Gtx 1080

[Life-Binder]

Fury costs a lot to make because its big and has expensive HBM1

they cant just drop prices on it

[Remilia]

Yeah... Fury is in a very odd position. Even if you compare similar sized silicon, say the GM200 to Fiji, Fury has the additional HBM and Interposer. While Interposer is considered a few bucks, the HBM is 2.5x(?) price of GDDR5 and about 2x more of GDDR5X. HBM1 is also only made by Hynix and isn't made standard by JEDEC, HBM2 however any manufacturer can have a go at. Even though AMD with Hynix (and lots of co) made HBM together, doesn't eliminate the cost of production still.

[Vash The Stampede]

The Fury cards sit in a niche where the excuse is that they're small. Which matters for those making small gaming PCs. But it's also AMD's fastest GPU, which isn't the markets fastest GPU. Since its release in June 2015, you would think the cost of making the Fury cards went down. But again, if they don't lower the price they will lose sales. The 1080 is just faster, with DX12 or without. But this isn't just with Fury cards, but with everything AMD does.

[Wries]

Ah.. Fury X.

It still costs 599 dollars when I check Newegg. That's the MSRP of the freaking 1080, but even the 1070, that soon will be available for much cheaper, can handily beat it. Right now the Fury is in such a small niche of usefulness that if it's like you say that they can't get the manufacturing costs down, I'd assume they have stopped producing new units at this point.

And HBM has effectively limited them to 4GB, making it look even less attractive from a future-proofing perspective.

[Egore]

anyone know at what time of day and timezone did the 1080 officially go on sale? looking to plan for the 1070 release. thanks

[Shinzai]

Well, as I said last year with all the cards AMD released, they were just far too late. It's genuinely aggravating that they put so much effort into the 390 series and Fury series, but came a year after the 970/980. The effects of such a gap in launches is permanently reflected in their sales, which can easily be testified to from the Steam hardware surveys.

The 3XX series and the Fury don't even have enough of a percentile share in the charts to even show up as calculable figures. They're in the sub 0.30% ownership margin. The 2XX series actually has 1.32% share by comparison.

That's why I made such a point of them being quite literally required to respond quickly and deliver a product this year, within a reasonable time-frame. If AMD delivers on a physical launch date at Computex, then manages to push for the 490/490X equivalents being released within the year, they can easily muscle in on the overall market share. I believe this is why Vega is rumoured to have been brought forwards. Perhaps, they'll launch GDDR5X variants of the Vega cards, that might have been originally planned to release after the HBM2 variants.

On another, related note, I genuinely believe that Nvidia's GP102 Titan P/1080Ti will be running with GDDR5X and that they may have foregone using HBM of any sort on any of their "Pascal" consumer-grade cards. The actual difference between GDDR5X and HBM 2 is quite small when you factor in what the cards are physically capable of at the moment. HBM is overkill even on a Titan P in my opinion. The 384 Bit bus Titan P that was 'leaked' the other day makes sense, in that it will give a 670~750 gigabytes per second transfer rate, compared to HBM 2's 1 TB/S. I 100% believe Nvidia will use the cheaper option, but still charge prices as though it had HBM 2 on board.

So now, I'm just more curious than ever as to AMD's plans for Vega and whether they'll be able to release it this year, or not. The 480 is extremely positive on a price/performance ratio, regardless of what may be said about it. If it performs as well as, or better than it does on paper, it'll probably be the most popular 1080p card since the 970.

[Artorius]

The problem with GDDR5X Vega is that they'd need to have 2 different memory controllers at the chip which consumes space.

Maybe small Vega is GDDR5X and big Vega is HBM2, who knows.

[Shinzai]

The problem with GDDR5X Vega is that they'd need to have 2 different memory controllers at the chip which consumes space.

Maybe small Vega is GDDR5X and big Vega is HBM2, who knows.

That's pretty much what I would expect. It's been mentioned many times before, by myself in particular, that AMD is rumoured to have a first come, first serve basis with SK Hynix's first batches of HBM 2, due to their level of assistance in its creation. However, it was also noted a number of times on various sites that AMD was rumoured to be tapping GDDR5X for some of their products. Which ones and what levels remains unclear however.

The 480 rocking GDDR5 is interesting however. That would imply to me that the 49X equivalents may have GDDR5X and the Fury equivalents would be the HBM 2 targets. This would mirror what Nvidia's done with its split usage of GDDR5 and 5X.

[Drunkenvalley]

So what's the verdict on custom cards so far? In terms of overclocking, did the additional 6 pin provide any improvement? I imagine cooling is better, but I don't really care about that one way or another, given my usual watercooling nature.

[Deleted]

So what's the verdict on custom cards so far? In terms of overclocking, did the additional 6 pin provide any improvement? I imagine cooling is better, but I don't really care about that one way or another, given my usual watercooling nature.

I just finished reading the review on MSI's Gaming card and they weren't able to go beyond 2.1ghz. Some games not even go beyond 1.9 or 2.0 even.

You can take a look and read.. Only hope right now for further OCing (if Voltage doesn't become a problem..) would be on MSI Sea Hawk and EVGA Hybrid

[Drunkenvalley]

Looking at the review, it doesn't really mention whether it was thermals or power that stopped them. :| Hmm.

[haval1976]

Wow, it is the king of videocards!

[Butler to Baby Sloths]

The problem with GDDR5X Vega is that they'd need to have 2 different memory controllers at the chip which consumes space.

Maybe small Vega is GDDR5X and big Vega is HBM2, who knows.

You can't just swap out the memory technology like that though. I'm fairly sure that small vega will be the lower-binned big vega GPUs and will have HBM2, because I don't think they want to inflate the cost of the chip by redesigning the memory controllers for small vega.

[Artorius]

You can't just swap out the memory technology like that though. I'm fairly sure that small vega will be the lower-binned big vega GPUs and will have HBM2, because I don't think they want to inflate the cost of the chip by redesigning the memory controllers for small vega.

I'd agree if it was a single chip and a cut-down version of it. But we have Vega 11 and Vega 10 so it's plausible to have 2 different memory controllers.

Example:

RX 490 - Cut-down small Vega with GDDR5/X
RX 495 - Full small Vega with GDDR5X
Fury successor - Cut-down big Vega with HBM2
Fury X successor - Full big Vega with HBM2

[DeltrusDisc]

Wow, it is the king of videocards!

I honestly don't know why people are "amazed" every time a new GPU comes out. I think it's cool when they show improvements, but I'm not amazed, like, we know it should beat older cards, it's newer tech, etc.

It would make no sense and be seriously ho-hum if they launched the GTX 1080, able to do what the GTX 980 can do, but just for 20% less power and created a little less heat.

So, of course it's the king of video cards, it shouldn't be a surprise, though...

[FlawlessSoul]

I honestly don't know why people are "amazed" every time a new GPU comes out. I think it's cool when they show improvements, but I'm not amazed, like, we know it should beat older cards, it's newer tech, etc.

It would make no sense and be seriously ho-hum if they launched the GTX 1080, able to do what the GTX 980 can do, but just for 20% less power and created a little less heat.

So, of course it's the king of video cards, it shouldn't be a surprise, though...

It's a spammer, Deltrus...

[Artorius]

I honestly don't know why people are "amazed" every time a new GPU comes out. I think it's cool when they show improvements, but I'm not amazed, like, we know it should beat older cards, it's newer tech, etc.

It would make no sense and be seriously ho-hum if they launched the GTX 1080, able to do what the GTX 980 can do, but just for 20% less power and created a little less heat.

So, of course it's the king of video cards, it shouldn't be a surprise, though...

Apple every year: "IT'S THE BEST IPHONE EVER MADE"

[DeltrusDisc]

Apple every year: "IT'S THE BEST IPHONE EVER MADE"

Wow, such shock, very news.

[JestersShadow]

I honestly don't know why people are "amazed" every time a new GPU comes out. I think it's cool when they show improvements, but I'm not amazed, like, we know it should beat older cards, it's newer tech, etc.
...

Usually the first cards to come in the Next generation don't beat the Premium best of the best ones from the last. Considering this 1080 beats the Titan it is a huge shock to many people. Hell there are still cards in AMD's 7970 range that still beat many of their newer cards.

[Artorius]

Usually the first cards to come in the Next generation don't beat the Premium best of the best ones from the last. Considering this 1080 beats the Titan it is a huge shock to many people. Hell there are still cards in AMD's 7970 range that still beat many of their newer cards.

"many people" don't have any clue about anything related to semiconductors. If the die-shrunk Maxwell with that clock boost from the manufacturing process change didn't outperform it's previous self then we'd have a problem.

You seem to forget that the node change was from 28nm planar to 16nm FF. They skipped 22nm entirely and did a huge leap. We had numerous generations built at the same manufacturing process and that's why the improvement was minor from gen N to gen N+1. This time around each tier is expected to deliver 60% better performance in comparison to last gen simply by density, then you add architecture improvements on top of it (ok we can ignore this part with "pascal") and the higher clocks that is made possible and suddenly the performance increase of the 1080 in comparison to the 980 isn't really impressive. It's expected.