Place your bets on Vega

[sparks]

IF amd had something they would be shouting from the roof top.
They don't

 

[NKD]

Polaris was hailed as the second coming of blowjob jesus, and it turned out to be a 970, literally the same TDP and performance of the then-year-old 970. It was simply a die-shrink of GCN, nothing more.

Now AMD are toutng Vega to be the next coming of titty-f*** moses malone skywalker, and we all know how this is going to turn out. I cant remember the last AMD launch that didnt feel like a sad handjob: sure, it beats nothing at all, but not by much.

Only by stupid people that expected. Why would a card that was hailed assecond coming of Jesus have only 36 CUs? Haters and fanboys hyped it up and haters enjoyed and fanboys were disappointed. Too many false rumors around it. Your post is a perfect example of how it was hyped. Second coming of Jesus lol. I didn't think that about Polaris. Only people with unrealistic expectations did and hype/fake rumors form wtftech.com

 

[JustReason]

Gladly. Official boost is 1733 which, with an OC of 2030mhz, would give us a 17.3% bump.

If we go by this average it's 14% bump.

Naturally this doesn't account for memory overclock.

For comparison, the RX480 cards would have to regularly clock to 1485mhz to match this OC headroom

RX 580 would have to hit 1571mhz regularly

I have seen higher but assume this is indicative if the whole. My 290 with a 947 base OCs to 1130 easily (maybe higher but don't want to add V) which correlates to 19%. So does that mean my 290 is better? No so talking about OC headroom, which most love leaving out the boosting clocks of Nvidia to inflate OC headroom, means little unless there is a tie and that is the lone deciding factor.

 

[Ieldra]

I have seen higher but assume this is indicative if the whole. My 290 with a 947 base OCs to 1130 easily (maybe higher but don't want to add V) which correlates to 19%. So does that mean my 290 is better? No so talking about OC headroom, which most love leaving out the boosting clocks of Nvidia to inflate OC headroom, means little unless there is a tie and that is the lone deciding factor.

Yes, precisely. Assuming a tie the card with the better headroom takes the win. The 980Ti went up to like 1120mhz boost at stock? At 1480mhz that's a 32% increase. You can see that reflected in the sales.

If a Vega card matches a Pascal card. Let's say a 1080, but does not have comparable OC headroom, it will lose in the performance charts.

1080Ti @ 2GHz = 14.336Tflops. Assuming stock big Vega is 12Tflop it needs an extra 20% just you match the specs on paper, and that's not to mention historical tendency for NV uarchs to make better use of shader resources than GCN

 

[TaintedSquirrel]

Ref 980 Ti is 1200~1250 MHz gpu boost.

 

[Ieldra]

Ref 980 Ti is 1200~1250 MHz gpu boost.

Not according to data.

maximum of 1202Mhz. Average of 1130mhz.

Most websites report the maximum boost clock, some report maximum boost clock per game, neither of those two values are very meaningful when the clocks are not sustained

 

[Armenius]

I have seen higher but assume this is indicative if the whole. My 290 with a 947 base OCs to 1130 easily (maybe higher but don't want to add V) which correlates to 19%. So does that mean my 290 is better? No so talking about OC headroom, which most love leaving out the boosting clocks of Nvidia to inflate OC headroom, means little unless there is a tie and that is the lone deciding factor.

The reference 780 Ti from the same generation boosted to around 940 MHz during gaming out of the box with a quoted 928 MHz boost from NVIDIA. [H] got it up to 1163 MHz in their testing. That is a 24% overclock.

 

[Snowdog]

Unless AMD makes massive, massive improvements in power efficiency for Vega, I just can't see how they will compete with a 1080Ti.

The RX580 draws about the same amount of power as a GTX 1080 while under load:
http://www.tomshardware.com/reviews/amd-radeon-rx-580-review,5020-6.html

The RX580 and GTX 1060 perform close to the same, but the GTX 1060 draws close to half as much watts under load.

The GTX 1080Ti is a 250W TDP card. If AMD doesn't improve power efficiency, then their equivalent Vega card will draw near 450W, which is just insane, and require a giant water block for the reference card. That might happen (like with the 295X2), but it will probably slot in close to the 1080 in performance, but 250W~280W TDP on air instead.

Unless ALL of the leaks are wrong a Single Vega GPU is NOT going to compete with 1080 Ti. Its a 4096 SP part, so about GTX 1080 performance level.

Also, multiple leaks all point to a dual GPU card. This is the way they will get better than 1080 Ti performance, Here they can downclock and undervolt a bit to keep power under control.

 

[bezant]

Nvidia stole all the HBM2.

/s

Yeah Nvidia just reported results and they did not give a good answer as to why their inventory had climbed that much. Remember their revenues were better than expected, so inventory should have declined.

 

[bezant]

[...] coming [...] blowjob [...] coming [...] titty-f*** [...] hand job [...] beats nothing at all, but not by much.

Exactly!

 

[Chimpee]

Yeah Nvidia just reported results and they did not give a good answer as to why their inventory had climbed that much. Remember their revenues were better than expected, so inventory should have declined.

Inventory is up due to buildup for Volta release.

 

[bezant]

Inventory is up due to buildup for Volta release.

Yeah, I understand that explanation, but it doesn't preclude stocking up on memory. If anything it implies it. Plus besides him saying that inventory was up because of new product launches, there were no specifics.

 

[razor1]

Q1 and Q2 tend to be less volume quarters, so nV might have miss calculated how much less Q1 would be and inventory built up.

If we don't see inventory changes for the better in Q2 then we can say its due to Volta, but prior to that, it can be just channel volume adjustments on a per quarter basis.

 

[bezant]

Q1 and Q2 tend to be less volume quarters, so nV might have miss calculated how much less Q1 would be and inventory built up.

If we don't see inventory changes for the better in Q2 then we can say its due to Volta, but prior to that, it can be just channel volume adjustments on a per quarter basis.

Thanks. Didn't they beat their own guidance? That would imply less inventory of existing products.

 

[Shintai]

It was specified in the conference that it was build up for new products.

 

[razor1]

hmm missed that, if that is the case, then yeah Volta build up could be possible, but I don't see why they would put that as inventory build up, since those orders would already by set, they would not put them in inventory, Weird.

Could be older products (left over maxwell parts)?

 

[razor1]

Thanks. Didn't they beat their own guidance? That would imply less inventory of existing products.

They beat their own guidance but guidance isn't the same thing as units sold

 

[bezant]

They beat their own guidance but guidance isn't the same thing as units sold

Let's say they expect to sell 100 units and guide for that many, they order 110 just in case, but sell 120. Inventory should decrease by 10. Margins did not go up, so units sold is where we are in terms of beating the guidance.

 

[razor1]

depends, margins on Pascal are higher then the left Maxwell parts, so, lets say they expected to get rid of left over stock of Maxwell parts at a certain rate but didn't and increased Pascal parts sales but, by volume just not as much as the drop in left over parts? Get more money but inventory goes up.

PS nV likes to keep its inventory around 70 days, which they aren't that much off, so just a little bit of change to what it is now at 77 can be rectified quite quickly.

 

[Chimpee]

Yeah, I understand that explanation, but it doesn't preclude stocking up on memory. If anything it implies it. Plus besides him saying that inventory was up because of new product launches, there were no specifics.

He was being vague since GTC is on the next day of earnings call and Huang doesn't want to discuss details of Volta. The only thing he really said buildup is for new product and wait until GTC to find out why inventory is up.

 

[Gideon]

depends, margins on Pascal are higher then the left Maxwell parts, so, lets say they expected to get rid of left over stock of Maxwell parts at a certain rate but didn't and increased Pascal parts sales but, by volume just not as much as the drop in left over parts? Get more money but inventory goes up.

PS nV likes to keep its inventory around 70 days, which they aren't that much off, so just a little bit of change to what it is now at 77 can be rectified quite quickly.

77 days is actually pretty deep if it continues to rise then Nvidia has a problem. 60 days or less is the usual goal in the manufacturing world. My guess is sales are tapering off, people that wanted to buy have and the rest are either waiting on Vega or figured to hold out till Volta.

 

[razor1]

77 days is actually pretty deep if it continues to rise then Nvidia has a problem. 60 days or less is the usual goal in the manufacturing world. My guess is sales are tapering off, people that wanted to buy have and the rest are either waiting on Vega or figured to hold out till Volta.

That is not, much for the past year they have been a 73 days , So no, its not much of a change. That is why when going from a major sales Q like Q4 to a season drop in Q1, inventory can shift a bit. Its not a big deal. And no this is nothing to do with future unreleased products and definitely nothing to do with AMD Vega.

nV when they had 60% marketshare were trying to keep inventory at 65 days, but they increased that as their marketshare grew, because season changes will affect them more, as seen q2 to q3 of last year. They need to ensure they have increased amount of stock now then before so no problems occur in the supply chain if there is an increased demand for what ever reason.

 

[bezant]

That is not, much for the past year they have been a 73 days , So no, its not much of a change. That is why when going from a major sales Q like Q4 to a season drop in Q1, inventory can shift a bit. Its not a big deal. And no this is nothing to do with future unreleased products and definitely nothing to do with AMD Vega.

nV when they had 60% marketshare were trying to keep inventory at 65 days, but they increased that as their marketshare grew, because season changes will affect them more, as seen q2 to q3 of last year. They need to ensure they have increased amount of stock now then before so no problems occur in the supply chain if there is an increased demand for what ever reason.

INteresting that you say that, because he said on the call that inventory is up because of new product. I read the transcript, and these days the SEC could go after him if he lies outright.

 

[TheHig]

Like a lot of folks here. 1080ish , a smidge better but eating more juice to get it done. Unless they got the efficiency down another notch or 3 after Polaris. I can keep my rx 480 under or near 150w (Afterburner numbers so def salt piles and not a scientific measurement) with an under volt and 1300 core. SO to get to 1080 and beyond Red will need to push the clock and the power. Nvidia just has such a huge lead in perf per watt I don,t see AMD catching up this round in that area.

But I am definitely looking for the 1070 competitor as an upgrade to keep an all AMD rig going unless the big Vega is really priced well that is.

 

[razor1]

INteresting that you say that, because he said on the call that inventory is up because of new product. I read the transcript, and these days the SEC could go after him if he lies outright.

Yeah they can but did what he stated hurt investors?

I'm not sure how Volta coming out for a specific market segment stopped Pascal HPC sales, those markets are more stringent on when and what to upgrade, unlike the general consumer segment. Its easier to keep those supply chains tighter.

 

[bezant]

Yeah they can but did what he stated hurt investors?

I'm not sure how Volta coming out for a specific market segment stopped Pascal HPC sales, those markets are more stringent on when and what to upgrade, unlike the general consumer segment. Its easier to keep those supply chains tighter.

Makes sense.

 

[Simplyfun]

Obviously nothing is going on in the Nvidia forum if we require an in depth discussion of Nvidias business prospects and management in a Vega thread... says something.

 

[CSI_PC]

Yeah they can but did what he stated hurt investors?

I'm not sure how Volta coming out for a specific market segment stopped Pascal HPC sales, those markets are more stringent on when and what to upgrade, unlike the general consumer segment. Its easier to keep those supply chains tighter.

Agreed and just to add.
HPC is one of the few segments that has multiple generations from both Nvidia and AMD still active.
Nvidia is still selling for now Kepler (mostly K80) into HPC as it has a competitive price and viability for those already on it and do not yet need latest tech solutions, and AMD still sells Hawaii based GPUs for similar segment.
And this will continue with Pascal and Volta.

Cheers

 

[Araxie]

Obviously nothing is going on in the Nvidia forum if we require an in depth discussion of Nvidias business prospects and management in a Vega thread... says something.

Vega thread related to Nvidia's card performance?.. wow.. of course is gona be talking about Nvidia if all the cards mentioned on the poll are from GTX 1070 and above.

 

[Simplyfun]

Vega thread related to Nvidia's card performance?.. wow.. of course is gona be talking about Nvidia if all the cards mentioned on the poll are from GTX 1070 and above.

Of course, but you side slipped the issue, the issue is who gives a flying fart about Nvidias stock levels. Amusing.

 

[Araxie]

Of course, but you side slipped the issue, the issue is who gives a flying fart about Nvidias stock levels. Amusing.

same people who cares about being always apologist because AMD have not enough income to have a high R&D cash for their GPUs so they don't deliver just a GTX 1080 (and maybe just barely) more than one year later?..

"Yeah we are AMD, we delivered a GTX 1080 comparable performance more than one year later"

 

[Simplyfun]

same people who cares about being always apologist because AMD have not enough income to have a high R&D cash for their GPUs so they don't deliver just a GTX 1080 (and maybe just barely) more than one year later?..

"Yeah we are AMD, we delivered a GTX 1080 comparable performance more than one year later"

Again, who cares, can you read the thread title? DO you understand what that means? Don't give me this BS. Open a thread in the Nvidia sub forum, I'm basically a green team believer, but I'm not in for thread crapping.

 

[GoldenTiger]

Again, who cares, can you read the thread title? DO you understand what that means? Don't give me this BS. Open a thread in the Nvidia sub forum, I'm basically a green team believer, but I'm not in for thread crapping.

Then why are you crapping on this thread?

 

[polonyc2]

performance will fall somewhere between the 1070 and 1080...you can bet the house on it

 

[Simplyfun]

Then why are you crapping on this thread?

Nice try.

 

[GoldenTiger]

performance will fall somewhere between the 1070 and 1080...you can bet the house on it

That's my guess, too. Too little, too late.

 

[razor1]

I'm still thinking its about the 1080 or just above it. There would be no reason to make this card if it wasn't.

These are the things we can say with almost certainty

225 watts for C1 or just above (based on AMD TDP ratings)

Doom performance around or just above 1080 gtx (depends on overclock of the 1080) for C1

8gb for C1

Unless there are major changes in drivers or pushing it into crazy ass power consumption ranges*which I doubt on both counts*, I don't see it going any more.

 

[{NG}Fidel]

Decent jump if they did go from fury x perf to 1080 perf. Disappointing to consumers but then that's based on price.

 

[Magic Hate Ball]

Decent jump if they did go from fury x perf to 1080 perf. Disappointing to consumers but then that's based on price.

It's really all about price.

If you ignore price, you could say "why even release the 1070? The 980TI already has that much performance, and it is older!"

PC hardware sales has always been bang for buck.

 

[yatz10]

The more I've thought about it, after initial pessimism, the new leaks etc., I've come around to expecting they'll manage to compete with the 1080ti. There will be compromises, e.g. it'll come at the expense of overclocking, TDP and cooling (and probably yield), but I figure they'll pull something out kinda like they did for the 390X vs. 980. Given they've waited so darn long, they kinda need to compete...