AMD Radeon R9 380X 'stacked memory' information emerges

[HEXUS]

AMD employees mention the 380X GPU and HBM in their LinkedIn profiles.

Read more.

[bae85]

so glad I waited to upgrade my GPU, this looks promising!

[Platinum]

300w jesus

[MrJim]

300w jesus

My thoughts entirely. I don't know how they can reconcile going on to say:

With an HBM implementation the graphics memory would be located on the GPU chip substrate, said to be similar to how the Xbox One architecture works. This should provide benefits in performance and power consumption.

It doesn't seem very power-efficient to me?

[LSG501]

unless 300w means the system as a whole, you usually see high end cpu's 'needing' 500w psu's.

Suppose we'll have to wait and see

[MrJim]

unless 300w means the system as a whole, you usually see high end cpu's 'needing' 500w psu's.

Suppose we'll have to wait and see

If you read the PC Perspective article they linked to, they seem to think that the card has a 300W TDP...

[CampGareth]

My thoughts entirely. I don't know how they can reconcile going on to say:



It doesn't seem very power-efficient to me?

total power vs power efficiency, going 1mph with 300mpg vs going 200mph with 300mpg, both are efficient but you'll go through more petrol per hour with the 200mph car.

[Domestic_Ginger]

More worryingly with the recent turnover at AMD whether they are smartening up their CV's.

On the other hand people who keep updating linkdin are usualy more concerned with making a career than making what they're supposed to.

[Jowsey]

total power vs power efficiency, going 1mph with 300mpg vs going 200mph with 300mpg, both are efficient but you'll go through more petrol per hour with the 200mph car.

Fair point. However, that means the 380X has to go very fast indeed.

I hope AMD really pulls a rabbit out. As we all know Nvidia is sitting on the 980ti which will be the true Maxwell flagship part. So no matter what AMD produces, Nvidia will respond to either battle AMD's new units or if AMD don't put up a fight, they will use the 980ti to embarrass AMD. I would just like some fierce competition please

[kalniel]

If you read the PC Perspective article they linked to, they seem to think that the card has a 300W TDP...

The 290X had a 300W TDP at some stage didn't it?

[Hicks12]

Maybe I am blind but I cannot see a mention of a specific 300 series gpu with regards to the 300w power statement so could it be that it is referring to the 395x? (or whatever the dual gpu version will be called). it could infact be great that they have gotten the tdp to such a level for a dual chip, maybe I am too optimistic but I wont buy in rumours till its released .

Still looks good that the stacked memory might be coming to the 300 series.

[gwp]

Is this new card likely to have less coil whine issues than the GTX 970 and 980?

[[DW]Cougho]

Is this new card likely to have less coil whine issues than the GTX 970 and 980?

My Crystal ball says the answer is "Horatio Nelson 1805"

In other words no body knows, coil whine is not fully understood on known cards - nevermind those that are unreleased.

[DanceswithUnix]

300w jesus

I wouldn't read too much into it. Probably just means the package & interposer technology can take 300W, not that it will actually generate 300W. I would expect to over engineer something like that, then eg overclockers won't blow it up too easily. Kind of like putting 200mph rated tyres on my diesel won't make it any faster.

ISTR 300W is supposed to be the power limit for a PCIe card, so it seems a fair target to work towards.

[jigger]

If AMD have gained 65% memory performance and used the 40% power saving to bump up the GPU performance I hope this card will pull 300 watts at full tilt. If the performance per MM2 is better than the R9 290 then a big voltage tolerant 28nm chip would be a performance beast and for me thats what highend is all about.

[AETAaAS]

Why are people worried about the TDP when not much else is known about the card? For all we know, this is a balls-to-the-wall card with the performance to match. You can't judge 'efficiency' without even knowing the performance. I'm interested in how this new memory setup will be leveraged and if this newfound memory bandwidth and architecture will have an overall effect on the GPU design. nudge that GCN revision number.