Poll: ** The AMD VEGA Thread **

[AllBodies]

Im quite excited to see what Navi will be like. 7nm and its the architecture which Raja had most involvement in. Not to mention is using "Next gen memory". Also the potential to make a big GPU out of higher salvageable smaller dies could be quite interesting.

The Next Gen Memory is just GDDR6 FYI...

And likely 14000 MHz on 256-bit for the first Navi cards out in late-2018, which will likely be RX480 sized. So we're talking a ~13 Tflop GPU with 448 GB/s bandwidth.

 

[opethdisciple]

I just hope Vega has something interesting about it.

It needs to have an x factor or it will be a bulldozer in GPU form. I.e a damp squib but the foundations for the next gen.

 

[AllBodies]

I just hope Vega has something interesting about it.

It needs to have an x factor or it will be a bulldozer in GPU form. I.e a damp squib but the foundations for the next gen.

Well it looks like the only saving grace is if analysis is correct that a lot of the new hardware features are currently turned off on FE-Vega. And turning those features on, and utilising them well in games, makes up a 20%+ performance boost.

 

[five8five]

wth a 484mm² die vega shouldn't be too far from the Ti, i for once think AMD might be sandbaging.

Why sandbag though? What have they got to gain? Nvidia will already know the performance of the RX. They're not going to release a competing card but they can reduce prices very quickly.
AMD have had lots of bad publicity from the FE release and they've done almost nothing to counter it - a few twitter / reddit comments, the world tour

Why haven't they released snippets of information, benchmarks etc to generate positive interest. It just strikes me as incredibly naive marketing.
If they are sandbagging it's a joke. They've lost sales to Nvidia that they potentially could've had.

 

[AllBodies]

Why sandbag though? What have they got to gain? Nvidia will already know the performance of the RX. They're not going to release a competing card but they can reduce prices very quickly.
AMD have had lots of bad publicity from the FE release and they've done almost nothing to counter it - a few twitter / reddit comments, the world tour

Why haven't they released snippets of information, benchmarks etc to generate positive interest. It just strikes me as incredibly naive marketing.
If they are sandbagging it's a joke. They've lost sales to Nvidia that they potentially could've had.

Yeah they could repeat the "Titan Hunting" marketing from Kepler-Titan vs R9 290X. No one believed it at the time, but it did indeed turn out faster. And a boat-load faster today, after driver optimisation and Kepler performance falling off a cliff.

Kepler-Titan vs 290X today would be a massacre.

EDIT: Turns out this may be due to reviewers reusing old scores, see post #6352

 

[opethdisciple]

Well it looks like the only saving grace is if analysis is correct that a lot of the new hardware features are currently turned off on FE-Vega. And turning those features on, and utilising them well in games, makes up a 20%+ performance boost.

Lets assume that this comes to fruition. With 20% + perf, where would that leave the FE? Faster than a 1080?

 

[Jono8]

Interesting post on Reddit -

"Reference models are already stockpiling at one company with which I have contacts. They apparently have "low thousands" ready for release. They usually have tens of thousands, though, but this is apparently a very expensive GPU. Supposedly priced above the 1080 at the top (my contact doesn't have pricing info outside of internal rumblings).

The company is a primary supplier, not a reseller."

Could be BS though.

 

[Jono8]

Lets assume that this comes to fruition. With 20% + perf, where would that leave the FE? Faster than a 1080?

Seems the 1080 ref model is about ~10% faster on average across a fair few games in this review -

http://www.gamersnexus.net/hwreviews/2973-amd-vega-frontier-edition-reviewed-too-soon-to-call/page-4

So realistically, even with 20% driver improvements you are probably only looking at a high overclock 1080 performance.

 

[AllBodies]

Seems the 1080 ref model is about ~10% faster on average across a fair few games in this review -

http://www.gamersnexus.net/hwreviews/2973-amd-vega-frontier-edition-reviewed-too-soon-to-call/page-4

So realistically, even with 20% driver improvements you are probably only looking at a high overclock 1080 performance.

Lets assume that this comes to fruition. With 20% + perf, where would that leave the FE? Faster than a 1080?

So the FE-Vega is about the same as a stock 1080. It loses/matches in Nvidia/neutral games, and beats the 1080 in AMD favoured games.

But that's with it jumping around 1440-1528 clockspeed stages. So if you get a solid ~1600-1650 MHz with a proper cooler and optimised voltage, then added 20%, you'd get between max-clocked 1080 at worst, and nipping at the heels of a stock 1080 Ti at best.

 

[r7slayer]

The Next Gen Memory is just GDDR6 FYI...

And likely 14000 MHz on 256-bit for the first Navi cards out in late-2018, which will likely be RX480 sized. So we're talking a ~13 Tflop GPU with 448 GB/s bandwidth.

Ohh well GDDR6 is still a step in the right direction and 448 at a mid ranged card is nowt to be sniffed at

 

[Rroff]

And a boat-load faster today, after driver optimisation and Kepler performance falling off a cliff.

Kepler-Titan vs 290X today would be a massacre.

As I keep banging on - its not as much as some reviews would make you think - watch the videos of side by side benchmarks like are linked in the two recent threads on here where you can see that actual proper boost clocks are in effect and the 780ti and 290X actually trade blows over a large number of games - the 290X has come on a long way and in some cases like The Division have some major wins while the wins on the Kepler side tend to be much reduced but it certainly isn't a massacre like some mainstream sites make out.

Its been quite a ride for the 290X cards though - when they were new I benchmarked my old 780GHz at max clocks against my friends 290X Tri-X at max clocks and it wasn't even a competition 90% of games I was clearly ahead - 18months to a couple of years later or so they were pretty much dead even with the 290X taking a few more recent titles.

 

[AllBodies]

As I keep banging on - its not as much as some reviews would make you think - watch the videos of side by side benchmarks like are linked in the two recent threads on here where you can see that actual proper boost clocks are in effect and the 780ti and 290X actually trade blows over a large number of games - the 290X has come on a long way and in some cases like The Division have some major wins while the wins on the Kepler side tend to be much reduced but it certainly isn't a massacre like some mainstream sites make out.

Its been quite a ride for the 290X cards though - when they were new I benchmarked my old 780GHz at max clocks against my friends 290X Tri-X at max clocks and it wasn't even a competition 90% of games I was clearly ahead - 18months to a couple of years later or so they were pretty much dead even with the 290X taking a few more recent titles.

Weirdly enough, I managed to find this super up to date: https://www.techspot.com/review/1418-geforce-gtx-780-four-years-later/

Although it doesn't have the 780 Ti, or the 290X, it does have the 970 and the 390.

So if we assume the 780 Ti, with appropriate overclock, would be at 970 level. And the 290X would be at 390 level. Then turns out you're right.

I guess other review sites must be reusing old scores from old drivers and low initial clockspeeds, as you said.

Also at the same time it's pretty ridiculous the 1050 Ti keeps up with a highly overclocked 780 most of the time. That's 132mm2 vs 561mm2, and only 1 node different, so the 1050 Ti would only be ~264mm2 if it was on the same node. Maxwell and Pascal were a disgustingly good lesson in efficiency

 

[Rroff]

I guess other review sites must be reusing old scores from old drivers and low initial clockspeeds, as you said.

I have no idea what they are doing - I assume running limited to on paper boost clock or something but look at this thread https://forums.overclockers.co.uk/t...o-anything-about-the-kepler-tanking.18784789/ and the other thread I link in one of my posts and there are links to several recent videos comparing a 780ti and 290X side by side where its possible to verify they are running at proper boost clocks and aside from 1-2 games where the 290X takes a good lead mostly they are closely matched.

However most 2nd generation Kepler cards are limited these days by their 3GB of VRAM versus the 4GB on the 290 cards and nVidia's updated DX11 drivers on Kepler seem to rely a lot on having decent system RAM bandwidth to perform their best.

My intention isn't really to defend Kepler as such as its had its day and has its inadequacies but the image of them projected by some mainstream sites is just way out compared to the reality - I mean I can't even underclock my 780GHz low enough to match some of the gamer's nexus scores with their 780 cards.

 

[Nablaoperator]

Interesting post on Reddit -

"Reference models are already stockpiling at one company with which I have contacts. They apparently have "low thousands" ready for release. They usually have tens of thousands, though, but this is apparently a very expensive GPU. Supposedly priced above the 1080 at the top (my contact doesn't have pricing info outside of internal rumblings).

The company is a primary supplier, not a reseller."

Could be BS though.

Finally some good news .. +1
Now i dont even care abt fake news .. the way amd marketing has mangled the launch..
zilch coverage of PDXLAN.. they bloody deserve it i say

 

[Nablaoperator]

Im quite excited to see what Navi will be like. 7nm and its the architecture which Raja had most involvement in. Not to mention is using "Next gen memory". Also the potential to make a big GPU out of higher salvageable smaller dies could be quite interesting.

i think its fair to assume that amd has treated us like dirt during vega launch.. its like my first experience with the company
and i am not talking about rx vega's performance.. its just the way they have managed the launch
i am very much convinced by rx vega [i am an 'on paper' kind of a guy] and will buy 2 of them, but i must say amd is trying hard to ensure that i rather buy 2 ti's instead

next time i would advise never to wait for their products.. there should be a massive backlash from the community and amd fans, if you dont provide them the right feedback how can they improve

 

[Steampunk]

Interesting post on Reddit -

"Reference models are already stockpiling at one company with which I have contacts. They apparently have "low thousands" ready for release. They usually have tens of thousands, though, but this is apparently a very expensive GPU. Supposedly priced above the 1080 at the top (my contact doesn't have pricing info outside of internal rumblings).

The company is a primary supplier, not a reseller."

Could be BS though.

That top end is going to be the water-cooled version, so expect it to be silly money.

 

[beany_bot]

i think its fair to assume that amd has treated us like dirt during vega launch.. its like my first experience with the company
and i am not talking about rx vega's performance.. its just the way they have managed the launch
i am very much convinced by rx vega [i am an 'on paper' kind of a guy] and will buy 2 of them, but i must say amd is trying hard to ensure that i rather buy 2 ti's instead

next time i would advise never to wait for their products.. there should be a massive backlash from the community and amd fans, if you dont provide them the right feedback how can they improve

Amazing story.... Considering RX Vega hasn't actually been launched yet... Lol.

 

[Mauller]

The Next Gen Memory is just GDDR6 FYI...

And likely 14000 MHz on 256-bit for the first Navi cards out in late-2018, which will likely be RX480 sized. So we're talking a ~13 Tflop GPU with 448 GB/s bandwidth.

Next gen memory can also mean anything after HBM2, such as LCHBM or HBM3.

 

[Mauller]

Where did you read that? I'd be interested in having a read up on it too if you can point me in the right direction please.

There is nothing in the API's that allows control over the rasterisation of the scene. Even nvidia's TBR is all driver controlled, even down to a per app basis if required.

 

[r7slayer]

i think its fair to assume that amd has treated us like dirt during vega launch.. its like my first experience with the company
and i am not talking about rx vega's performance.. its just the way they have managed the launch
i am very much convinced by rx vega [i am an 'on paper' kind of a guy] and will buy 2 of them, but i must say amd is trying hard to ensure that i rather buy 2 ti's instead

next time i would advise never to wait for their products.. there should be a massive backlash from the community and amd fans, if you don't provide them the right feedback how can they improve

Don't wait then buy what ever is available. I really don't get your post it's kinda seems like a spoilt kid kicking off. Get two Ti then mate. Ive got one and its fantastic. Does not mean i'm not enthusiastic to see what RX Vega does or Navi because im a fan of technology and enthusiast. AMD hasn't treated anyone like dirt they just have kept quiet and took too long to release the high end for this gen. It's your first experience with the company? How? Enlighten me because if your waiting for vega your either new to PC gaming or just rambling on about nothing.
Anyways do what ever you want get a Ti but posting things like AMD deserve backlash seems kinda ridiculous. They just need to realise keeping thier heads in the sand does not help either.