The Fermi Thread - Part 3

[GoldenTiger]

Why not?

Since you posted those benches I assume that's the games you play. Out of 7 it's a hell lot faster at six of'em (it's almost a tie in the single one the 480 beats the 5970), it consumes less power, it doesn't sound like 747 taking off and it was available in November 2009...

The 5970 also happened to be chronically unavailable, $700, and dependent on dual-GPU profiles. I've said it in the past and I'll say it now... I do not like multi-GPU setups and avoid them as much as possible.

I stated a few times in this thread that hardwarecanucks lost my respect when they gave out results in the nvidia 'big bang 2' that contradicted everyone else, then argued that the games they used were not given to them by nvidia.

Anandtech blew that out of the water when they said nvidia tried to force those games on reviews.

Something like that anyway, my old failing memory sometimes gets the better of me.

So the new fermie results they are showing, don't help matter in my trust in them.

Fair enough, but the other reviews in the chart show similar results. Are you trying to say that ALL those sites are not credible? I know you're not, but the pattern matches... even if the numbers are a couple percent different either way. Thus we can assume they are properly done.

[Motiv]

Fair enough, but the other reviews in the chart show similar results. Are you trying to say that ALL those sites are not credible?

It depends what you say are similar. Having a greater score of between 5-10% compared with all the others, makes me wonder how their testing differs.

[GoldenTiger]

It depends what you say are similar. Having a greater score of between 5-10% compared with all the others, makes me wonder how their testing differs.

What I mean is, that even though the numbers vary ~5% or so (which is in margin of error for manual run-throughs and various setups), the pattern is the same: game A performs roughly X% better, game B performs roughly Y% better, etc. across all the reviews. For example, Far Cry 2 is a clear, large winner across reviews; Crysis is a very small bit better across them. We can assume that if a review fits inside of the pattern on how its games are performing, that it's pretty close if not spot-on. If for instance some review showed Crysis as 30% better when most of them are showing it as around 3-5%, we'd know something was up.

[insurgent]

What I mean is, that even though the numbers vary ~5% or so (which is in margin of error for manual run-throughs and various setups), the pattern is the same: game A performs roughly X% better, game B performs roughly Y% better, etc. across all the reviews. For example, Far Cry 2 is a clear, large winner across reviews; Crysis is a small bit better across them. We can assume that if a review fits inside of the pattern on how its games are performing, that it's pretty close if not spot-on.

Of course it performs better, it was designed that way.. but it doesn't perform far better as you would say, it's just okay despite coming 6 months later. ATI and NV took different paths and currently I prefer ATI's way. It's efficient.

[zalbard]

eVGA has 725Mhz superclocked edition - how brave of them to increase clocks by 3.5%.

GTX480 - melt an igloo in seconds
GTX480 - helping with global warming has never been easier!
GTX480 - when you see the juices on the burger, its time to flip one more time and you're done.
GTX480 - We finally beat the freezing aliens of Crysis!
GTX480 - nVidia - helping teenagers fake fevers since 1999.
GTX480 - hey baby, its hot in here.. why dont you take off that bra *wink*
GTX480 - Its time to kick as and chew bubble gum, and stop whining about a little heat - I've been to hell and back.
GTX480 - may cause dizziness, blurred vision, dry mouth, dehydration, shortness of breath, headaches, naussea, explosive diahrea

Haha, you crack me up!
Hope you're not actually serious, though.

Then came the part of the show I was having a hard time keeping a straight face. To promote the GTX 480 and 470 Nvidia came up with a new slogan. and ... wait for it ... here it is!

Oh my God...

[Motiv]

What I mean is, that even though the numbers vary ~5% or so (which is in margin of error for manual run-throughs and various setups), the pattern is the same: game A performs roughly X% better, game B performs roughly Y% better, etc. across all the reviews. For example, Far Cry 2 is a clear, large winner across reviews; Crysis is a very small bit better across them. We can assume that if a review fits inside of the pattern on how its games are performing, that it's pretty close if not spot-on. If for instance some review showed Crysis as 30% better when most of them are showing it as around 3-5%, we'd know something was up.

It ends up 11% different to anand/hexus, that's a fairly substantial margin of error there.

the only cause for this which someone has mentioned is they ran at 4ghz, which if this is the case, means the fermie is better at higher speeds.

But to blanket say 'it's margin of error' is taking the piss a little.

[GoldenTiger]

It ends up 11% different to anand/hexus, that's a fairly substantial margin of error there.

the only cause for this which someone has mentioned is they ran at 4ghz, which if this is the case, means the fermie is better at higher speeds.

But to blanket say 'it's margin of error' is taking the piss a little.

I'm not calling it margin of error, I'm calling it within the pattern other reviews set for relative performance differences.

[Mk]

Still not seeing the "massive fail" with 1920x1200 or 1920x1080, plus 4x AA, or 2560x1600 4x AA (16x AF). Minimum framerates 20-30% higher, average 15-20% across most reviews. Some higher, some lower... . The main well-done review belonging to Hardware Canucks shows this very clearly. $400 vs. 500, is about 25% difference in price, not that out of hand for a top-end single-GPU card. Yes, it's hot... yes it draws a lot of power, but no it isn't a "massive fail".

/agree

[Olivon]

Fermi arch' is quite good but it's too big, too complex for 40 nm. 28 nm will be more interesting for it IMHO.

[Caparroz]

(...)I've said it in the past and I'll say it now... I do not like multi-GPU setups and avoid them as much as possible.(...)

(...)I'm *this* close to selling my 280 and going for the 295(...)

Nevermind me, I'm just teasing you. I'm piss drunk DL'ing a GIGANTIC file from a client just to email the bastard to say "yep, I got it". Heck, in this very thread I said that Fermi was a flop and probably will be SLI'ing two of them.

To tell you the truth I think the gtx480 is great card if you h2o it.

[GoldenTiger]

Nevermind me, I'm just teasing you. I'm piss drunk DL'ing a GIGANTIC file from a client just to email the bastard to say "yep, I got it". Heck, in this very thread I said that Fermi was a flop and probably will be SLI'ing two of them.

To tell you the truth I think the gtx480 is great card if you h2o it.

Yeah, I never did end up going for it but I was tempted at the time .

[tifosi]

Posted this in the benchmarks thread as well, but this is what i could gather from reading Anandtech's review of Fermi...

THE GOOD:
- Fastest single GPU
- Size, only 9.5 inches

THE BAD:
- Support for a maximum of 2 monitors only
- HDMI audio still not at par with AMD/ ATI. HTPC users must take note of this
- Cost of the card

THE UGLY:
- 95-96 degree Centigrade at load. HTPC users must take note of this
- Noise. HTPC users must take note of this
- Power consumption in idle is more than a pair of 5870's CF'd idling
- Power consumption at load is about 100W-150W more than 5870 (diff people have quoted diff figures)
- 250W TDP for GTX 480 (with only 480 shaders). Charlie is right with his 295W TDP figure, though that must be for the 512 shaders part.

Then again, AMD/ ATI has covered some ground with Stereo 3D support now included with the 10.3 drivers... so as far as i as a normal user is concerned, i'd rather have a 5870 and if i feel the itch for more power... two of them 5850's OC'd would be more than handy in the price range to kill/ negate any performance advantage of Fermi... they still wouldn't suck as much power as this big mutha does...

Another big disadvantage which is noticed as per my usage was the lackluster support for the audio over HDMI... AMD/ ATI is doing it... now also supporting bit-streaming with their 5 series cards since 6 months or so. Why could or rather wouldn't Nvidia cover this? Oh i do not only game, but watch a lot of hd flicks with hd audio tracks on my pc... which a lot of people do and they should note this.

All in all, i can't seemingly justify a purchase of a GTX480... so i'll catch hold of a sucker who wants to sell two 5870's (yes i know one such guy :P )... will get them both for about or less than $500, and under warranty...

[Snow Crash]

Crank that S#!T UP ...
reminds me of the ads for Frank's RedHot Sauce where the little old lady says "I put that S#!T on eveything!"

http://www.franksredhot.com/

[dogchainx]

This thread has turned from okay, to bad, to a ***king joke. I might as well have joined in on a conversation of who can piss the furthest. Reviews are out, so I'm ignoring this thread.

EDIT: JUST LOOK AT THE STUPID POST BELOW

[xlink]

My response to Fermi: LOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLO LOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLO LOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLO LOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLO LOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLO LOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLOLO LOLOLOLOLOLOL


Fermi = 2900XT version 2

[grimREEFER]

2900xt was never the fastest single gpu
the nvidia cards are a little faster, but when you take into account electricity and price, the results are basically ambiguous.
you can basically argue a purchase of any of the gpus(ati or nvidia) against its various competing gpus.

[annihilat0r]

god. it's as if those people really wanted Fermi to be the most gigantic flop ever. Well it wasn't good enough, but it isn't a big flop either, but these guys have practically conditioned themselves to that, and with some info that backs their points, now they virtually can't be stopped from spewing BS all around.

[zalbard]

Fermi = 2900XT version 2

Nah, more like 5800 Ultra version 2.

[Carfax]

god. it's as if those people really wanted Fermi to be the most gigantic flop ever. Well it wasn't good enough, but it isn't a big flop either, but these guys have practically conditioned themselves to that, and with some info that backs their points, now they virtually can't be stopped from spewing BS all around.

Just wait till the Forceware 200.xx drivers come out

[insurgent]

god. it's as if those people really wanted Fermi to be the most gigantic flop ever. Well it wasn't good enough, but it isn't a big flop either, but these guys have practically conditioned themselves to that, and with some info that backs their points, now they virtually can't be stopped from spewing BS all around.

If you cared so much why don't you buy yourself a 480? Most of your posts seem so contrived, as if you're trying too hard to come off as objective.

[annihilat0r]

Yep, I think other people would do good if they tried hard to be objective as well. I own a 5870 and don't need Fermi. Forgive me if my posts aren't as sincere and as objective as FERMI LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL

[zalbard]

Yep, I think other people would do good if they tried hard to be objective as well. I own a 5870 and don't need Fermi. Forgive me if my posts aren't as sincere and as objective as FERMI LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL LOL

Hey, looking at your avatar I think it's not too late for your posting style to change!


Yeah, yeah... just kidding.

[Macadamia]

god. it's as if those people really wanted Fermi to be the most gigantic flop ever. Well it wasn't good enough, but it isn't a big flop either, but these guys have practically conditioned themselves to that, and with some info that backs their points, now they virtually can't be stopped from spewing BS all around.

Personally I'm waiting for the RMAs before I start to LOL.

This thread needs some RMAs- OOPS PAPER LAUNCH


BTW I would have bought the 470 for CUDA/Premiere CS5 if it didn't use as much power as a GTX275. but nooooo, it had to hog so much power while being only a bit faster. I don't have the PSU the time nor the patience for some RMAs.

[Olivon]

Nah, more like 5800 Ultra version 2.

Except 5800 Ultra losed badly in multiple wayz against 9800/9700 Pro. Power consumption and heat are really problematic but performance are better than HD 5870. That's not the same thing IMHO.

Fermi arch' is good but too complex for 40nm. They believed that TSMC can handle that but that was a fail. TSMC have big problems with 40 nm.

[H2O]

Remember that most of the heat problems will go away if you watercool the card. If you already have a watercooling setup, GTX480 seems to be a nice bet as it outperforms the HD5870 by ~>10% (and the 2GB version by ~6%) and it's drivers have a chance to mature and hopefully the performance gains will continue to increase.