AMD: Speculation, Rumors, and Discussion (Archive)
- Thread starter iMacmatician
- Start date
- Status
Considering AMD's operating margins are in the toilet I doubt their production costs are a major factor in pricing. If nVidia sold 150w 1070's at $199 they would make a crap ton more money than AMD even with a larger die.
gamervivek
Regular
If it's just using a single 6-pin connector, then it's pretty likely that it's not 150W on the dot. Guru3D peg gtx1070 power usage at 161W.
D
Deleted member 13524
Guest
Last edited by a moderator:
It's a little premature to make deep conclusions based just on TDP numbers, I think.
My conclusion for now is that AMD has managed to catch up with Nvidia in terms of perf/W, and probably perf/mm2. (It remains to be seen if they've also improved their ability to use BW effectively. The fact that it needs the same 256 bit memory system as a 1070 is not encouraging.)
I think that's good news for them and it will make them much more competitive than they were, without distancing themselves in any way from their competitor.
I'd give them a cookie for all of that if it weren't for the fact that it took them 2 more years to get to this point. This is business, not kindergarten.
Is it TDP or AMD's Typical Board Power, which Fury X has shown has a margin of error roughly equivalent to one Polaris 10?
At least with Polaris and unlike other AMD boards, the 6-pin connector gives them less wiggle room on "typically" not blowing out the marketing number for board consumption. It's not a perfect limit, but it's more definitive than what we've gotten for a while.
At the dawn of the DX12 age, I'd rather wait and pay the premium, or pay the early adopter tax. That's just me, but I don't have the patience to pay more of my time into every game the developer/AMD louse up with every release/patch/driver update.
D
Deleted member 13524
Guest
Getting a dual RX 480 for "similar performance" of a GTX 1080 is indeed a less safe bet to get day-one performance with good scaling (at least until DX12's mGPU matures properly like 3dilettante said).
However, if you're going to get an adaptive sync display, the cost of the cards alone isn't the only money saver. You'll also have to deduct some $150 or more of G-Sync premium over FreeSync.
We can now find 34" IPS Curved Ultra-Wide monitors with Freesync for close to 800€, but I can't find a G-Sync equivalent for less than 1100€.
2* 8GB Polaris 10 + 34" FreeSync = 1300€
GTX 1080 + 34" GSync = 1800€
This is a 40% price difference. It's a premium that might make you think twice.
However, if you're going to get an adaptive sync display, the cost of the cards alone isn't the only money saver. You'll also have to deduct some $150 or more of G-Sync premium over FreeSync.
We can now find 34" IPS Curved Ultra-Wide monitors with Freesync for close to 800€, but I can't find a G-Sync equivalent for less than 1100€.
2* 8GB Polaris 10 + 34" FreeSync = 1300€
GTX 1080 + 34" GSync = 1800€
This is a 40% price difference. It's a premium that might make you think twice.
Lol, you want a source for a hypothetical scenario? The source is me.
A $199 1070 will generate a lot more sales and revenue than a $199 480 for obvious reasons - stronger brand and higher performance. The relative die sizes would be inconsequential.
Nvidia must not want money then.
Why do you say that? They will make even more at $379+.
My point is that price has nothing to do with die sizes or cost. Price is a measure of perceived value.
It must costs nvidia nothing to make bigger dies.
Love_In_Rio
Veteran
I would not bet the silicom cost is cheaper with GloFo 14nm Polaris 10 process than with TSMC 16nm 1070 even taking into account the die is larger in Nvidia's case.
PlanarChaos
Newcomer
GCN excluding Fiji seems to preform equal to NVidia cards of about 10-15% less FLOPS in the real world when boost clocks are in play.
Silent_Buddha
Legend
Revenue =/= profit. It's entirely possible to have incredibly high revenue and still be operating at a loss. It's also entirely possible to have low revenue and operate at a profit. For example, back during the GTX 280/260 versus Radeon 4870/4850 price wars, Nvidia had high revenue and operated at a loss (AMD also was operating at a loss, I believe).
A 199 USD GTX 1070 would have significantly less profit margins (assuming it had a profit margin at all at that price) than a 199 USD Rx 480.
It'll sell a whole lot more obviously if both were priced equally, AMD would just have to lower the MSRP and still have the same or higher profit margins.
I doubt either company is wanting to get back into a price war again. As long as Nvidia has healthy sales, they aren't going to price pressure AMD just to price pressure AMD. The past few years Nvidia has been VERY focused on their operating margins. I don't see that changing. The only way it does change is if Rx 480 massively eats into GTX 1070/1080 sales. And I just don't see that happening, they are in 2 completely different market demographics.
Regards,
SB
Last edited:
PlanarChaos
Newcomer
Isn't everything I have heard that TSMC has a ton of demand? We will learn when Apple A10 comes. If they have lots of demand than Apple gets cheap price and TSMC raises price on other wafers to boost margins a bit?
GloFo is probably more of excess supply and so lower costs. Plus AMD has WSA. Even making 1 million off of 100 million in wafers is better margin than throwing 100 million away. Right?
I think this is a pretty smart move on AMD's part. It looks like it will very likely offer more performance and longevity than any GP106 device (assuming it isn't 256bit here). And $199 is an incredibly important price point, so kudos to AMD for recognizing & capitalizing on this.
Also looks like they are finally catching up in perf/w, but my educated guess is that they are likely still behind Nvidia (though probably very close now).
The one concern I have long-term is the clock speed. Obviously, if they could clock it higher they would... While it probably won't be a huge issue for Polaris10 / RX480, since it looks to sit above GP106 (though if it comes at 1.8GHz+ base clock who knows) and have a time to market advantage as well. But I don't see how an RX490 has much hope of competing with the 1080.
Still, definitely the right move for AMD to make, just wish I could buy one tomorrow (for my own sake and theirs)....
Also looks like they are finally catching up in perf/w, but my educated guess is that they are likely still behind Nvidia (though probably very close now).
The one concern I have long-term is the clock speed. Obviously, if they could clock it higher they would... While it probably won't be a huge issue for Polaris10 / RX480, since it looks to sit above GP106 (though if it comes at 1.8GHz+ base clock who knows) and have a time to market advantage as well. But I don't see how an RX490 has much hope of competing with the 1080.
Still, definitely the right move for AMD to make, just wish I could buy one tomorrow (for my own sake and theirs)....
nV hasn't had an operating loss in a long time, And I think that was the last time they reported a loss. But Revenue also dropped though in those Q's (q1 2009 and q2 2009)
AMD was aslo reporting a loss at the time.
Silent_Buddha
Legend
We'll have to wait until actual reviews to see where AMD is with regards to perf/watt. A 6 Pin connector just means that the power required is greater than 75 watts but lower than 150 watts. If the wccftech piece is anywhere close to what Rx 480 ends up at it'll be anywhere from 110 watts up to 130 watts. If it's closer to the former then that's pretty good. If it's closer to the latter then at least they are closer to Nvidia than they have been for the past few years.
Regards,
SB
Regards,
SB
- Status
