AMD Hawaii GPU might launch on September 25
Click here to post a comment for AMD Hawaii GPU might launch on September 25 on our message forum
GeniusPr0
Titan/780 technically existed at the same time as the GTX 680. NVIDIA just didn't release it. GTX 680 needs around 170W in games. That should be enough to tell you it's not a real high-end part. It's a mid-range.
---TK---
you have evidence gk110 was ready on March 22, 2012? Would love to read up on it.
GeniusPr0
*Technically* existed. GeForce cards have been above the 200W line since GTX 2xx series. To drop so much power consumption means there's an obvious window for more potential. It's especially true when it doesn't even require architecture tweaks. It should ring bells.
---TK---
I was looking for something factual ,thanks anyways
yasamoka
Typical 170W power consumption which we know is that because not all games and benches reach TDP. 190W TDP which makes sense because of four things:
1) Crippled compute, simpler chip design, less power draw
2) Not as much of a jump from Fermi (GTX580), compared to the jump from GTX280 to GTX480.
3) Artificial power limits.
4) Compare to 7970, the added power consumption (which is not that much, mind) is most probably due to the wider memory bus and the non-crippled compute performance.
To have GK110 ready at GTX680's launch with that TDP (which I think is being hit more frequently than the GTX680's power limit at similar overclocks) does not make much sense, bar the problems Nvidia had with the GK104 yield at the time.
Remember GTX680 stock? That card was mythical. Then, surprise, the GTX670 hits a month later at a very sweet price point and promised to be a great card at its price point, offering slightly less performance than the GTX680 for $100 less. Yields.
Chillin
It was ready as the Tesla K20, problem is that they had the supercomputer order to fulfill first; the Oak Ridge National Laboratory Titan.
Loophole35
^ Yeilds just weren't there the process was not mature enough GK110 could have never happened at that time for the enthusiast market.
sykozis
GK110, compared to GK104, is a very large chip. It was built strictly for massive performance with little, if any, regard for power consumption....unlike GK104.
Kepler was a new architecture. It's not unrealistic for a new architecture built around reduced power consumption to show such results. You can't determine potential or even market position by power consumption.
---TK---
they could not even stock the 680`s in any abundance at launch. it was almost vaporware. they had a good supply of 670`s though that were not good enough to be 680`s shader wise
GeniusPr0
What is potential determined by then? Since no one in this thread has really made a killing point. Yields were bad but that's not the point. Once the architecture exists, and based on maximum wattage thresholds, it seems completely reasonable to compared and extrapolate within those constraints. I don't see anything better to be used as tools.
The way I see it is AMD is releasing Hawaii but it took them a new architecture to possibly beat Kepler, which effectively means Kepler gets the longest lifespan ever.
I still remember when the NVIDIA CEO said Kepler would be unstoppable. He wasn't kidding.
---TK---
you cannot use tdp alone to base gpu prowness or "potential". that is a silly argument.
GeniusPr0
It's not alone. Once there is an existing architecture to argue for along with power consumption and constraints, it's an easy argument. Now, instead of making an empty statement without explaining...
CPC_RedDawn
Hoping this new GPU will be priced VERY aggressively against the TITAN and 780. If its more than a 780 they can forget it and I will just pick up a second 7970 and be happy for at least another two years. If it comes out at £399 or even better £350 I shall be all over that like a dog on heat :banana:
GeniusPr0
My guess is $600.
---TK---
wrong, you based the 680 not being high end based solely on the lower tdp compared to previous generations higher tdp.
yasamoka
Let me add to that. If GK110 was at the time of release of the GTX680 release, we'd have a doubling in performance over the GTX580 while maintaining the same power consumption in a generational jump. Is that a trend you see in generational jumps?
GeniusPr0
Not solely. I said it's 195 mostly using 170W which is way under the norm for what's potentially there including what has been done in the past. And I'm waiting on your argument about TDP.
GeniusPr0
http://tpucdn.com/reviews/NVIDIA/GeForce_GTX_Titan/images/perfrel.gif
Not quite
yasamoka
Well look at the jump anyways.
GeniusPr0
10% shy. 2880 cores with slightly higher mem speed would seal the deal I think. Not sure how good the clocks would be though.
