The TDP can be 1000w or 50w, less than 10% is what the actual power usage difference is.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
No GK110. Instead GK114 will be the GTX 780
- Thread starter Gregster
- Start date
The TDP can be 1000w or 50w, less than 10% is what the actual power usage difference is.
Is the 780 going to be the big Kepler then with better than atrocious compute power?
Regardless, I think nvidia is in a strong position. A 680 with a 384-bit bus would have a similar heat output to a 7970, possibly even less, but would outperform it. It remains to be seen though if they can make a big chip with decent clock speeds and some compute power with decent yields.
Regardless, I think nvidia is in a strong position. A 680 with a 384-bit bus would have a similar heat output to a 7970, possibly even less, but would outperform it. It remains to be seen though if they can make a big chip with decent clock speeds and some compute power with decent yields.
Give it up Humbug. Digging deeper when you should have just left it alone
Explain how a lower TDP gives more power head room than one with a higher TDP, and do it without ignoring actual power consumption results.
Only Nvidia will know what they will be releasing and right up untill release, we will hear many rumours of which 99% will be wrong. NviDiA (NDA will be in full force) and I wish they would give out a little info but that won't be happening
Soldato
- Joined
- 18 May 2010
- Posts
- 23,684
- Location
- London
It feels like Nvidia are reaching the max potential of the architecture and not able to make such great leaps from series to series.
I can see AMD having the faster cards when the new cards do get released.
I can see AMD having the faster cards when the new cards do get released.
Only Nvidia will know what they will be releasing and right up untill release, we will hear many rumours of which 99% will be wrong. NviDiA (NDA will be in full force) and I wish they would give out a little info but that won't be happening
True. All I want to know to begin with is will they be voltage locked? If so, team red again for me.
Because higher TDP goes hand in hand with higher power consumption - ergo, a card with a lower TDP typically has lower power consumption such as the 680 does with respect to the 7970. Which has already been pointed out means that Nvidia can boost performance merely by beefing up the existing design.
^^^^ Yet in practice its actually only 10% at best, you can give a device a TDP plucked out of thin air, what matters is what its actually doing.
No, i think its the same for AMD, the architecture is reaching its full potential, another 15% for the 8### series and thats the lot.
I think we will stand as we are today GTX 7## vs 8### series, with the acception of Nvidia catching up with AMD at high res due to better vRAM
No, i think its the same for AMD, the architecture is reaching its full potential, another 15% for the 8### series and thats the lot.
I think we will stand as we are today GTX 7## vs 8### series, with the acception of Nvidia catching up with AMD at high res due to better vRAM
Are you kidding me?
You keep on about power but this isn't the point. Kepler uses a lower TDP, it is cooler at running than Tahiti.
Basic common sense tells you that something cooler has headroom to give more grunt. I am no expert on dye sizes or compute technology, Duff man/Xsistor/Drunkenmaster will bring you up to speed on these things but the TDP shows that Kepler has more to give than it's AMD counter part (unless AMD can get it running cooler).
If there is one thing with computers we should all know, is that there is always a work around even if they are voltage locked
Last edited:
10% is a lot more to play with than you think - your understanding of TDP as people have pointed out is a little wonky.
EDIT: Theres an excellent presentation on it somewhere I think it was about ARM core development but I can't find it easily
Last edited:
You do realise that the two are linked - if it's too small the card overheats and burns out, too big and the cooler is oversized for the duty and makes it cost more than it needs to. It's not plucked out of thin air!
Its not that simpe Gregster. AMD could easily have released the 7970 at 1.1v and the TDP/heat output would be a lot closer to the 680. It would obviously have a lot less over clock headroom though on stock volts, also similar to the 680. As a result of this they had almost certainly had better yields than the 680, but higher heat output.
Think of it another way, if the 680 was released locked at 1.225v they would have had a higher TDP, and higher over clock headroom at stock.
Think of it another way, if the 680 was released locked at 1.225v they would have had a higher TDP, and higher over clock headroom at stock.
I never have and don't deny Kepler runs cooler, it uses up to 10% less power so its bound to, but its a few degrees, nothing more.
Its also very GPU cooler dependant.
TDP is not a measure of how hot or cold something actually runs, or how much power its actually using compared with a device with a different TDP.
The reason i keep going on about actual power consumption is because that is what actually matters, "a device with a TDP written on the box that is a lower number than another device means the device with the lower number uses less power and runs cooler, fact" no its not, the device which actually uses less power in practice, is the one that runs cooler, fact. yes.
in that kepler does indeed use less power and run cooler, but 10% at best to me is not enough to be significant.
Not sure I follow but put it this way:
680 locked at 1.175V and a 7970 locked at 1.175V, which card would perform best?
Another way to see it is, Kepler runs cooler at full load over Tahiti. If Nvidia do it properly, they have more room to add grunt than Tahiti does.
Edit:
I give up explaining it Humbug. Google is your friend
Last edited:
This isn't true by the way...
SB or IB
Which one is cooler? SB
Which one uses less power? IB
What I'm saying is the 7970's were released with artificially high voltage for their clock speed. Hence the high over clock headroom. They could have easily been released at 1.1v and have a lower TDP, which is why TDP isn't everything.
We know Kepler is more efficient however from the 7990 vs 690.
What we don't know is how much the difference can be attributed to the extra vram, 384-bit bus or the extra compute units.
For the 8xxx GCN could be stripped of gpu compute units and instantly it is far more efficient and has a lower TDP, think Pitcairn. A Pitcairn scaled up to 350mm^2 would have nvidia worrying.
We know Kepler is more efficient however from the 7990 vs 690.
What we don't know is how much the difference can be attributed to the extra vram, 384-bit bus or the extra compute units.
For the 8xxx GCN could be stripped of gpu compute units and instantly it is far more efficient and has a lower TDP, think Pitcairn. A Pitcairn scaled up to 350mm^2 would have nvidia worrying.
True, there are other variables. like bOrking the TIM to use your example.
That is spot on and the 384bit memory is why it leaves the Kepler standing at higher resoloutions. If you was to drop to a 256 bit bus or raise keplers bus, things would be different.
.