Vega, Massively faster than GTX 1080 in DOOM 4K.

[cdawall]

assuming you were running well over 100 fps... seems about right. These are only about 10-15% apart. Thus is why I hate talking in fps especially without context. If it was 10-20 fps faster at 50 fps, it's 20-40% faster... which its clearly not.

/done splitting hairs..

Nah my card was thermally throttling like a fat hot pig. It was 20+ % faster going to a single 480 with an overclock.

 

[Kanan]

Have an R9 390 8 GiB and it's equal or better than the RX 480 (1366 MHz) 8 GiB in most games. The only downside is that the R9 390 sucks a lot of power.

If you overclock both the 480 should be faster, however I was talking about the 470. Also efficiency is something great and not something that should be dismissed easily. That said, the 470/480 are the better Gpus.

 

[Nergal]

Choose one - you can't have both.

And again - it's a total failure if it doesn't frankly demolish the 1080. It needs to target the 1080ti.

Depends on the price.
A card that demolishes the 1080 should be worth more than the 1080
So, 999$?

Lol, doubt that

 

[ratirt]

(just 780, not Ti though)

Vega is 500mm^2 (bigger than titan pascal) and is using expensive memory. nVidia, on the other hand, has 314mm^2 1080 that sells like cookies, despite being heavily overpriced.

AMD can't win a price war with that in mind (and nVidia WILL enter it, if AMD cards start grabbing market share)

Well that's just 780 not Ti.


So it's bigger? and? if 1080 is overpriced and it is. even if Vega is using expensive memory modules 1080 is overpriced so much that if vega shows up it will still be cheaper. NV is feeding on the lack of competition and gets a lot of cash from customers for it's products.

 

[Flow]

I don't know about that though, when amd released the fury cards I was kinda disappointed about the price they asked for them. Granted, performance was close to the nvidia flagships at the time, but still.
Where are the days you had a real bang for your buck with amd?
So while I'm excited about new amd flagships, I'm also worried they will price it to its performance rating versus nvidia.

 

[FYFI13]

Guess I'm dumb. Bought a used 7970 3GB off ebay for $250, it failed playing games, with lots of crashes, even sitting in windows idle it would blue screen every hour, just a burned out card. Sold it back to another miner as is on ebay after some unsuccessful RMA repairs for $350 and bought a 770 2GB for $320. I'd have considered buying AMD for the second card purchase but they were so price inflated from the mining, I guess I was dumb for going with a cheaper better performing card and making money in the process . If I bought a new AMD at that time, I'd have to pay $500 for a R9 280X.

I hope they gave you some sort of "best buyer" signature or something? Where do you even find prices like that?

 

[Recon-UK]

One of my more successful threads then i guess.

 

[ratirt]

I don't know about that though, when amd released the fury cards I was kinda disappointed about the price they asked for them. Granted, performance was close to the nvidia flagships at the time, but still.
Where are the days you had a real bang for your buck with amd?
So while I'm excited about new amd flagships, I'm also worried they will price it to its performance rating versus nvidia.

We'll have to wait and see but i doubt AMD will get fixed price matching NV. You have to remember that AMD needs market share and that will be achieved only if the price tag is right and performance is great no other. Think about rx 480. It is a great card and they didn't bump the price up to match GTX 970 which AMD's card was competing with in the mid range segment. it was way cheaper. Hopefully they will do same thing with Vega. If vega price tag gets around 1070 it would be great. Despite if it's faster or matches 1080's performance

 

[Lox]

Hopefully AMD will release a few press samples for reviewers soon.
While things might still change, at least we’ll have an idea about performance for more than just doom running on Vulkan.

I’m currently looking at buying 3x 1440p 144hz displays for my racing rig with either freesync or gsync, but unsure if I want to go green or red.

 

[kruk]

If vega price tag gets around 1070 it would be great. Despite if it's faster or matches 1080's performance

Small Vega 11 which has smaller die and GDDR5X can be priced at 1070 level for sure. The big Vega 10 with HBM2 and almost 500 mm^2 can in no way be cheaper than 500$. The full Vega 10 die will probably be 650$ and the cut die 550$ at launch (same as Fury). They cannot win a price war with a smaller Pascal GP104 die, therefore the cut die must beat GTX 1080 by ~10% or more and the full die must comfortably beat the upcoming 1080 Ti. I think this is achievable with all the changes Vega brings, but let's just wait for benchmarks .

 

[HD64G]

Small Vega 11 which has smaller die and GDDR5X can be priced at 1070 level for sure. The big Vega 10 with HBM2 and almost 500 mm^2 can in no way be cheaper than 500$. The full Vega 10 die will probably be 650$ and the cut die 550$ at launch (same as Fury). They cannot win a price war with a smaller Pascal GP104 die, therefore the cut die must beat GTX 1080 by ~10% or more and the full die must comfortably beat the upcoming 1080 Ti. I think this is achievable with all the changes Vega brings, but let's just wait for benchmarks .

The cut Vega 10 should clearly win 1070 in average and get close to 1080. And its price to bleed nVidia could be at 1070 levels. Full Vega 10 should do the same to 1080 and be very close to 1080Ti at DX12/Vulcan.

 

[Kanan]

Optimal case would be that Vega10 beats Titan X / 1080 Ti and Vega11 is about as fast as 1080 or faster, but that's probably me dreaming. Actually I'm wondering why they insisted to use HBM2, knowing its hard to manufacture, has shortages and is expensive - why didn't they use G5X? I can only guess because it costs a lot of power, the controller is bigger, takes some chip size, and the power consumption of G5X is clearly higher than HBM2. That said, I hope it pays off for them that they are going with the premium solution here, else G5X would've been probably the better solution because it's cheaper, if Vega fails to challenge Titan X/1080 Ti.

 

[ratirt]

Optimal case would be that Vega10 beats Titan X / 1080 Ti and Vega11 is about as fast as 1080 or faster, but that's probably me dreaming. Actually I'm wondering why they insisted to use HBM2, knowing its hard to manufacture, has shortages and is expensive - why didn't they use G5X? I can only guess because it costs a lot of power, the controller is bigger, takes some chip size, and the power consumption of G5X is clearly higher than HBM2. That said, I hope it pays off for them that they are going with the premium solution here, else G5X would've been probably the better solution because it's cheaper, if Vega fails to challenge Titan X/1080 Ti.

For me if vega hits 1080 then I don't think I will buy it and I think a lot people will do same. AMD should think about the pricing before drowning in their decisions.

 

[the54thvoid]

Just read more rumours about a Pascal refresh. If Nvidia do release a Pascal 20X0 range, Vega absolutely cannot afford to only best the GTX1080. If the rumoured refresh is 10-15% faster and it comes out in Summer/Autumn, AMD's card has to seriously perform.
May can't come soon enough frankly. At least it gives AMD time to debug the shit out of Vega and squeeze it for every ounce of performance.

 

[BiggieShady]

Vega absolutely cannot afford to only best the GTX1080

AMD cannot afford to lose money, but they can afford to make less than nvidia ... sure, Vega is a bigger silicon but even just below inflated GTX 1080 price point it'll still make AMD money and leave room for price battle

 

[Nergal]

At least it gives AMD time to debug the shit out of Vega and squeeze it for every ounce of performance.

Now that is something that won´t happen with AMD.
One of the few things we can relay on with AMD is that their cards perform better in time.
Some extra quicktime(see what I did there) isn´t going help.

 

[Tatty_Two]

AMD cannot afford to lose money, but they can afford to make less than nvidia ... sure, Vega is a bigger silicon but even just below inflated GTX 1080 price point it'll still make AMD money and leave room for price battle

Whilst I get what you are saying, I am not so sure, with NVidia's market share they are selling more therefore they in theory at least can sell for less and still ensure profitability, AMD does not currently have that luxury, although I appreciate that may change, what is clear though is that this is very much a new architecture that AMD has spent a lot of time and resource in developing so they need to recoup pretty quick...... just my thoughts.

Now that is something that won´t happen with AMD.
One of the few things we can relay on with AMD is that their cards perform better in time.
Some extra quicktime(see what I did there) isn´t going help.

Again I agree however if they lose a share of initial sales because of early driver review's they have lost part of their potential market, by the time their driver evolution has improved performance by 10 - 15% some who have waited until launch to compare may have gone green, those consumers are unlikely to sell 4 months later when they see the gap (may) have closed.

 

[the54thvoid]

Whilst I get what you are saying, I am not so sure, with NVidia's market share they are selling more therefore they in theory at least can sell for less and still ensure profitability, AMD does not currently have that luxury, although I appreciate that may change, what is clear though is that this is very much a new architecture that AMD has spent a lot of time and resource in developing so they need to recoup pretty quick...... just my thoughts.



Again I agree however if they lose a share of initial sales because of early driver review's they have lost part of their potential market, by the time their driver evolution has improved performance by 10 - 15% some who have waited until launch to compare may have gone green, those consumers are unlikely to sell 4 months later when they see the gap (may) have closed.

This is the truth. Unlike an aged whiskey (which you can't buy until it's aged anyway - so not the best analogy) it is never in a company's interest to release something that gets better as it gets older unless it is also the best at release. If it's not the best at launch - that is what buyers go on. Also, with GCN, on DX12 and Vulkan the improvements seen in the past on DX11 wont be as obvious. Close to metal API's require less driver overhead so AMD's 'time served' advantage will mean less moving away from DX11.

 

[Kanan]

Well Vega isn't GCN anyway, it builds up on GCN 4 which already had way better utilization than GCN 3 and 2, and reading about their new tech, it's obvious they invested a lot of time to have higher utilization of their engines this time (High Bandwidth Cache etc.). AMD very well knows what their problems were in the past and they are not gonna repeat them, so it's very well possible that Vega can use "every ounce of its performance". I wouldn't bet on it, but its possible.

 

[Palladium]

Just read more rumours about a Pascal refresh. If Nvidia do release a Pascal 20X0 range, Vega absolutely cannot afford to only best the GTX1080. If the rumoured refresh is 10-15% faster and it comes out in Summer/Autumn, AMD's card has to seriously perform.
May can't come soon enough frankly. At least it gives AMD time to debug the shit out of Vega and squeeze it for every ounce of performance.

Yeah that is what I don't get either: AMD comes out with products that is slightly better and cheaper than the competition, except they are also somewhere between 1 year (vs GP104), 2 years (vs GTX 970) to 6 years late (vs Intel) while expecting us to kowtow to them as if time value of money never mattered.

 

[medi01]

This is the truth. Unlike an aged whiskey (which you can't buy until it's aged anyway - so not the best analogy) it is never in a company's interest to release something that gets better as it gets older unless it is also the best at release. If it's not the best at launch - that is what buyers go on. Also, with GCN, on DX12 and Vulkan the improvements seen in the past on DX11 wont be as obvious. Close to metal API's require less driver overhead so AMD's 'time served' advantage will mean less moving away from DX11.

Well, except price on AMD cards is competitive at launch most of the time, so that FineWine thing is just a nice bonus on top of stuff such as FreeSync.
It's possibly not good for AMD in that they could have sold more chips for at a higher price, but it's a clear win for the customers.

 

[the54thvoid]

Well, except price on AMD cards is competitive at launch most of the time, so that FineWine thing is just a nice bonus on top of stuff such as FreeSync.
It's possibly not good for AMD in that they could have sold more chips for at a higher price, but it's a clear win for the customers.

Vega won't be cheap but as long as it gives Nvidia a shock, it will be worth it.

 

[ratirt]

Vega won't be cheap but as long as it gives Nvidia a shock, it will be worth it.

If there will be (and rumors say it will) 20X0 release of refreshed Pascal, and you state there's a possibility for that this year, than if vega ends up slower then the planned to released revisions of pascal maybe price will go down. Maybe it would be wise to wait a bit longer. on the other hand we may be waiting to infinity. Either way my expectations for Vega cooled down a bit knowing there's a revision of Pascal approaching. IF AMD foreseen it and they got something in store for new pascal that would be great? Well that would be a neat trick. Vega ain't out yet so having something better not likely.

 

[medi01]

knowing there's a revision of Pascal approaching.

What is the source of this "knowing"?

We are in 1H 2017, when nVidia asks for 700$ for a 314mm^2 chip with GDDR5x memory

 

[ratirt]

What is the source of this "knowing"?

Rumors only but not mine. I was going to ask the54thvoid. He mentioned it but didn't give any links. Rumor is still only a rumor but it is something.