If the 680 is 10% faster than the 7970 with a lower TDP, it would be a real winner.
If the 680 is 10% faster than the 7970 with a lower TDP, it would be a real winner.
Just gonna throw these rumors in here. Good picture of the card too. I do have a really hard time believing the rumored TDP.
http://www.geeky-gadgets.com/nvidia-...ed-14-03-2012/
http://www.engadget.com/2012/03/14/r...-prove-it-doe/
"There cannot be true dispair without hope." - Bane
Yea the TDP thing throws me off. Gets me thinking about how long ago Intel and AMD were bashing each other about how TDP is measured. Wondering if they changed their methodology.
Do you know what was the biggest fault of the Fermi architecture ? Why it never was the monster it was supposed to be ? Why there was no GTX 490 ?
Power consumption. The performance per watt was horrible and AMD was miles ahead in that area.
If the Kepler architecture is more power efficient than the AMD's, this will really change the market and create some interesting situations.
So yes, TDP is really, really important.
Last edited by haxartus; 2012-03-14 at 07:16 PM.
NDA might be lifting 22nd 6am PST.. Still trying to get confirmation for it though.
Thank you for redundancy, haxartus. But that's honestly simply not a big deal to me when I'm buying a graphics card. Performance and price are far, far bigger priorities. Unless the price/performance between the two are really, really close the TDP usually* won't bother me.
*But if it's fucking batshit then it's absolutely relevant, yes.
I have never even looked at the TDP of the GPU's i have brought, only for perfomance and how much they cost.
Performance and price are the deciding factors in buying GPU's for most, as it should be. But when looking over architecture history and TDP, it's been turned around, and can indicate NV's position with their architecture. I mean if they can have the same performance while having lower TDP on the same manufacturing processes when in the past they have always had a higher TDP to remain competitive is a strong indicator NV engineers have done some great changes. It may also indicate the headroom in the architecture. So my assumption is that it's very possible this is not going to be their top end GPU. Or they will save higher binned GPU's for the HPC market where they can gain some good revenue. Most likely testing the waters to see how AMD will respond.
I built a new machine back in October and put a 570 in it. I'm greatly looking forward to see what the new line is like!
Except that power efficiency doesn't translate directly into performance and as a consumer the most important factor is Price / Performance.
Hm, I guess I'm different then. I'm mostly interested in TDP and performance, I believe I save more in doing so.
Especially when considering things like when you have to buy a new PSU and airflow and multi-card setups.
What nVidia really need to do, for their own sake and survival in the Desktop graphics business, is a new 8800 (GT/GS/GTX et c).
Geez, what gave that away?
Moreover, are you going to keep your whole wannabe-teacher thing going? Because it's extremely uninvited, random and redundant. What you're saying are all things I know, but none of it has meaning to me, because usually the TDP numbers thrown around are fake rumors, and the only time we actually get straight up fact is the same day I'm reading a review on the final product itself.
When we reach such a point there's no turning back, the numbers are there to stay. And my attention will be drawn to price/performance once more. Certainly other factors are there, but price/performance compared to competing products is just going to be really big by comparison.
TL;DR: TDP has little meaning to me. I'm an end user who's (window-)shopping.
Last edited by Drunkenvalley; 2012-03-15 at 12:10 AM.
Nobody should be shocked by the 195W TDP rumor.
The GTX 680 has 5 VRM phases.
The HD 7970 has 6 VRM phases.
The GTX 680 has 2GB of memory.
The HD 7970 has 3GB of memory.
The GTX 680 has a ~320mm^2 die.
The HD 7970 has a ~352mm^2 die.
The GTX 680 has 6+6pin PEG connectors.
The HD 7970 has 6+8pin PEG connectors.
Again, the assumption is the GTX 680 slots in between the HD 7950 and 7970. The rumored specs support that. If the GTX 680 beats the HD 7970, that might just be because is clocking its 28nm chips higher. Maybe a 1005MHz GTX 680 is roughly 5% faster than a 925MHz HD 7970, but if both are clocked at 1100MHz, then the 7970 will be faster...
What does 195W vs 215W TDP mean for single-GPU cards? Almost nothing. The only way it has any effect on the market is when they'll have to dial back clockspeeds or disable CUs in dual-GPU cards. The Nvidia side may get interesting if the dual-GPU card is based on GK104 and a single GK110 card is also released in the meantime.
There has been a lot of back and forth "bickering" going on in here. If you really want to just direct comments at each other and argue nuances, please feel free to do so via PM. Keep this thread clean with constructive discussion pertaining to the GPU and it's architecture. Thanks.
I do care about the tech, but when it all comes down to it I care a whole lot more about the results.
Indeed. To be fair, AMD were quite modest when they sold the cards. Most of them can at minimum reach a 1k MHz, and a lot of them casually max out all the sliders in Overdrive.