PDA

View Full Version : I am REALLY hoping Nvidia wows us with their drivers



perkam
03-26-2010, 05:37 PM
...because the numbers right now look grim.

I was one of those hoping for an 8800 type launch where everyones eyebrows are hitting the ceiling, but its far from it.

All those waiting for price drops on ATI cards are now seeing price increases instead.

I have been a fan of ATI for the longest time, but if I know one thing well, its that the only reason they have cards like the 5x00 series out there today is because Nvidia has been applying pressure since day one.

Perkam

Apokalipse
03-26-2010, 08:00 PM
if I know one thing well, its that the only reason they have cards like the 5x00 series out there today is because Nvidia has been applying pressure since day one. Or you might say that Nvidia only makes faster GPU's because ATI is being so agressive on development. Both are largely untrue.

They both need to try and develop GPU's faster to compete with each other. That's as much as is true.

Sushi Warrior
03-26-2010, 08:06 PM
It is not an amazing product, it simply ties the 5870 I would say (~15-25% better FPS, 20% more cost but hotter). It is actually surprising to see such linear FPS/$, usually high end cards make you pay a lot more for not so much FPS.

But yeah, it's really hot and draws a lot of power. No surprises there.

Apokalipse
03-26-2010, 08:09 PM
Cypress gets much better performance per watt.

I wonder how far it can overclock given that much heat output?

Sushi Warrior
03-26-2010, 08:15 PM
Cypress gets much better performance per watt.

I wonder how far it can overclock given that much heat output?

Reviews say GTX480 can go to 850/1700/4500. But I doubt that card is chilly :shrug:

And no surprise Cypress gets better FPS/W, Fermi is not built for gaming it seems :shakes:

But IMO it is the GTX470 that is the most impressive. Trades blows with the 5870 sometimes....

Serpentarius
03-26-2010, 08:17 PM
seriously ... i've read some of the reviews .... they kept comparing 480 with 5870 ........ as if 5970 doesnt exist!
wth Nvidia is doing? it's like digging his own grave ...

all the anticipation built up, now burst like a dead balloon ... man this sux

Apokalipse
03-26-2010, 08:19 PM
But IMO it is the GTX470 that is the most impressive. Trades blows with the 5870 sometimes....doesn't the 470 cost the same? and use more power?

Sushi Warrior
03-26-2010, 08:31 PM
doesn't the 470 cost the same? and use more power?

And perform 20% better at 16x10 and 19x10 with 4xAA on.... pick your poison, FPS or power bill :shakes:

ferrari_freak
03-26-2010, 08:36 PM
Well according to the Anandtech review they're looking at a $350US price point (so of course it will be a bit more in retail) putting it neatly in between the 5850 and 5870. Power consumption is right up there, just slightly lower than the 5970. Argh this is reminding me of GeForce FX.... At least this time the performance is on their side.

I don't think power consumption should be too much of a problem though, I mean everyone has been through the 4870 phase and the power consumption on those wasn't nice either.

Apokalipse
03-26-2010, 11:06 PM
And perform 20% better at 16x10 and 19x10 with 4xAA on.... pick your poison, FPS or power bill :shakes:20% better? are you sure you're talking about the 470? maybe the 480 will get those results.
Or are you saying 5870 is 20% better?

demonkevy666
03-27-2010, 06:26 AM
what is mind boggling is a few review said the card eat 500 watts at load :shocked:
what happened to the 225 watts and 300 watts ? :wth:

side note looks like that heat sink needs the nickel sanded off to copper.

merkk1
03-27-2010, 07:31 AM
Well for me the thing is do you really want to put a video card that run at 100c in you PC ? :shrug: Someone running a air cooling rig this may lead to trouble . I go with ATI just base on the temp. the card runs at . There no need to put up
with that with the hardware we have to pick from today .

Blkout
03-27-2010, 09:32 AM
Nvidia can spin the numbers and the hype as much as they want, but the sad fact is that even though on paper it appears they've recaptured the performance crown, that's not really the case when you consider ATI certainly has something faster RIGHT NOW due to the fact that they were 6 months ahead of Nvidia. So technicially, it looks like Nvidia is ahead when the truth is they're way behind.

I'm 100% certain ATI has a new product that's considerably faster right now but simply has no need to release it just yet as the price/performance/heat/power ratio of the 5800 series is still dominating Nvidia in a big way. Consumers know it too which is why the GTX 400 series isn't going to sell that well and ATI is in no rush to release the next generation just yet. Instead we'll see a mild refresh of the 5800 series with higher clocks and some slight refinements to bring the performance crown back to ATI, which is only a stop gap until the next generation. At this point, I believe Nvidia is actually one year behind ATI in development even with this paper launch of the GTX 400 series.

Nvidia went about this generation all wrong, made terrible design and engineering decisions and now it's time to pay for their mistakes, and they will. Nvidia has got to make major changes to get the power consumption and heat reduced before they can even consider getting more performance out of this design. ATI is in a much better position of not having to deal with this engineering issues and can concentrate completely on performance improvements. Nvidia hasn't been ahead of ATI since the 8800 GTX. From the 4800 series forward, ATI has been in a commanding lead as far as price/performance/heat/power ratios. Nvidia still has a LOT of ground to make up and it doesn't look good for them.

Blkout
03-27-2010, 09:41 AM
But IMO it is the GTX470 that is the most impressive. Trades blows with the 5870 sometimes....


Huh? Depending on which review you read, the 470 is going head to head with the 5850 and getting crushed in price/heat/power. The 470 isn't even close to competition for the 5870 especially when you consider it costs just as much.

Apokalipse
03-27-2010, 11:24 AM
what is mind boggling is a few review said the card eat 500 watts at load :shocked:
what happened to the 225 watts and 300 watts ? :wth:Total system load.

But still, it does use a lot of power.

Sushi Warrior
03-27-2010, 11:28 AM
Huh? Depending on which review you read, the 470 is going head to head with the 5850 and getting crushed in price/heat/power. The 470 isn't even close to competition for the 5870 especially when you consider it costs just as much.

It's 20% faster at 16x10/19x12/19x10 at 4xAA while only $50 more. Yes, it is MUCH hotter and needs more power (not actually that much, same as a GTX285. but more than a 5850 for sure) but it is hardly "crushed". I think drivers will bring Nvidia slightly ahead over the next few months.

Chickenfeed
03-27-2010, 01:15 PM
One thing I feel worth noting is GF100 appears to scale much better than Cypress when clock speeds are increased. The difference of 850-1000 on a 5870 isnt that amazing where even a moderate clock speed bump on the 470/480 seems to make a very nice difference. Now given the heat issues, I'm hessitent to say they are great overclockers however if the heat issue were removed ( aftermarket cooling / die shrink ) they will be quite nice. As a reference product though, dissapointing is the first word that comes to mind. I have no doubts that SLI overclocked 480s on water will be amazing but the said system required ( as well as the cooling ) will cost more than 4 5970s on air...

texasreefer
03-27-2010, 03:14 PM
Well the die is supposed to be 40nm, why does the chip look like a damn CPU? Am I confused...I know the die shrink refers to the fabrication process and shriking the dies produces more chips. Is there any correlation with the size of the chip and heat production?

Frag Maniac
03-27-2010, 03:42 PM
I am REALLY hoping Nvidia wows us with their driversI'm really hoping they wow us with a 32nm process. I don't think the drivers are the problem at all. Most every site says they're stable and functioning quite well considering they're basically betas. The only real problem has been 2560x1600 scaling in some games, which Nvidia claims is just driver optimizations that have yet to be implemented.

Maybe I'm being a bit too picky. After all, radical new architecture typically takes some time to perfect. I guess it's that I've held off stepping out of the P4 dinosaur age so long (since end of '07) that I've envisioned a sparkling new, state of the art rig with no glaring caveats.

It's been such a long wait, I'm tempted by the thought that waiting for another die shrink wouldn't be much longer. Though I supposed a case with a bit better ventilation (though louder noise) than I'd planned on, and bumping the GPU fan speed to 70% (which supposedly drops the peak temps to 80c within reasonable noise levels), wouldn't be such a bad compromise.

I'm also wondering if I could time the purchase of an EVGA GTX 480 well enough to make use of their upgrade program on a possible smaller die version. Such a release would no doubt be guarded by quite a bit of NDA BS though.

Well the die is supposed to be 40nm, why does the chip look like a damn CPU? Am I confused...I know the die shrink refers to the fabrication process and shriking the dies produces more chips. Is there any correlation with the size of the chip and heat production?The 400 series does use a 40nm process. There's fabrication process die size, then there's physical die size in area. You're merely confusing the two it sounds. The 40nm process refers to the thickness of the silicone layer, not the physical area of the chip. Thinner layers dissipate heat better. The physical area of the chip is determined more by the micro architecture than the fabrication process. You can only make a chip with this many features so small with existing micro architecture.

Micro architecture and fabrication process do work hand in hand to some degree though. For instance Intel's use of Hafnium for data pathways allowed them to make chips smaller and thinner. Hafnium is a superior conductor, so using it means smaller pathways that leak less current and heat, in turn allowing for smaller, thinner chips.

However as technology advances, they also pack more features into the chips. So as with Fermi, they may actually end up being bigger vs smaller. Rest assured though on an older, bigger fab process, Fermi would have been about the size of the palm of your hand vs what it ended up being.

Intel now has 32nm Hex Core CPUs as a result of Hafnium and the micro architecture it allowed (i9s), which will soon hit the retail market. IMO this is the kind of thing Nvidia should have pursued with Fermi. Perhaps they are, but it would have been MUCH nicer to see the 400 series release on a 32nm process. As is it's basically a work in progress.

You have to keep in mind though when a manufacturer is chomping at the bit to A) get back the speed crown, B) implement DX11 fully, and C) facilitate the HPC industry, with the resulting rad architecture required to do all that, acceptable temp levels becomes anything within a reasonable percentage of the GPUs max heat handling level, which is supposedly 105c. It may be much higher in TDP and much less efficient in performance per watt than competing chips, but it does do quite a bit more and within 10% of it's temp barrier when under extreme gaming loads, and that's with the stock fan setting of only 60%.

skoreanime
03-27-2010, 05:28 PM
Anyone else notice how eerily similar this situation is when Nvidia had the 8800GTX and ATi had their 2900XT?

It's not surprising, as history does repeat itself (especially in this industry). I'm keeping silent until we see a refresh. Although it's way late into the party, the ones who show up late usually gets the party going again :P

Blkout
03-27-2010, 08:25 PM
Well the die is supposed to be 40nm, why does the chip look like a damn CPU? Am I confused...I know the die shrink refers to the fabrication process and shriking the dies produces more chips. Is there any correlation with the size of the chip and heat production?

Its a 40nm fab process but it has 3 billion transistors, that's why it's so large.