The Fermi Thread - Part 3

[SoF]

I forecast saaya to come in with some infos shortly

[saaya]

How much faster does a SLI gtx 480 have to be to match quadfire 5970s. Considering the only resolution where such power shows itself is 2500*1600 where video memory will play a limiting effect and dualcard sli will scale better than quadfire, I can imagine for people with 30" monitors the gtx 480 might look like a better card in alot of ways. Would SLI gtx 480 at 1.2(5870) be match quadfire 5970 at 2500*1600 considering scaling issues with the 5970?

The power consumption looks awful those. This near 300 watts is a disaster in so many ways. I think classified editions of such cards won't be all for not anymore, as this card looks like it can use another 8 pin or something. I hope a company comes out with a custom PCB without the hole and a standard waterblock.

The refresh looks like the card to buy those the gtx 485 should be an overclockers dream, if they are indeed already hitting 700/1400 with this current revision.

gtx480 sli, gtx485... come on man, get a grip

have you guys seen this?
http://www.techpowerup.com/117897/Ge...s_Surface.html

GeForce GTX 480
shader cores 480
clocks 700/1400/3700
295W TDP
499$
5870+7%

GeForce GTX 470
shader cores 448
clocks 607/1215/3350
225W TDP
349$
5850+7%

the good:
low prices

the bad:
high tdp
low 470 clocks
low mem clocks
only 480 cores for gtx480

the ugly:
performance...

i wonder how well these scale on ln2... if they clock well on ln2 they will def break all world records

[onethreehill]

Latest update : TDP lowered to 250W for GTX 480

http://forums.vr-zone.com/7792771-post44.html

[Vardant]

there won't be any 20 percent if shader difference between 470 and 480 is 32 only imho

Are you sure?

just a try:

	GTX 480  GTX 470  470->480  470->480	480->470	
			     %      (% less)  	(% more)	
sps	     480   448	     93,33%	6,67%	7,14%	(= % tmu units)
core	     700   607	     86,71%	13,29%	15,32%	(= % rop speed)
shader	     1401  1215	     86,72%	13,28%	15,31%	(= %  tmu speed)
mem	     1848  1674	     90,58%	9,42%	10,39%	
bus width    384   320	     83,33%	16,67%	20,00%	(= %  rop units)
mb	     1536  1280	     83,33%	16,67%	20,00%	
tdp	     295   225	     76,27%	23,73%	31,11%	

pixel fillrate		     72,26%	27,74%	38,39%	
texture fillrate             80,94%	19,06%	23,54%	
flops		             80,94%	19,06%	23,54%	
bandwidth                    75,49%	24,51%	32,47%	
"average"	             77,41%	22,59%	29,49%

[Sentential]

Very very very VERY dissapointing. If it really has 480 shaders then my interest in buying these cards just took a serious nose dive. 3 spins and they cant even get all the shaders to work? Why in the world would I drop $500 for a broken card, the 470 looks better and better now especially if it holds to its $350 price tag.

[Zed_X]

GTX 480s TDP is 250W many months, that 29xW was total BS! launch date is not postponed! Nordic Hardware is wrong, launch is still 26th.

[NaMcO]

Very very very VERY dissapointing. If it really has 480 shaders then my interest in buying these cards just took a serious nose dive. 3 spins and they cant even get all the shaders to work? Why in the world would I drop $500 for a broken card, the 470 looks better and better now especially if it holds to its $350 price tag.

I don't think it's an issue with "can't make them work". More like TDP issues.

[Carfax]

Very very very VERY dissapointing. If it really has 480 shaders then my interest in buying these cards just took a serious nose dive. 3 spins and they cant even get all the shaders to work? Why in the world would I drop $500 for a broken card, the 470 looks better and better now especially if it holds to its $350 price tag.

Seriously, with all the erroneous info floating, I would be hesitant to take any of this as fact.. The Galaxy website showed the GTX 480 having 512sp after all.

And the speculation about the GTX 470/480 only being 5-10% faster than the HD 5850/5870 should be taken with a healthy dose of salt..

Just wait for the launch date.

[zalbard]

Hopefully GTX 470 will have a decent OC and not significantly lowered volts, 750/1500/1950 on stock cooler is whereabouts I hope the max clocks on stock cooler/volts would be.

I think the issue will be heat, not overclockability... And stock volts will have to be pretty low in order to keep the power consumption under control.

How much faster does a SLI gtx 480 have to be to match quadfire 5970s.

Quadfire 5970 is supposed to be A LOT faster.

GTX 480s TDP is 250W many months, that 29xW was total BS! launch date is not postponed! Nordic Hardware is wrong, launch is still 26th.

250W, hm, not so bad I guess, for 3B transistors...

[bfar]

Looking forward to next week!
Haven't decided what to do yet... based on rumours I'm looking at b) or c), but here's hoping they're wrong!

a) Get a GF100 depending on price/perf
b) Get a 58xx if there's a price drop
c) Hang onto my gtx280 and wait until July/Aug for refresh or price drops

Big question - does ATI have a Cypress refresh waiting in ambush?

[Vardant]

Quadfire 5970 is supposed to be A LOT faster.

It could have changed greatly, but it was only 35% faster. I actually think that HD 5870 in Tri-Fire is a better choice.

[saaya]

a 470 at 480 clocks is how much slower than a 480?
shouldnt be much, right?

the 470 doesnt sound bad at all for benching, but on air... 5850s and 5870s clock pretty well on air...

gf100 is not going to make a lot of people switch from ati to nvidia, but for nvidia fans and benchers a 470 should be a very nice card

[Sentential]

Seriously, with all the erroneous info floating, I would be hesitant to take any of this as fact.. The Galaxy website showed the GTX 480 having 512sp after all.

And the speculation about the GTX 470/480 only being 5-10% faster than the HD 5850/5870 should be taken with a healthy dose of salt..

Just wait for the launch date.

I would agree wholly however what I am concerned by is the long-term rammifactions of these chips. If they are broken now how will their long term reliability going to be 1 year or even 2 years after purchase? People's 285GTXs are dropping left and right for thermal related bump issues and there weren't nearly as many respins required for that silicon to be deemed acceptable. How are we to believe that this given how many respins it took; that this round going to be any better. This chip just feels like all kinds of "half-assed" and Im not sure how comfortable I feel buying either of these cards.

If ATi drops the price of the HD5870 down to the $300-350 I am likely to jump ship; *especially* if Thuban only works on an AMD chipset which by all accounts could happen.

[annihilat0r]

470 should be a good card but its TDP is still way high. For performance lower than 5870 it uses up to 30W more.

MSRP's look good, though.

[Picao84]

470 should be a good card but its TDP is still way high. For performance lower than 5870 it uses up to 30W more.

MSRP's look good, though.

And price is really what matters no? How many people look at power? 30W more is peanuts, comparing to one TV or a Microwave :p

[saaya]

yeah i dont think the tdp matters for power consumption... tdp only matters indirectly by reducing the clocks...
the 1950pro was a very popular card as well and it had a huge power consumption... but it was priced well and performed well...

[cegras]

And price is really what matters no? How many people look at power? 30W more is peanuts, comparing to one TV or a Microwave :p

It's not peanuts when you're considering your power supply. I'd wager that's why people are mad.

[annihilat0r]

And price is really what matters no? How many people look at power? 30W more is peanuts, comparing to one TV or a Microwave :p

Yeah, from a practical perspective this is correct. However, more power for less performance doesn't feel right and it's technically ineffective.

[Picao84]

It's not peanuts when you're considering your power supply. I'd wager that's why people are mad.

We are talking about GTX470 here, which rumour says it has a 225W tdp. I dont think anyone who buys cards in this range will have to update the power supply...

[Picao84]

Yeah, from a practical perspective this is correct. However, more power for less performance doesn't feel right and it's technically ineffective.

Yes you are right. But in the end, it will be the bottom line that matters. If price is good, and performance fits the price, it will be bought. On the other hand, if ATI lowers prices (I really dont think it will) things can change.

[eric66]

Are you sure?

yeah i am sure and that weird graph won't change anything

[MrDiaz]

You know what's sad? People will still buy these cards even if:

1 - Power consumption is off the charts
2 - There is < 10% performance increase over the 5xxxx series
3 - More expensive than the 5xxxx series

/shrug

[ryboto]

yeah i dont think the tdp matters for power consumption... tdp only matters indirectly by reducing the clocks...
the 1950pro was a very popular card as well and it had a huge power consumption... but it was priced well and performed well...

I think you're confusing cards here, the pro was on a newer process(80nm) if I recall vs 90nm for the XTX's. It actually had very low power draw, I should know, I owned one.

[Leeghoofd]

Not all users buy cards coz they are the fastest in games or benches...
Some are loyal to one brand, no matter what. Some prefer to stick with a brand coz they personally never had any issues with it. Some do other things with videocards then game on them...

[MrDiaz]

Not all users buy cards coz they are the fastest in games or benches...
Some are loyal to one brand, no matter what. Some prefer to stick with a brand coz they personally never had any issues with it. Some do other things with videocards then game on them...

I realize that... which is why i said it's sad.