Computerbase did a test with undervolting the card, big drop in power usage while the performance drops just a little.
Looking at the power efficiency of this, and what it'll mean for where they can scale Vega to, you do have to wonder how on earth they're going to get there R&D money back.
Gibbo@OCUK said:Over 700 now, I expected good sales, but this is beyond my expectations for day one. Several lines have sold out infact.
I see competitors have also moved prices on 8Gb part up so I guess they are low on stock, I reckon I can hold my 8GB pricing as I did have well over 1000 units in stock for launch.
Shame on the 4GB part though, going to push AMD hard to manufacturer more, or put 8GB cards in 4GB boxes haha!
Computerbase did a test with undervolting the card, big drop in power usage while the performance drops just a little.
All these people claiming that all that matters is perf per dollar and that performance per watt is the devil and no one has purchased anything from the links I have posted of far, far greater perf per dollar (as well as better performance).
Makes me wonder if people claiming that they want to upgrade from their Fury X to 480s might not be telling the truth.
Hmmmm....
http://www.ebay.com/sch/i.html?_odk...0+-"R9+280x"+-"R9-280"+-"R7+240"&_sacat=27386
http://www.ebay.com/sch/i.html?_odk...nkw=R9+290x+-"R9+280x"+-"R7+240"&_sacat=27386
http://www.ebay.com/sch/i.html?_odk...0+-"R9+280x"+-"R9-280"+-"R7+240"&_sacat=27386
I think my 7950B is more like 230-250ish watts, I was really hoping this card would be around 390X performance and at 150ish watts, seems to be 15% slower and close to 180W.
price is actually to high for me in Canada anyways seems to be >300 for the 8GB.
I wonder if it is constrained by 32 ROPs vs 64 from the previous generation?
Sure I was. I said so myself several times. Making comparisons to GK104 and Pitcairn when GTX 680 came out, I figured 390x performance was reasonable at 1440p. That would have been a slight regression vs. GP104 when comparing GK104 with Pitcairn.
But nope. We're looking at 390 levels with absolutely no headroom and GTX 1070 power consumption. Like I said, unless Nvidia prices remain elevated or the used market doesn't come down in price, I'll likely get a ~$250 GTX 980. If used 980's are $280+, then I will look into a $199 RX 480.
But i looked further and found that the 380X has 8 ace and the rx 480 only has 4 ace.
http://www.anandtech.com/show/9784/the-amd-radeon-r9-380x-review
Same number of ACE's in the 480. They grouped up 2 ACE's and called them HWS. 1 HWS = 2 ACE.
4ACE + 2HWS = 8 ACE total.
I'm not quite ready to breathe a sigh of relief just yet, but it does seem this issue may be limited to certain unlucky cards rather than a systematic fault with the whole series. In particular with point #1 there, already I've seen the tinfoil hats going on with people suggesting AMD are deliberately pumping the cards out of spec to drive the performance up. That would be a tremendously risky not to mention grossly irresponsible move which also assumes they've deceived the external compliance tests as well for the sake of a relatively tiny amount of extra performance. I've always been a believer in cock-up before conspiracy and this to me looks like another example of that.1) The RX 480 has passed PCIe compliance testing with PCI-SIG. This is not just our internal testing. I think that should be made very clear. Obviously there are a few GPUs exhibiting anomalous behavior, and we've been in touch with these reviewers for a few days to better understand their test configurations to see how this could be possible.
2) Update #2 made by the OP is confused. There is a difference between ASIC power, which is what ONLY THE GPU CONSUMES (110W), and total graphics power (TGP), which is what the entire graphics card uses (150W). There has been no change in the spec, so I would ask that incorrect information stop being disseminated as "fact."
We will have more on this topic soon as we investigate, but it's worth reminding people that only a very small number of hundreds of RX 480 reviews worldwide encountered this issue. Clearly that makes it aberrant, rather than the rule, and we're working to get that number down to zero.
I think my 7950B is more like 230-250ish watts, I was really hoping this card would be around 390X performance and at 150ish watts, seems to be 15% slower and close to 180W.
price is actually to high for me in Canada anyways seems to be >300 for the 8GB.
We're all worried about future prospects. AMD has only now achieved parity with Maxwell in performance/watt. What happens when you have a 250W Vega vs a 250W 1080 Ti, both using HBM? How do we fit 2 Vega chips in that small gap between 160W and 250W?If you want to buy a used card, go and buy one. To act childish and nitpick on everything for a card that is placed in a bracket with no competition is downright silly.
If another card comes around in the next few years and provides a similar performance to dollar ratio, you bet that it will be purchased.
You melted cables 4 times and decided to do exactly the same thing a 5th time?As the owner of 5, count em, 5 melted PSU cables from the R9 290 8+6 pin debacle I would highly recommend you go with no less than 8+6 pin for the 480.
If I were less lucky I might have burned to death in the fire those would have absolutely caused.
What happens when you have a 250W Vega vs a 250W 1080 Ti, both using HBM?
I mean at arstechnica they only mention 4 :
http://arstechnica.com/gadgets/2016/06/amd-rx-480-polaris-review/
They only have 2 hws. You got it wrong. Or Arstechnica
We're all worried about future prospects. AMD has only now achieved parity with Maxwell in performance/watt. What happens when you have a 250W Vega vs a 250W 1080 Ti, both using HBM? How do we fit 2 Vega chips in that small gap between 160W and 250W?
The Polaris architecture enables a range of performance per watt improvements over previous-gen products. It really just depends on what two parts you're comparing, and where they fall on the frequency vs. voltage curve. The RX 480 is like 1.7-2X, depending on the apps you're looking at. The RX 470 is a more power efficient design, which does gives us the ~2.7X you're looking for.
So if this is true, does this mean the 470 could end up a much better deal for the low end user?
One is going to be $399 and the other will be $1299. The $399 one wont be able to be priced any higher because it will be slower than a 1070. Realizing this, it is entirely possible that AMD might just scrap the entire project, because what is the point in making it unless GF is willing to produce 500 mm^2 dies for under $150 apiece?
It was talked about when Fiji came out, one HWS can have the same functionality as 2 ACE units. I don't remember all the details, I'll have to go back and see if I can find it from the Fiji info.
A GPU that supports asynchronous compute can use multiple command queues and execute these queues simultaneously, rather than switching between graphics and compute workloads. AMD supports this functionality via its Asynchronous Compute Engines (ACE) and HWS blocks on Fiji.
Asynchronous computing is, in a very real sense, GCN’s secret weapon. While every GCN-class GPU since the original HD 7970 can use it, AMD quadrupled the number of ACEs per GPU when it built Hawaii, then modified the design again with Fiji. Where the R9 290 and 290X use eight ACEs, Fiji has four ACEs and two HWS units. Each HWS can perform the work of two ACEs and they appear to be capable of additional (but as-yet unknown) work as well.
I thought that too.
But the 380X has 32 ROPs, is lower clocked and has less memory bandwidth.
That cannot be the case.
Either aots is highly optimized for GCN 1.1 and 1.2 or there is a surprise waiting for us in the near future.
But i looked further and found that the 380X has 8 ace and the rx 480 only has 4 ace.
http://www.anandtech.com/show/9784/the-amd-radeon-r9-380x-review
What beating?apparently a main stream gpu beating a xx70 gpu from last gen is just catching up hahahaha
do these guys type with a straight face? or do they honestly believe that?
http://www.hardwarecanucks.com/foru...9-radeon-rx480-8gb-performance-review-24.htmlMoving on to DX12 and we see AMDs new architecture really coming into its own against the NVIDIA cards. It absolutely demolishes the GTX 970 across the board (even in NVIDIA-friendly games like Tomb Raider) and even manages to run circles around that once-expensive GTX 980. These tests show Maxwells performance in current DX12 applications is nothing short of embarrassing and proves this architecture simply wasnt designed with these types of workloads in mind.