The G80 currently sits at 420mm². Time for a diet?
Here's a real short news bit from the guys over at
DigiTimes:
"Nvidia has placed urgent orders with Taiwan Semiconductor Manufacturing Company (TSMC) for GeForce 8000 GTX and GTS series chips as demand for graphics cards supporting DirectX 10 is picking up, according to market sources, as cited by the Chinese-language Commercial Times.
Monthly shipments for the urgent orders, which will begin in March, will be equivalent to 3,000-4,000 12-inch wafers using 80nm technology, the sources were cited as saying. "
Since the chips are being migrated to 80nm, we can expect some pretty hefty changes in the way the 8800 series of cards are sold. As well, the change should have some particularly beneficial performance enhancements.
First of all, the die shrink should reduce the core's surface area from 420mm² to a bit under 380mm² if the process shrink involves perfect scaling. Power consumption will also drop due to the shorter distances between transistor gates. The two of these things combined mean that we should see some lower temperatures from the next round of G80s, which would be a blessing and a half - current models are not unknown
to hit a blistering 90C under load.
The second blessing will come with cost - the die shrink means that there will be quite a few more chips per wafer. On the 90nm process, NVIDIA gets about 80 chips gross (meaning before defective chips are thrown out) out of each wafer. The move down to 80nm should increase this turnout to roughly 90 chips, a 12.5% increase. The increased yield will allow NVIDIA to drop the cost of the chips a bit, which we will hopefully see filter into the marketplace sooner than later.
As mentioned, the order expects that the chips themselves should be in production by March. Could we be seeing an April or May price drop for the G80? That would put it just behind (but still close to)
the tentative March 2007 release date of ATI's R600.
We'll keep you informed as we know more. In the meantime, let us know your thoughts
in our forums.
Look at the current VAT free prices if you want an indication od the price drops we can expect and it isn't likely to be far wrong imo.
Can I say ****?
P.S. I'm back from being ban for a year.
Hmmm...when you say that, I hope you remember to group NVidia and ATI (or, the ATI branch of AMD) in that same group. IIRC, rumor has it that ATI's R600 GPU is going to debut with a power draw even higher than NVidia's regular 8800GTX.
I don't think either GPU company intends to work too hard on meeting lower power requirements any time soon...
PS: welcome back from your ban...based on your use of ****'s, I'm guessing you were banned for language?
http://forums.bit-tech.net/showthread.php?t=118823 always respect people here.... it can get you a long way.
Anyway I'm thinking the 80nm versions of the G80 is going to be the G81's or 8900's
I hope so , that way i can immediatly buy one of those :D
Otherwise it'l be like when i had mt 7800 and the 7900 came out , i just had to have one :p
It's rumor that the r600 has an updated frame buffer, that's include into the xbox 360 gpu.
from now on please behave and obey the moderators.
I really don't think you can point the finger at Nvidia only. As mentioned above, ATI are just as guilty of upping the power reqs of gpu's to extraordinary levels.
I mean, we now have 2000 watt power supplies! Madness... running two machines are home and a laptop, I have noticed my electric bills higher this year than last.
/sigh -- I want mini components, not massive!
Not sure why that has anything todo with power consumption though Tim?
Interestingly, although domestic prices have shot through the roof, oil prices have actually dipped recently. We should see a knock-on in a few months, because suppliers buy their stock on the wholesale market in blocks a fair few months in advance.
I defy you to spec up a desktop system that will tax even a 1kW PSU. Even with an overclocked QX6700 (at 130W stock, you might push that up to 200W with overclocking), a couple of 8800GTX in Sli (again, with overclocking, you might push the 185W stock power draw up to a max of say 250W. Multiply by two and you've got 500W). The rest of your components, assuming a couple of Raptor X in RAID 0, a couple of big drives for storage, two optical drives, a sound card etc., MIGHT just make up another 100W peak. IF you could find a way to push it all to the max at the same time (something like the bit-tech stress test, nothing you would actually DO day to day), you might be able to push 800W.
Now scale back to what 'normal' enthusiasts are using - a high spec C2D processor, overclocked drawing ~100W, a single high end graphics card drawing maybe 200W, a couple of drives etc., and you're looking at comfortably within 500W.
So with the current trend of G80 being Max 185W will that rating drop? It will still affect my overall systems power consumption if I don't use all that 185W because idle power draw will still be quite a lot due to inefficiency.
Will the 80nm also lead to a smaller PCB due to higher efficiency and a smaller cooler needed?
So essentially (after asking questions like the above to myself) What will the new G80s TDP be?
Hmm...I'm not sure how the world oil market works...but here in the US, we've been hearing that oil is DOWN near $50 a barrel recently since the midwest is getting much warmer winter weather than usual (and thus need less "heating oil", so demand is low). I can remember some months back (perhaps over the summer) when the oil prices were listed at over $70/barrel on the news here...
my x1950xtx is a hot gpu, not going over about 75*C on stock, but 90 is just bad... lol