AMD: Speculation, Rumors, and Discussion (Archive)

Status
Not open for further replies.
Hmmm, I guess Samsung is out of the picture? These are all made at Global Foundries.
Was this mentioned in the presentation? I didn't watch it.

Going with GF makes total sense if they're still bound to that agreement that requires them to pay for wafers that they aren't even using: it's almost as if they can produce the GPUs for no additional cost! ;)
 
Of course it is the price. Or to be exact price/performance. Yes, Pascal is amazing, but it is way outside of my budget. This one is within my budget thus I'm far more excited by RX480 than Pascal. Actually, pricing wise, it is even better than my expectation.
Also one other thing that I like is the potentially smaller card because my case is really cramped. I was going to buy 380 until I saw the length of the card could barely fit my case (at least for the cheaper brand that is within my budget. There are other more expensive brand that is smaller).
 
  • Like
Reactions: xEx
So for now, the only thing that's really impressive about the RX 480 is the price?

Great for the consumer, of course, but it's only something to brag about as a company if you can achieve that because of some amazing engineering that brought cost down as well: anyone can achieve a low price by cutting into margins.

And right now, I don't see any engineering exploits. Performance between a 970 and a 980 is not exactly overwhelming.
If we assume that a GP106 will be exactly 1/2 of a GP104, with 10 SMs and a 128-bit 8Gbps GDDR5, the 480 should outperform a 1060, but only barely, with very similar power.

So the 480 will have a quite formidable competitor once the 1060 gets to market.

Still, it's way better than having to use the abomination that was Tonga against gm206.

If GP 106 performs under Polaris 10, I'd imagine Nvidia will have to price it lower than they'd like. I'm guessing that they were aiming for the 249-299 USD price bracket for GP 106 with GTX 1070 starting at 379 USD.

If Nvidia have to price GP 106 at under 200 USD that leaves a large gap between GP 106 and GTX 1070.

Regards,
SB
 
I wonder how utilization went down to 51%, even mixing nvidia with amd's last gen still gets 90-100% gpu utilization in sli on ashes.
 
If GP 106 performs under Polaris 10, I'd imagine Nvidia will have to price it lower than they'd like.
Yes, but that's always a risk when you're in a market where one party primarily competes on price. After all the proclamations of the last 5 months, I don't think the price will be a huge surprise.

More important is that the 480 doesn't even come close to touching the 1070. That would have been a much bigger worry.

I'm guessing that they were aiming for the 249-299 USD price bracket for GP 106 with GTX 1070 starting at 379 USD.

If Nvidia have to price GP 106 at under 200 USD that leaves a large gap between GP 106 and GTX 1070.
Such gaps can easily be filled with a further cut down GP104. Say, 13 SMs with a 192 bit bus?
If one has to believe some forum experts, the yields on that chip must be low enough to make such a SKU feasible. ;)
 
I wonder how utilization went down to 51%, even mixing nvidia with amd's last gen still gets 90-100% gpu utilization in sli on ashes.


Well we don't know even know the settings used and what not, I wouldn't even use that test as anything other than proof of concept.
 
Yes, but that's always a risk when you're in a market where one party primarily competes on price. After all the proclamations of the last 5 months, I don't think the price will be a huge surprise.


Such gaps can easily be filled with a further cut down GP104. Say, 13 SMs with a 192 bit bus?
If one has to believe some forum experts, the yields on that chip must be low enough to make such a SKU feasible. ;)

I'd imagine Nvidia would be adverse to doing that if they can sell the dies for significantly more in a 1070 which is already a salvage die. Their yields would have to be extremely low if they were to sell a further cut down GP 104. Either that or demand wasn't there for GTX 1070/1080. Both of those prospects I view as extremely unlikely. Especially the past few years where they've been carefully growing/maintaining their operating margins.

Although I suppose if Rx 480 (and presumably a future Rx 480x?) were to start eating into GTX 1070 sales, selling further cut down dies meant for GTX 1070 would be better than not selling them at all.

Regards,
SB
 
More important is that the 480 doesn't even come close to touching the 1070. That would have been a much bigger worry.

How so? Seems like it is within reach if; 1. RX480 is not a full P10 or 2. Custom AIB cards have free reign.

EDIT- RX480 looks like it is ~15-25% under a GTX1070.
 
I'd imagine Nvidia would be adverse to doing that if they can sell the dies for significantly more in a 1070 which is already a salvage die. Their yields would have to be extremely low if they were to sell a further cut down GP 104.
Yes, I agree. And I do think that their yields are just fine.

Although I suppose if Rx 480 (and presumably a future Rx 480x?) were to start eating into GTX 1070 sales, selling further cut down dies meant for GTX 1070 would be better than not selling them at all.
Or they could just lower the price of 1070 GPUs without explicitly changing the MSRP. Other than the GPU and maybe the cooler, the 8GB BOM of a 1070 and a 480 should be pretty close, so if AMD can afford to sell the whole package at $230, it means that the current 1070 silicon markups can withstand healthy cuts while still maintaining excellent gross margins.
 
Well we don't know even know the settings used and what not, I wouldn't even use that test as anything other than proof of concept.
Keeps my hopes up of possibly seeing a single 480 near a 1080 in ashes benches.
How so? Seems like it is within reach if; 1. RX480 is not a full P10 or 2. Custom AIB cards have free reign.

EDIT- RX480 looks like it is ~15-25% under a GTX1070.
In crossfire fury x achieves about 66% increase in performance, assuming similar for this would put it [single 480]at 36% slower than a 1080 at ashes [going by the benchmark]. In most benches 1070 is 20 to 25% slower than 1080. So that would put a single 480 at 10-15% slower than a 1070 as compared to a 1080[ in some dx 12 benches].

But since fury x with 980 ti sli achieves near 100% utilization, I believe fury x crossfire should achieve near 100% utilization. So the 480 crossfire at 51% utilization may not be commesurable, true 480 performance may thus be higher.
 
Or they could just lower the price of 1070 GPUs without explicitly changing the MSRP. Other than the GPU and maybe the cooler, the 8GB BOM of a 1070 and a 480 should be pretty close, so if AMD can afford to sell the whole package at $230, it means that the current 1070 silicon markups can withstand healthy cuts while still maintaining excellent gross margins.

I'd agree with that. It's what would make the most sense. We'll just have to wait and see where Rx 480 performance in the real world actually ends up, and if people find the budget proposition more attractive than paying a much larger premium for better performance.

Regards,
SB
 
Status
Not open for further replies.
Back
Top