AMD VEGA confirmed for 2017 H1

[Besty]

So what do we reckon... £350 for 1070 perf, or £350 for 1080 perf?

Don't want to get too optimistic after the 480 and 580 cost upwards of £250 (offering the same perf/£ as the 290/390 had for years...)

1070 perf for £350 would be massively underwhelming tho, since it doesn't raise the bar in any way.

The card that has been announced is a $1000 card excluding tax. It will be as quick in 4K as a stock 1080ti in Some scenarios. It won't be quicker outside of that and may trade blows with a custom 1080.

 

[N19h7m4r3]

The card that has been announced is a $1000 card excluding tax. It will be as quick in 4K as a stock 1080ti in Some scenarios. It won't be quicker outside of that and may trade blows with a custom 1080.

I didn't know a custom 1080 was around the performance of a 1080ti reference card.

Thought there was nearly a 30% performance difference between the two reference cards.

 

[Howling]

The card that has been announced is a $1000 card excluding tax. It will be as quick in 4K as a stock 1080ti in Some scenarios. It won't be quicker outside of that and may trade blows with a custom 1080.

Sauce?

 

[pandatm]

Estimated based on the Sniper Elite 4 demo. I would wait for benchmarks before jumping to any conclusion. Besides, the Frontier Edition is optimized towards compute (as per what AMD says).

 

[spoffle]

I didn't know a custom 1080 was around the performance of a 1080ti reference card.

Thought there was nearly a 30% performance different between the two reference cards.

30% isn't much. A decent overclock on a 1080 could see it getting into stock 1080Ti performance, situation dependent I imagine.

 

[Erik13]

This a dejavu, seriously one would think that they should have learned few lessons from Fury shortage but no it is Amd we r talking about. I guess they really need to hit the bottom in order to take certain actions

 

[eddyr]

I'm not sure what else could be done. They need to launch Vega this quarter to keep to their schedule. Releasing it in a higher margin, lower volume professional variant first is sensible.

 

[akarypid]

Not according to Raja (https://youtu.be/FwcUMZLvjYw?t=8m42s). This HBCC is a hardware solution, devs won't see any difference. It will work with any application, not just games using DX whatever. They demoed this with render applications beside the Deus Ex and Rotr demos.

There is no safe data to be moved outside the cache or data to not be moved. It's all determined based on how much it is accessed. This is a personal guess, as AMD did not actually specify how the fine grain managing works. Can't say if they ever will...

I really wish that this is the entire truth and that we can now enjoy full performance even with less memory but I will remain skeptical until I see it in action.

He really does say that "it'll just work as is" but I suspect that means "if the developer doesn't do it, we will do it for them, analyse the game and add a profile in our drivers".

...which is perfectly ok since they've been doing an amazing job at that recently.

Me, I'd still rather go for the 8GB Vega than the 4GB one. And if it turns out to work as advertised, I will be happy I got a card that can effectively handle up to 16GB.

 

[Steampunk]

The card that has been announced is a $1000 card excluding tax. It will be as quick in 4K as a stock 1080ti in Some scenarios. It won't be quicker outside of that and may trade blows with a custom 1080.

Frontier is a professional workstation card. It's not a gaming/consumer card. Sure, you can game on Frontier, just like you could take a Challenger tank to do your local shopping, but that's not really what it's for.

 

[Gregster]

30% isn't much. A decent overclock on a 1080 could see it getting into stock 1080Ti performance, situation dependent I imagine.

Thought I would test this and ran Timespy with my CPU and 1080Ti at stock and not even close in this test.

My GPU scored 10013 for graphics and the highest 1080 scored 8355 in our own Timespy thread that Kaap runs.

 

[N19h7m4r3]

30% isn't much. A decent overclock on a 1080 could see it getting into stock 1080Ti performance, situation dependent I imagine.

30% is rather huge. That's the difference between 1080 and Titan XP. I've never heard of someone bridging that gap for gaming performance.

Besides if Vega suddenly performed with a variance of over 30% depending on games that would be extremely odd.

Imagine if you had GTX 1080 and in some cases it's slower than 980Ti. That's odd.

Edit: Thanks for testing mate. That's a pretty substantial gap still.

Thought I would test this and ran Timespy with my CPU and 1080Ti at stock and not even close in this test.

My GPU scored 10013 for graphics and the highest 1080 scored 8355 in our own Timespy thread that Kaap runs.

 

[nashathedog]

It is to do with the way a integrated circuit design is transferred to the silicon - dies are made using a process not a million miles removed from old film photography - grossly simplified DUV and EUV refer to the light used to expose the design involved in this.

Grossly simplified I can do

I get it now, Thanks.

 

[nashathedog]

Yes, by all accounts GloFo is offering 7nm before the end of 2018. And it was designed by IBM originally, so people have high hopes it'll be good. AMD have also mentioned 48-core Zen-2 arch Server chips on 7nm in 2018/2019.


EDIT: This Anandtech article is good, and not really out of date http://www.anandtech.com/show/10704/globalfoundries-updates-roadmap-7-nm-in-2h-2018

Basically we're expecting 7nm from both GloFo and TSMC to start production 1H 2018 ish, and ship before the end of the year (so mid-Navi, and Volta 1080Ti). Then 7nm+ from GloFo in 2019 sometime (Navi Fury?). Then 5nm from TSMC in 2020, with GloFo also offering 5nm around then (Navi and Volta successors from both companies).

So after the 28nm stagnation, we're actually back on track with steroids by the looks of things.

Summary of edit:

Mid-2017 to Early-2018 14nm/12nm: Vega Fury, and mid-Volta (GTX 1080 size)

Mid-2018 to Late-2018 7nm: mid-Navi (RX480 to GTX 1080 size), and MAYBE large Volta (if not then Early-2019)

Early-2019 to Mid-2019 7nm+: Navi Fury, MAYBE later large Volta

Mid-2020 to Mid-2021 5nm: mid-sized successors to Navi and Volta from both companies. And we'll probably be complaining "where are the 8K capable cards" around now

Thankyou, I'll go and have a good read now.

 

[alec]

This a dejavu, seriously one would think that they should have learned few lessons from Fury shortage but no it is Amd we r talking about. I guess they really need to hit the bottom in order to take certain actions

Tricked by SK Hynix promises

 

[spoffle]

I meant to say territory. The boosting feature of nVidia cards really does make overclocked versus stock comparisons very difficult. But don't 1080s overclock quite a bit? I had the impression that they tend to overclock 30%>. My luck with GPUs that overclock well might be skewing that as I've had a decent amount of cards that I was able to get a 50% overclock on.

The essence of what I was getting at though is that 30% isn't that big of a difference that it's going to be night and day in games. It's really the difference between just about playable, and playable with a bit in reserve for dips.

 

[nashathedog]

I meant to say territory. The boosting feature of nVidia cards really does make overclocked versus stock comparisons very difficult. But don't 1080s overclock quite a bit? I had the impression that they tend to overclock 30%>. My luck with GPUs that overclock well might be skewing that as I've had a decent amount of cards that I was able to get a 50% overclock on.

The essence of what I was getting at though is that 30% isn't that big of a difference that it's going to be night and day in games. It's really the difference between just about playable, and playable with a bit in reserve for dips.

50% wow, which cards?

 

[Gregster]

30% is rather huge. That's the difference between 1080 and Titan XP. I've never heard of someone bridging that gap for gaming performance.

Besides if Vega suddenly performed with a variance of over 30% depending on games that would be extremely odd.

Imagine if you had GTX 1080 and in some cases it's slower than 980Ti. That's odd.

Edit: Thanks for testing mate. That's a pretty substantial gap still.

No probs and my own curiosity got the better of me

 

[spoffle]

I can't remember them all, but I had a good few 7950s that would overclock more than 50%.

I haven't had the need to overclock for quite a while. Due to a change in circumstances (bought a house and I haven't got a dedicated computer room set up with a desk and all that, yet...), I do all my gaming on my TV with a lapdog, and I tend to really only play PVZ Garden Warfare 2, which is a fairly light game. So I have no need to overclock until I finish said computer room and start playing more games.

 

[Gregster]

I meant to say territory. The boosting feature of nVidia cards really does make overclocked versus stock comparisons very difficult. But don't 1080s overclock quite a bit? I had the impression that they tend to overclock 30%>. My luck with GPUs that overclock well might be skewing that as I've had a decent amount of cards that I was able to get a 50% overclock on.

The essence of what I was getting at though is that 30% isn't that big of a difference that it's going to be night and day in games. It's really the difference between just about playable, and playable with a bit in reserve for dips.

Depends on what way you look at it. That 30% for me was well worth it and the 1080 was great for 1440P and allowed me to run with ultra settings in pretty much everything but switching up to 3440x1440 had it struggle and noticeable slow downs in a couple of my more demanding games and with the same settings but then switching the 1080 out for a 1080Ti was night and day and allowed me to not even consider having to lower settings.

 

[aoaaron]

If you're not overclocking, is there much point in having faith in AMD at this point?

If they're just going to release a card which has 1080 level performance at a 1080 price point this LATE in the game, I think I'd rather take that initial hit on the NVIDIA tax in regards to gsync and then have access to high end cards which push current levels of performance without having to wait a year for AMD to play catchup.