WYP
News Guru
Last edited:
ASRock Radeon VII Phantom Gaming Graphics Card Pictured
- Thread starter WYP
- Start date
Dicehunter
Resident Newb
Looks to be just a reference card with Asrock stickers ^_^
WillSK
Active member
I think that is EXACTLY what it is. I don't understand the point of this in that case?Reminds me of the Fury X debacle all over again. AMD locking out AIB and only selling reference once more? :S
Dicehunter
Resident Newb
Same happens with all GPU's in general.
Can't see it happening, AMD will want to squeeze every last drop of performance out of it as possible and that in general needs people like Sapphire and MSI, Reference is good but to push things better power delivery etc... is needed, Unless AMD have already maxed the card out and the only thing to make better is cooling.
Last edited:
I agree. But I would have thought they wanted that with the Fury gen cards given the losses AMD were making at the time.
If anyone can pull out a killer Radeon card, its Sapphire!
Last edited:
I think it's probably more to do with the fact relatively few coolers are designed to dissipate such a high amount of thermal power from such a small surface area, due to 7nm this is a tiny 331 mm die but it seems to be running with similar power targets as Vega. Obviously there are now also two flanks of HBM2 on either side of the die so the total surface area of the interposer will be as wide as Vega to fit them but now even longer to accommodate the second bank, so you need a large surface mount but one that ideally has separate heatpipes or a well shaped vapour chamber that ensures the HBM areas that aren't in contact with the much hotter central pipes/chamber which also have to be quite beefy to take around 250W from a ~2-3cm wide area.
So far all the AIB manufacturers announcing cards have been kinda OEM ones who rarely do their own high end custom cooler designs, assuming Sapphire is the original manufacturer for this reference cooler.
So far all the AIB manufacturers announcing cards have been kinda OEM ones who rarely do their own high end custom cooler designs, assuming Sapphire is the original manufacturer for this reference cooler.
Last edited:
AlienALX
Well-known member
This is all because Maxwell smacked AMD for six and they've been unable to cope.
If you cast your mind back to pre Fury X AMD were calling it, and I quote "an overclocker's dream" . In reality? It had a water cooler on and you were lucky to get 50mhz out of it.
However, I am sure that Fury X would have been clocked far lower, and thus have far more headroom left in it had Maxwell not been so damn good. AMD have been on the back foot ever since. Their last success was the 290 and 290x which did compete well with Kepler Titan, but Maxwell smashed them to bits so they've been releasing fully clocked power guzzlers ever since.
And that won't change now. "Radeon VII" is already 2 years behind.
It's not got much to do with design philosophy, more to do with upfront costs, AMD hasn't had the money to design dies nearly as large as NVidia's largest since Bulldozer flopped & NVidia's kept releasing more enterprise sized dies to consumers as they've realised there's a market in areas previously thought to expensive for consumers with the Titan. Maxwell was designed more like a mobile GPU with a tile based rendered but also introduced us to the first 600mm^2+ consumer die ever, it was efficient partly because it could be such a wide architecture.
With 14nm the effect of the budget gap grew further, as AMD had to use 14LP(Low power) across their whole stack while NVidia could mix it up with 16FF on the high end parts, allowing them to also have better clock speeds or better efficiency at slightly lower speeds when compared to AMDs high performance parts.
Presumably, with 7nm AMD will still be using 200mm^2 dies at most with the first two "small" Navi parts and won't release something larger than R7 until 7nm has matured somewhat in a year or so. I'd expect them to still use a 64-CU design for their top end but with larger, wider more efficient cores with a greater instruction set.
Essentially, AMDs approach likely cost a tiny fraction of NVidia's but obviously still kept them relatively in the fight albeit with lower efficiency/top end performance.
With 14nm the effect of the budget gap grew further, as AMD had to use 14LP(Low power) across their whole stack while NVidia could mix it up with 16FF on the high end parts, allowing them to also have better clock speeds or better efficiency at slightly lower speeds when compared to AMDs high performance parts.
Presumably, with 7nm AMD will still be using 200mm^2 dies at most with the first two "small" Navi parts and won't release something larger than R7 until 7nm has matured somewhat in a year or so. I'd expect them to still use a 64-CU design for their top end but with larger, wider more efficient cores with a greater instruction set.
Essentially, AMDs approach likely cost a tiny fraction of NVidia's but obviously still kept them relatively in the fight albeit with lower efficiency/top end performance.
Last edited:
Dawelio
Active member
Yup, unfortunately I feel kinda the same here to be honest.
Well if this is true, then they are shooting themselves in the foot...
