicated to hardware for multimedia/display/pcie and memory interface. Not to forget the often touted hardware scheduling, although you could argue it is integral to AMD's performance in their current design so should remain in the calculation but it would require additional power and the added complexity likely impedes clockspeed. It also doesn't account for future games that may play to AMD's arch or for AMD to workout how to get extract better utilisation. Considering the limited info available that's an argument for another day between those who care enough to argue it!

Nvidia is using GDDR5X though and we don't know how bandwidth limited Polaris 10 is. AFAIK,Nvidia is far more efficient using memory bandwidth due to the way their GPUs function.
 
This is what I am saying. At worst it will match it.

But I think people are confused because they are looking at Polaris and using that as comparison. Lol.




Waste of time looking at those tbh. So many lies for click bait about the 480 if you recall not so long ago.

They are well within the realms of possibility though.
 
They are well within the realms of possibility though.

True, but so were a lot of things they said which turned out to be inaccurate in the past. But I personally do not bother with them much anymore. Edit: By them I mean rumours.

Did Kaaps not have a thread where he would post about rumours to see if they turned out to be true or not, whatever happened to that? :p
 
True, but so were a lot of things they said which turned out to be inaccurate in the past. But I personally do not bother with them much anymore. Edit: By them I mean rumours.

Did Kaaps not have a thread where he would post about rumours to see if they turned out to be true or not, whatever happened to that? :p

Rumour has it it crashed and burned.
 
Ha, got a sneaky feeling the full fat Vega will smoke the Titan XP by a fair amount (;

I am putting some pennies aside...

DX12? Yes I think it will more than likely (pretty embarrassing for AMD if it doesn't), and probably equal-ish on DX11....

I *do* think that they'll charge for it though tbh, at least try. But they've been out of the top-tier game for so long I'm not sure consumers will stomach it! :)
 
True, but so were a lot of things they said which turned out to be inaccurate in the past. But I personally do not bother with them much anymore. Edit: By them I mean rumours.

Did Kaaps not have a thread where he would post about rumours to see if they turned out to be true or not, whatever happened to that? :p

Must get that thread back up and running.

Also need more people to post more rumours in it.:)
 
This is what I am saying. At worst it will match it.

But I think people are confused because they are looking at Polaris and using that as comparison. Lol.


Well we can only go on what we have seen already.

Now with 64 compute units (4096 stream processors) the 24TF of FP16 (half precision) 225W TDP that the latest rumour tell us, they are talking about twice the cores over twice the compute power and 87% more power (all compared to a RX470)

That points to a twice a 470 but a bit better, which would have trouble beating a 1080 and most certainly not be anywhere near a TitanX Pascal.

Now of course, AMD might have a much bigger improvement than that, I would guess that if they had a much bigger improvement than from Fiji to Polaris with Polaris to Vega, then they might be able to pull it off and get close to the TitanX Pascal.

Have they had the research budget to do that?

Only time will tell.
 
Well we can only go on what we have seen already.

Now with 64 compute units (4096 stream processors) the 24TF of FP16 (half precision) 225W TDP that the latest rumour tell us, they are talking about twice the cores over twice the compute power and 87% more power (all compared to a RX470)

That points to a twice a 470 but a bit better, which would have trouble beating a 1080 and most certainly not be anywhere near a TitanX Pascal.

Now of course, AMD might have a much bigger improvement than that, I would guess that if they had a much bigger improvement than from Fiji to Polaris with Polaris to Vega, then they might be able to pull it off and get close to the TitanX Pascal.

Have they had the research budget to do that?

Only time will tell.

Even if you double the cores of the RX470/480 you wont get double the performance due to other diminishing returns and bottle necks.
 
Even if you double the cores of the RX470/480 you wont get double the performance due to other diminishing returns and bottle necks.

Yeah this is true. We can only hope Vega is a genuine upgrade to the architecture and isn't just 'double polaris'.

Fiji suffered massively from bottlenecks, whatever the full technical reason turned out to be. It had 45% more cores than a 290X and certainly was not 45% faster.
 
Yeah this is true. We can only hope Vega is a genuine upgrade to the architecture and isn't just 'double polaris'.

Fiji suffered massively from bottlenecks, whatever the full technical reason turned out to be. It had 45% more cores than a 290X and certainly was not 45% faster.

Which is why it makes me laugh when people start rumouring about 234535345435435 cores on the next iteration of a GPU as more often than not it won't be the case due to architectures running into pipeline latency issues and so on diminishing performance.
 
AMD must have sorted those bottlenecks, and it must be at least as fast as the 1080.

If not there is no reason for the card to exist because it wouldn't be any faster than the Fury-X.

In which case just add another 500 shaders to the RX 480 and sell it for £300.
 
AMD must have sorted those bottlenecks, and it must be at least as fast as the 1080.

If not there is no reason for the card to exist because it wouldn't be any faster than the Fury-X.

In which case just add another 500 shaders to the RX 480 and sell it for £300.

You'd certainly hope so yes.

As mentioned, no one's exactly sure what the bottleneck was on Fiji (or what combination of things). It could be just clockspeed getting up to ~1400+ MHz sorts things out. It could be something much more fundamental.

The only thing is, if the rumoured specs are true then it certainly looks like they're trying again, but this time without the underlying issues?

Let's just hope it's Fury X-mk2, a monster. And not Fury X-mk2 electric boogaloo.
 
You'd certainly hope so yes.

As mentioned, no one's exactly sure what the bottleneck was on Fiji (or what combination of things). It could be just clockspeed getting up to ~1400+ MHz sorts things out. It could be something much more fundamental.

The only thing is, if the rumoured specs are true then it certainly looks like they're trying again, but this time without the underlying issues?

Let's just hope it's Fury X-mk2, a monster. And not Fury X-mk2 electric boogaloo.

If not then its not a very expensive card, and that suits me just fine.
 
Status
Not open for further replies.
Back
Top Bottom