Will AMD release any high performance GPU's in Q2 or Q3?

[Deleted]

I wonder if fanboys of any colour actually have any true technical interest. Vega will be an engineering spectacle, whether it runs dx11 like Nvidia's cards or not. An engineer does not say "no change is sexy, no change makes me sultry, I love no change"; instead both innovation and alternate methodologies are welcomed, both in hw and sw, as this is what drives evolution and provides a better product to the end user.

[Life-Binder]

hopefully they havent forgotten that ultimately fps is what matters in a gaming card

[Remilia]

what? farming what with a gpu? that wasnt dead cause of asics?

Ethereum can't be farmed with ASICs.

- - - Updated - - -

I wonder if fanboys of any colour actually have any true technical interest. Vega will be an engineering spectacle, whether it runs dx11 like Nvidia's cards or not. An engineer does not say "no change is sexy, no change makes me sultry, I love no change"; instead both innovation and alternate methodologies are welcomed, both in hw and sw, as this is what drives evolution and provides a better product to the end user.

Personally I want to see how the HBCC translates in professional and gaming. I don't really care too much about the numbers but how it works. It's also why I was disappointed in consumer "Pascal" (not P100) as it was very much the same as Maxwell. It wasn't the performance part that disappointed me but the technological side of it was boring.

[Deleted]

hopefully they havent forgotten that ultimately fps is what matters in a gaming card

FPS in what though, that's the thing. Do we aim to cater for an API that wasn't intended to utilise modern hw, or embrace the future that massively enhanced parallelism provides, and build/expose architecture for that?

Personally, I would be perfectly happy with a Vega card which performs as well as the equivalent priced Nvidia card in dx12/vulkan, but accomplishes it in a different manner. dx11 is a hardware dog, it can't be binned soon enough; no more middleware/bloated drivers, no more "driver game optimisations", you have either written a shit engine or you haven't. I hope AMD's architecture encourages transition to the lower level API's befitting modern hw. At the end of the day, as gamers we want fps, dx11 is an extremely inefficient way of delivering this, why would we want to encourage it's retention when there is better tools for the job?

[Life-Binder]

dx11 aint going anywhere for at least several more years (will get slowly reduced/phased out)

its tried and true, and familiar


cant live in a vacuum, devs play a super important role in this ecosystem and they currently know dx11 much better then they do 12 & vulkan

and majority of them aint that great at pc versions optimizations, so they want simple when they can get it



- - - Updated - - -

you have either written a shit engine or you haven't.

the former option will pop up much more frequently then you or me would like

- - - Updated - - -

id rather transition slow and careful and steady, then take an immediate plunge



thats why I like Nvidias approach:


Maxwell - very good at DX11, worse at dx12, which didnt really exist that time

Pascal - better at Dx12 and async, but not all in on it .. still just as good @ dx11

Volta - will see, but the independent thread scheduling looks like it will be good for dx12s parallel/async approach .. maybe some other things they will put there

[Amalaric]

AMD disappoints yet again and the much hyped Vega will fizzle out into oblivion.

[Low Hanging Fruit]

Truth is most people don't care. Idiots are stupid enough to just scream about which side is best because they like the side best and not really care about which is really best. Mostly because people like things for different things but that is unacceptable to most sadly. So we just get to hear about people hanging onto their one or two things they care about and ignoring the rest while also the other side does the exact same thing.

All I care about is competition between the two to keep the market in a more fair position. Which is best I could give zero fucks about. I will do my research each time I do a build and do my best to meet the budget and performance scales I require.

[VooDsXo]

AMD disappoints yet again and the much hyped Vega will fizzle out into oblivion.

That is too funny, because the new gpu has specs pretty much on par w/ the 1080Ti w/ only literally slight differences. So, like the 7xxx series, we will be seeing a lot of pure AMD builds, we have solid cpu's and now vega with info coming out about everything but the rops but that is simply because the specs are slightly off kilter to normal.

- - - Updated - - -

Truth is most people don't care. Idiots are stupid enough to just scream about which side is best because they like the side best and not really care about which is really best. Mostly because people like things for different things but that is unacceptable to most sadly. So we just get to hear about people hanging onto their one or two things they care about and ignoring the rest while also the other side does the exact same thing.

All I care about is competition between the two to keep the market in a more fair position. Which is best I could give zero fucks about. I will do my research each time I do a build and do my best to meet the budget and performance scales I require.

I'm with you, Idc what the base brand is Nvidia or AMD both are worthwhile companies, and have had battles back and forth on and off for years. The thing I don't like is how many sub par companies snatch the chips up to make horrible cards.

[Life-Binder]

http://www.fudzilla.com/news/graphic...stack-cost-160

According to our multiple, well informed industry insiders, the 4GB HBM 2 memory stack cost around $80 USD. Vega needs two HBM stacks to get to 8GB memory putting the price all the way up to $160 just for the memory.

Our sources have implied that in normal cases GDDR5 / GDDR5X memory cost a few times less than the $80 for 4GB. This makes Vega or any other HBM 2 card much more expensive, as the Bill of Materials (BOM) explodes compared to traditional GDDR5 / GDDR5X cards.

wow ...

[Butler to Baby Sloths]

http://www.fudzilla.com/news/graphic...stack-cost-160

wow ...

How is this surprising? Everybody and their dog knew that HBM is expensive.

[Life-Binder]

seen a bunch of ppl claim "HBM2 is not much more expensive then GDDR5" / "Vega is not more expensive to make then GP102" etc.

[Deleted]

seen a bunch of ppl claim "HBM2 is not much more expensive then GDDR5" / "Vega is not more expensive to make then GP102" etc.

We don't know how good yields are of either, so in grand scheme it may well be cheaper to make. No way to tell rly.

[Life-Binder]

well I see no reason why TSMC would have worse yields on GP102s smaller chip then on Vegas

[mrgreenthump]

well I see no reason why TSMC would have worse yields on GP102s smaller chip then on Vegas

Well Ryzen 8 core apparently has 80% yield, so if they use the same manufacturing for Vega, it's not too bad. Also memory is not all of what a GPU costs, it's probable that HBM will push prices higher, but not by a lot.

[Life-Binder]

afaik they use GF for Ryzen and TSMC for Vega

either way Ryzen 8 die is 192 mm2, while Vega is atm estimated @ ~500-550 mm2

[mrgreenthump]

End of July it is then.. Which pretty much means, that the drivers still aren't ready. And/or they are waiting for HBM2 pricing.

[Gray_Matter]

End of July it is then.. Which pretty much means, that the drivers still aren't ready. And/or they are waiting for HBM2 pricing.

The chat on the AMD reddit says that it's drivers. The presentation on Vega was underwhelming. I'm guessing that there are some serious issues to iron out.

[Life-Binder]

they show any direct 4K gaming becnhmarks ?

- - - Updated - - -

ok so apparently they showed a Prey Demo at 4K, unknown settings, on 2x Vegas (!) but no FPS counter

a single 1080Ti does up to 80 fps in Prey 4K max graphics I believe


that doesnt look so good, if the perf were better then they would have showed more stuff I think .. a different game/more games, an fps counter, no dual GPUs



/sigh, such a contrast between their CPUs and GPUs atm, Intel is scrambling, while Nvidia just walks all over them

[Triggered Fridgekin]

The only thing which Vega can do for me is fill in that pricing/performance gap between a 1060 and a 1070. If AMD somehow, somehow pushed a 4GB Vega for $425 CAD then that would park it right inside the gap which separates mid-range to high-end. Speculating of course but I want to say it'll be more powerful than a 1070 so for that price it would be fairly substantial for a larger portion of gamers who don't have $800+ or even $500+ to spend on a graphics card.

4GB with HBM isn't bad either (as shown with DE:MD) due HBCC providing it's a passive benefit rather than another layer of coding required and we all see how lazy most developers already are...

Otherwise, it's not something which really concerns me if I'm being entirely honest. My 290x is an over-built jet engine and will easily last until the next wave of mid-range cards which will hopefully offer more a meaningful performance boost than what we have now to justify the cost.

[Gaidax]

End of July it is then.. Which pretty much means, that the drivers still aren't ready. And/or they are waiting for HBM2 pricing.

Nah, most likely supply. There was this rumor that they have only 16k GPUs on hand so they probably want to prio all that for the professional market first, while they get some stock up.