The Fermi Thread - Part 3

[saaya]

ive always spend most of my money on the monitor. 150$ gpu is good enough to enjoy games, 100$ cpu, and everything else under 100$. my monitor is worth 750$, i got a IPS back when i had no idea that was good, i just wanted a 27" 1920x1200 monitor, and they were all about 1000$ then.

wow, 27" 1080p ips sounds veeery nice... thats a great idea... you get a huge screen without a massive pixel count youd need multiple gpus for to drive...

it was like a whole new computer going from 1280x1024 to that, and just last year i was testing a 30" 2560x1600 for a friend who was getting it for xmas, and so i got to check for dead pixels by using it for a few weeks. and i really did not want to give it up. in WoW that 40% extra res really makes a life of difference, and cause the dot pitch is smaller i can pull it closer and have to turn my head to see the edges of the screen.

hahah, but why would you move your head that close?
idk how big the pixels are on your 27", but on my 23" i need to get REALLY close to see them

Fallout3 was stuttering to hell, but ive never felt so immersed. would i pay 1000$ for one, no, but if they came down to 600$ i would have jumped on it quick. but honestly now with multi monitor setups, im just waiting for a no bezel, or practically no bezel screen to appear, and jump on 3 of them for some 3600x1920 goodness, for hopefully under 800$

i share that dream with you... but i think its just a dream... especially the 800$ part ^^

i think if youd get a 30" screen youd regret it after some time... cause youll be doomed to life at the the edge of plabale fps or having to play below native res
and everytime a new killer graphics game comes out you will probably have to upgrade...

but i totally agree, upgrading the display is a much much much better investment and makes a much bigger difference than upgrading the vga, cpu or memory... id say the most useful upgrade right now is an ssd, after that a big display and after that a nice vga...

[Chickenfeed]

i think if youd get a 30" screen youd regret it after some time... cause youll be doomed to life at the the edge of plabale fps or having to play below native res
and everytime a new killer graphics game comes out you will probably have to upgrade...

Totally agree I wish someone would put out a nice 1080P panel for a change though ( everything is TN now days... ) I might end up using an Samsung LED for a monitor at this rate

[Philip_J_Fry]

Am I the only person who wants to see more GTX 295 Vs GTX 480 comparisons?
The GTX 480 is a strange card, yes it is rather powerful and certainly not an Epic fail, however it is that late to the party most of the guests are leaving or quite intoxicated and have fallen asleep.
IMHO this card has December 2009 wrote all over it. Impressive nether the less, but falls short of greatness.
The temps and fan noise are too much, a GTX 295 Single PCB is a shade quieter and cooler, for a dual GPU card that is some achievement!
I hope the GTX 480 Ultra or GTX 512 coming in Q3 has a revised cooling solution and improved fabrication process to bring the temps down.
John

lol agree, i was expecting more.. like more than 30% faster than 5870 in crysis, bf 2 bc, Aliens vs. Predator @ 1080p with or without aa for the heat its generating and ~ 6 months after cypress

the price is reasonable i will say , ~ USD $100 more than 5870. before price gouging that is heh

i think i will wait for both camps' die shrink

[***Deimos***]

http://hwbot.org/hardware/videocard/geforce_gtx_480

An unvoltmodded 800 MHz stable on air is quite nice for an Nvidia card, no doing that on a GTX285.

1165 MHz and 30K vantage under LN2 is also very nice, I'd like to see some more results with a overclocked CPU.

nVidia was stuck in a narrow 75Mhz zone, between 574Mhz - 651Mhz, for many years. (From 7900 GTX, to 8800 GTX, 8800 ULTRA, 9800 GTX, GTX280 all the way up to GTX285).

Until a couple days ago the only exception was 9800GTX+/GTS250. Even the 40nm GT240 clocks in at meagre 550Mhz.

On air, 800. With better cooling 900. Perhaps they use higher voltage than necessary to compensate for TSMC channel length variance. They could use various tricks around it, and TMSC could improve. With tighter control and reduced voltage near 200W might be possible (at expence of little OC headroom). But, with this 6 month delya, what if they already used up this "4890" trick?

@TweakTown review. how does a 10% overclock get 30% performance boost on HAWX?

Dunno. But they aren't just changing core clock - mem too.

Am I the only person who wants to see more GTX 295 Vs GTX 480 comparisons?
The GTX 480 is a strange card, yes it is rather powerful and certainly not an Epic fail

At launch I recall, GTX280 was < 9800GX2. In contrast, Fermi is consistently > GTX295. Performance, especially DX11, (in my mind) is outstanding.

The *only* problem with Fermi is power/heat. If they "fixed" it before maybe they can do it again:
GTX280 600Mhz 230W "close" to 250W Fermi - really more like 300W.
GTX285 650Mhz 180W

5870 scales poorly with clocks. At 950 or even 1000Mhz, it wont make up Fermi's 25-30% lead in DX11+AA. So a 200W, 750-800Mhz "GTX485", would just extend that lead.

Problem with GTX280 analogy is that 55nm was available and used by AMD for 1.5 years... 40nm is already the half-node shrink of 45nm.

[highoctane]

I wonder what impact Nvidia implementing a short of hyperthreading/hts/gigathread scheduler had on its performance, such as the strong min fps performance.

[Manicdan]

hahah, but why would you move your head that close?
idk how big the pixels are on your 27", but on my 23" i need to get REALLY close to see them

the dot pitch makes a world of difference when ur reading web pages at native res, or anything with text really. in WoW i had to up the size of the chat log, or pull the screen almost a foot closer. yeah i got the 27" cause it was bigger than the 24" options, for about the same price. since the 30" was just on borrow, i still use the 27" after over 3 years. infact its so old it only has 1 dvi input and a vga, lol. and the monitor i had before then was only vga. it really was worth every penny, and IPS prices took a very long time to drop, so i know its still worth a good 400-500$ today (if its still sold). my 400$ 2900 with 150$ waterblock was probably the most expensive thing i bought that fell like a brick.

[saaya]

At launch I recall, GTX280 was < 9800GX2. In contrast, Fermi is consistently > GTX295. Performance, especially DX11, (in my mind) is outstanding.

o rly?
most of these are old games, but still...
its interesting to see that almost 2.5x the hw results in less than 2x the performance... maybe nvidia castrated fermis texturing power too much?

gtx480 up to 20% slower than 295 in cod4
http://www.techpowerup.com/reviews/N...0_Fermi/7.html
gtx480 10% slower than 295 in qw
http://www.techpowerup.com/reviews/N..._Fermi/13.html
gtx480 15% slower than 295 in fc
http://www.techpowerup.com/reviews/N..._Fermi/14.html
gtx480 10% slower than 295 in fc2
http://www.techpowerup.com/reviews/N..._Fermi/15.html
gtx480 20% slower than 295 in prey
http://www.techpowerup.com/reviews/N..._Fermi/17.html
gtx480 10% slower than 295 in q4
http://www.techpowerup.com/reviews/N..._Fermi/18.html
gtx480 20% slower than 295 in ut3
http://www.techpowerup.com/reviews/N..._Fermi/22.html

[damha]

nVidia was stuck in a narrow 75Mhz zone, between 574Mhz - 651Mhz, for many years. (From 7900 GTX, to 8800 GTX, 8800 ULTRA, 9800 GTX, GTX280 all the way up to GTX285).

Until a couple days ago the only exception was 9800GTX+/GTS250. Even the 40nm GT240 clocks in at meagre 550Mhz.

On air, 800. With better cooling 900. Perhaps they use higher voltage than necessary to compensate for TSMC channel length variance.



Dunno. But they aren't just changing core clock - mem too.



At launch I recall, GTX280 was < 9800GX2. In contrast, Fermi is consistently > GTX295. Performance, especially DX11, (in my mind) is outstanding.

The *only* problem with Fermi is power/heat. If they "fixed" it before maybe they can do it again:
GTX280 600Mhz 230W "close" to 250W Fermi - really more like 300W.
GTX285 650Mhz 180W

5870 scales poorly with clocks. At 950 or even 1000Mhz, it wont make up Fermi's 25-30% lead in DX11+AA. So a 200W, 750-800Mhz "GTX485", would just extend that lead.

Problem with GTX280 analogy is that 55nm was available and used by AMD for 1.5 years... 40nm is already the half-node shrink of 45nm.

You buy a GTX480 first!

And live with:
- $200-$300 premium over the 5870
- lower oc (if you oc)
- super hot case
- A family complaining about the noise levels in the house, your PC is competing with the vacuum cleaner
- at least $7/month additional cost on your monthly bill if you game an average 2 hours/day and your cost is $0.8/kWHr (for us it varies from 0.7 to 0.9) (check here for your state. If you live in NYC, OWNED)
- If you don't have a 1000W psu, another $200 in cost
- No oc headroom if you don't get a waterblock or don't water cool (water block? +$200)
- High chance of failure. If the GPU runs at 100°C I can't imagine what the VRMs are running at. I wish someone tested this card with an infrared scanner. There is no way these cards will not have high failure rates, logic + history have proven this to us
- Poorer AF quality (you can claim it is not noticeable, doesn't change the facts)
- PhysX/CUDA
- 10-15% faster performance than the 5870

If I get a 5870, I live with:
- No CUDA/PhysX
- 10-15% slower performance

All I can say is, if that satisfies you then by all means do not let anyone stop you from experiencing the greatness of the 480.

[saaya]

You buy a GTX480 first!

And live with:
- $200-$300 premium over the 5870
- lower oc (if you oc)
- super hot case
- A family complaining about the noise levels in the house, your PC is competing with the vacuum cleaner
- at least $7/month additional cost on your monthly bill if you game an average 2 hours/day and your cost is $0.8/kWHr (for us it varies from 0.7 to 0.9) (check here for your state. If you live in NYC, OWNED)
- If you don't have a 1000W psu, another $200 in cost
- No oc headroom if you don't get a waterblock or don't water cool (water block? +$200)
- High chance of failure. If the GPU runs at 100°C I can't imagine what the VRMs are running at. I wish someone tested this card with an infrared scanner. There is no way these cards will not have high failure rates, logic + history have proven this to us
- Poorer AF quality (you can claim it is not noticeable, doesn't change the facts)
- PhysX/CUDA
- 10-15% faster performance than the 5870

If I get a 5870, I live with:
- No CUDA/PhysX
- 10-15% slower performance

All I can say is, if that satisfies you then by all means do not let anyone stop you from experiencing the greatness of the 480.

its supposed to be 100$ more, but lets see... i dont think itll be 300$ though...
and fermi seems to scale very nicely when oced... from what ive seen a fermi pushed to max on air vs 5870 pushed to max on air, the fermi actually gains on the 5870!

[Coldon]

http://www.newegg.com/Product/Produc...gtx480&x=0&y=0

http://www.newegg.com/Product/Produc...D+5870&x=0&y=0

round about 80~90 dollars difference between 5870 and 480 on newegg, no stock obviously. Right now these pointless arguments are getting ridiculous and really pedantic, wait two weeks then we'll see what the OC of the cards is like once a lot more guys get hold of the cards, also lets see what more mature drivers can do, it is a completely new architecture. The argument that nvidia had 6 months to improve drivers is silly cause then you could easily say that ATI had plenty of time before launch as well, and siz months down the line even the 10.3s aren't perfect...

Whereas i was going to buy a 480, I'm now gonna wait around a month and a half and see what happens with prices and drivers

[Helloworld_98]

its supposed to be 100$ more, but lets see... i dont think itll be 300$ though...!

if you look at UK prices, cheapest 5870 is £300 and the most expensive GTX 480 is ~£477, which the difference is equal to $264, or $310 if you count the cheapest 5870 as the one on amazon.co.uk which is never in stock

[zalbard]

The only "IDIOT" and "not a real gamer" here is you. Downgrade 30" 2560x1600? You should tell that to people also that have 120"-144" screens for home theater that have 1080P they should downgrade to 50-60" screens so that 1080P looks sharper....

Ever think that maybe people with 30" 2560x1600 can afford multi-gpu solutions, BTW please explain and show me when 7900 GTX SLI, 8800 GTX SLI, GTX 280 SLI or GTX 480 SLI has "stutter" problems? Huge power consumption, I can afford $50 extra a year for power, noise? with home theater sound or headphones people aren't listening to their fans spinning up while gaming.

It is clear you don't have a 30" 2560x1600 display, but calling people idiots or not real gamers really shows a lack of intelligence.

I'm sorry, Saaya, but I agree with him. Your comment was inaccurate and totally unnecessary. I, for one, would love to upgrade to 2560x1600 screen to game on. And I game a lot.

How loud is GTX470? (video)

WTF? This is a joke, right?

Fermi on Atom Platform

Yeah, awesome idea... NOT!

if you look at UK prices, cheapest 5870 is £300 and the most expensive GTX 480 is ~£477, which the difference is equal to $264, or $310 if you count the cheapest 5870 as the one on amazon.co.uk which is never in stock

When is the next batch of Fermi coming? They only shipped 5000 cards, didn't they? Wonder when the prices are going to go down...

[Manicdan]

http://www.newegg.com/Product/Produc...gtx480&x=0&y=0

http://www.newegg.com/Product/Produc...D+5870&x=0&y=0

round about 80~90 dollars difference between 5870 and 480 on newegg, no stock obviously.

the fact they are sold out at 500$ shows one or two things, either low stock, or the price wasnt too high. so even if nvidia laid down a high launch price, it was the right choice, and who knows how quickly that will come down.

[saaya]

also lets see what more mature drivers can do, it is a completely new architecture. The argument that nvidia had 6 months to improve drivers is silly cause then you could easily say that ATI had plenty of time before launch as well, and siz months down the line even the 10.3s aren't perfect...

and the 197 drivers for the gt200 based cards are perfect? theres no such thing as a perfect driver with constantly new hw and sw being introduced into the eco system...
i think drivers for gf100 will improve, but not all that much... maybe 5% on average... cards that come out with a big delay usually already have the drivers tweaked for performance...

if you look at UK prices, cheapest 5870 is £300 and the most expensive GTX 480 is ~£477, which the difference is equal to $264, or $310 if you count the cheapest 5870 as the one on amazon.co.uk which is never in stock

yeah but comparing the cheapest 5870 with the most expensive gtx480 isnt exactly fair, is it?
the uk is the uk, theres a reason its called the ripp off island even by locals
but i dont think there will be a constant 300$ difference of 480 and 5870...maybe 200$... but even that would surprise me... why would they ripp off nvidia users more than ati users?

I'm sorry, Saaya, but I agree with him. Your comment was inaccurate and totally unnecessary. I, for one, would love to upgrade to 2560x1600 screen to game on. And I game a lot.

heh, go and try it... go and try it... and then lets talk again...
ive played on a 2560x1600 demo rig and i found it very frustrating to have this awesome screen i just couldnt feed with gpu horse power... idk, its been a while so maybe things have improved by now... but from what i see in reviews 2560x1600 perf is usually still in the 30s fps wise if not less... and thats average fps...

so wait, gf100 is pretty much sold out already? wtf? must have been very few samples then...
didnt somebody here claim 60.000 cards within the first weeks after launch?
and those supposedly sold out in a day?
and in this entire thread only 3 or 4 people managed to buy one?
something smells about this...

[fiskov]

I don't see why the "High chance of Failure" is a problem..
BUY NOW.. Get a great card.
If it fails?
Most likely your get the revision.. So really, it's a win/win situation.

And this "fan noise" is so exaggerated;

Fan Noise

Others have also said it's no louder than a 280 GTX, which is no where near as loud as all this bashing is making it out to be.

In the end i managed to pre-order the ASUS 480 GTX @ £418.57 inc. shipping, compared the the cheapest HD 5870 @ £295.87

Which is only £122.70 difference.. I would honestly pay that little extra to avoid CCC and hotfixes for every new game release alone!

[RSC]

if you look at UK prices, cheapest 5870 is £300 and the most expensive GTX 480 is ~£477, which the difference is equal to $264, or $310 if you count the cheapest 5870 as the one on amazon.co.uk which is never in stock

That's really fair...

[zalbard]

didnt somebody here claim 60.000 cards within the first weeks after launch?

Not 5000?

Rumors abound that Nvidia will only have 5,000 to 8,000 Fermi GF100s, branded GTX480 in the first run of cards. The number SemiAccurate has heard directly is a less specific 'under 10,000'. There will have been about two months of production by the time those launch in late March, and Nvidia bought 9,000 risk wafers late last year. Presumably those will be used for the first run. With 104 die candidates per wafer, 9,000 wafers means 936K chips.

[Manicdan]

yeah but comparing the cheapest 5870 with the most expensive gtx480 isnt exactly fair, is it?
the uk is the uk, theres a reason its called the ripp off island even by locals
but i dont think there will be a constant 300$ difference of 480 and 5870...maybe 200$... but even that would surprise me... why would they ripp off nvidia users more than ati users?

the top 480s already have waterblocks, thats a good 100$ preimium, it seems that 75-100$ separates the two

heh, go and try it... go and try it... and then lets talk again...
ive played on a 2560x1600 demo rig and i found it very frustrating to have this awesome screen i just couldnt feed with gpu horse power... idk, its been a while so maybe things have improved by now... but from what i see in reviews 2560x1600 perf is usually still in the 30s fps wise if not less... and thats average fps...

the same guy who got the 30" just had me rebuild his machine, a phenom quad 3.4ghz, 4GBs of ddr3, and a 5850, plays CoD MW2 maxed out, and i mean maxed the F out, as smooth as can be.

[bill_d]

heh, go and try it... go and try it... and then lets talk again...
ive played on a 2560x1600 demo rig and i found it very frustrating to have this awesome screen i just couldnt feed with gpu horse power... idk, its been a while so maybe things have improved by now... but from what i see in reviews 2560x1600 perf is usually still in the 30s fps wise if not less... and thats average fps...

at 8xaa maybe but at 25x16 2xaa i can get 60fps on most all game with two 4890's a few game i can get 4xaa

[Helloworld_98]

That's really fair...

the guy who said $200-300 difference never put any rules down on it other than it was between a 5870 and a GTX 480.

[RSC]

the guy who said $200-300 difference never put any rules down on it other than it was between a 5870 and a GTX 480.

How about common sense? It's usually a very good rule...

[Coldon]

and the 197 drivers for the gt200 based cards are perfect? theres no such thing as a perfect driver with constantly new hw and sw being introduced into the eco system...
i think drivers for gf100 will improve, but not all that much... maybe 5% on average... cards that come out with a big delay usually already have the drivers tweaked for performance...

I never said they were perfect did i? 5%? strange how the new ATI drivers gave a good ~10% increase across the board for an old architecture. You could then argue as to why these improvements weren't in the drivers at release since its an old architecture, they have technically had years to perfect and tweak it, and yet the launch drivers for the 5870s were horrendous.

This is a completely new architecture, and you expect them to have drivers optimized in 6 months?! how long did it take for the g80 drivers to get tweaked and optimized, the move to USM/USA needed a lot of driver tweaks and optimization. The same can be said here as well, though not to such a great extent as the g70 -> g80 change over.

I'm starting to feel that you are just arguing and disagreeing for the sake of disagreeing and voicing your opinion. Some people might like larger displays simply cause they are older and their eyes get tired staring at smaller displays, or they prefer sitting further back from their screens. Yeh, SLI/crossfire is not a great gaming experience but they can simply turn off some eye candy on a single GPU and still have excellent framerates. High levels of AA at such a high resolution is not as essential as it is at lets say 1680x1050.

[tajoh111]

I never said they were perfect did i? 5%? strange how the new ATI drivers gave a good ~10% increase across the board for an old architecture. You could then argue as to why these improvements weren't in the drivers at release since its an old architecture, they have technically had years to perfect and tweak it, and yet the launch drivers for the 5870s were horrendous.

This is a completely new architecture, and you expect them to have drivers optimized in 6 months?! how long did it take for the g80 drivers to get tweaked and optimized, the move to USM/USA needed a lot of driver tweaks and optimization. The same can be said here as well, though not to such a great extent as the g70 -> g80 change over.

I'm starting to feel that you are just arguing and disagreeing for the sake of disagreeing and voicing your opinion. Some people might like larger displays simply cause they are older and their eyes get tired staring at smaller displays, or they prefer sitting further back from their screens. Yeh, SLI/crossfire is not a great gaming experience but they can simply turn off some eye candy on a single GPU and still have excellent framerates. High levels of AA at such a high resolution is not as essential as it is at lets say 1680x1050.

+ 1 to both your points.

Fermi has inconsistent performance across the board. Games the old architecture was good at perform comparatively bad on fermi. Its a new architecture and whatever people may say. 197.xx still is primarily focused in bringing higher performance pre fermi architecture. 58xx driver optimizations apply to 4xxx because the architecture are super similar and the drivers should be super similar. 5% is worse than even charlie would suspect even.

Additionally size and resolution matter. Bigger size needs higher resolution to look good. If gameplay framerates were everything, eyefinity would be crap because eyefinity penalizes framerates even more and in addition has driver issues. More power to fuel higher resolutions is one of the fundamental reason why videocards have to get more powerful.

[damha]

More power to fuel higher resolutions is one of the fundamental reason why videocards have to get more powerful.

Let's start taking bets: ATI's next generation GPU will be more powerful and run cooler than the GTX480.

Is that a far-fetched assumption?

Video cards will get more powerful, but not Nvidia's version of powerful.

[LiquidFiction]

Let's start taking bets: ATI's next generation GPU will be more powerful and run cooler than the GTX480.

Is that a far-fetched assumption?

Video cards will get more powerful, but not Nvidia's version of powerful.

nVidia will have to sell the company by 2011 Q1