The GeForce GTX 480 Update: 3-Way SLI, 3D Vision, And Noise

Page 4 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.

youssef 2010

Distinguished
Jan 1, 2009
1,263
0
19,360
Just one question before I post my large comment,Is there a good point of comparing 2 5870s to 3 480s costing twice as much?.I don't know about the value part,but a chart displaying the performance of 4 5870s vs. 3 480s would've been nice as they're nearly @ the same price($60 difference).
 

youssef 2010

Distinguished
Jan 1, 2009
1,263
0
19,360
[citation][nom]cangelini[/nom]Unfortunately, the only 2GB cards here are a pair of Eyefinity 6 boards, which cost $100 more each. That extra 1GB does not establish parity with GTX 480, even at those higher resolutions. For more on the numbers there, please see this piece.[/citation]
[citation][nom]cangelini[/nom]Unfortunately, the only 2GB cards here are a pair of Eyefinity 6 boards, which cost $100 more each. That extra 1GB does not establish parity with GTX 480, even at those higher resolutions. For more on the numbers there, please see this piece.[/citation]

Please don't talk about cost when you're comparing two setups with one costing literally half the other.And if you need a 2GB non-eyefinity edition, here you go

http://www.newegg.com/Product/Product.aspx?Item=N82E16814102893&cm_re=5870-_-14-102-893-_-Product
 

youssef 2010

Distinguished
Jan 1, 2009
1,263
0
19,360
I've only been through the first page and you're comparing 3 480s against 2 5870s.That means that you're comparing a $1497 setup (lowest price on Newegg is $499 per card) against a $780 setup (lowest price was $390 on Newegg). The 480s cost almost twice as much as the 5870s (not factoring in the beefier PSU needed to drive those 480s).The least to do is compare setups that cost the same amount of money, meaning 4 5870s (for $1560) vs. 3 480s ($1497).Now, that's a difference of $60 (acceptable over twice the price).You could've at least done 3 vs.3.Although,4 5870s are technically feasible on this motherboard using powercolor's LCS version (because the fourth slot isn't located properly to allow a normal setup). This is more expensive but can be used only to gather data.
 

juliom

Distinguished
Jul 21, 2009
84
0
18,630
It's quite impressive that NO ONE can see that the severe lack of performance at 2560 x 1600 with AA is only because of the smaller framebuffer of the ATI cards... Chris, what about mentioning it in the article?
 

eyefinity

Distinguished
Mar 6, 2010
1,106
0
19,310
You have to understand toms is very much an Nvidia favouring website. Why else would they benchmark a game like metro 2033 and not benchmark battlefield bc2?
 

bk420

Distinguished
Apr 24, 2009
264
0
18,780
I see a much different story over at overclockersclub... Strange, but whatever, TH has been falling behind lately anyhow.

I think TH also forget to mention the E6 editions can function at much higher res where nvidia can't even compete (0 fps). Just my thought since they did mention a lot of 3D hoopla.
 

Tallbastard

Distinguished
May 3, 2010
1
0
18,510
@ juliom: Chris is more concerned with negatively framing the platform than he is with providing a good review of a likely real-world setup. Once again, running trl sli 480s with an i7 930 is pointless, as the scaling on the third card clearly demonstrates. When he states that "The addition of a third card helps boost frame rates a bit, but not by enough to justify another $500+, in my opinion" Well, maybe you should benchmark with a cpu comparable to the quality of the gpu config, you know to maybe, uhh, feed enough data to your top end gpu setup. Your opinion is based on a crippled rig that NO ENTHUSIAST worth 2 cpu cycles would ever build.

No one but an enthusiast is going to purchase three 480s. Set up an enthusiast rig to bench them. Putting 3 480s into that setup is like taking the engine out of an Enzo and putting it an Audi.
 
[citation][nom]dreamphantom_1977[/nom]See, exactly what I'm talking about, the lowest price 5870 is $389.00, the lowest price gtx 480 is $499.00, last I checked that is not "almost twice as much", that is $110 which is only 26% more. And considering the fact that to add physX to a ati system with the hack would cost about $100 bucks,which would also consume more electricity, wich makes the cost about the same, and considering the better performance, the nvidia card is still a better buy, because even though these one sided benchmarks don't show, there are some situations were the gtx 480 cards run almost twice as fast as the ati 5870, like heaven benchmark, or metro 2033 at ultra high resolution.[/citation]

No, you idiot, he was talking about the price of 3 480s vs the price of 2 5870s, which is nearly twice as much, which doesn't factor in the extra PSU cost. Even enthusiasts have budgets.
Stop talking about physX, nobody cares about it.
I didn't get to read the full article yet, so I can't comment on the benchies yet, but after I read them, I'll see if I agree with your claim of biased or not.
 

notty22

Distinguished
This was not a direct flagship video card challenge. The focus of this article was the GTX 480 itself,in SLI+ tri sli. The other cards there were to set a baseline or a comparison. Not to to demonstrate that the GTX 480 is faster. Everyone should know this anyways. As far as heat + power. A more powerful card uses more of both, this is true with the 5850 as compared to the 5870.

The telling graph showing your performance per dollar is the same or better than the 5870 is the icing on the cake. You want the absolute most powerful single gpu in the world. You choose the GTX 480. Imho, there are countless intangibles that make Nvidia gaming better. For me, its not JUST price/performance. But thats impossible to argue with the current atmosphere, here at Semi Accurate,oops TH :)
 

dragonsqrrl

Distinguished
Nov 19, 2009
1,280
0
19,290
It's good to know that GTX 480 2-way SLI has good scaling on an 17-930, but I have a few issues regarding 3-way SLI. I don't know about anyone else, but I personally think it's unrealistic to run a setup containing 3 GTX 480's in tri-SLI with an i7-930. That's $1500 worth of graphics hardware running on a $300 processor? I'm assuming anyone willing to spend a disproportionately higher amount on graphics hardware relative to their processor would at least attempt to overclock 'the hell' out of that disproportionately cheaper processor. Otherwise, I'm thinking most people who have $1500 to burn on graphics alone probably won't have much trouble investing in a faster processor, say an i7-980x.

As soon as I read the specs for the test system I already knew how these benchmarks were going to turn out. Of course your going to see terrible scaling going from 2 to 3 GPU's. 3 way SLI scaling on the fastest processors available today + massive OC's is already mediocre, switching to a 3.3 GHz i7 is only going to make the situation worse.

There have been reviews suggesting that the biggest issue with running the GTX 480 in tri-SLI is the processor, and not scaling problems inherent in the forceware dirvers. Scaling was better on an i7-980x when comparing the benchmarks, and improved even more when OCing to 4 GHz. However, one thing was certain across all benchmarks, even at 4 GHz the i7-980x was still running into CPU bottlenecks with the GTX 480 in 3-way SLI.
 

dragonsqrrl

Distinguished
Nov 19, 2009
1,280
0
19,290
[citation][nom]backin5[/nom]1) What about temperatures under load? Were you afraid to scare us off with the results, Chris? Or maybe I just somehow missed them when reading? 2) I'm surprised Nvidia recommends the SilverStone Raven RV02. The simple idea behind it is great, but sadly tests revealed that the design in practice (or at least SilverStone's implementation of it) doesn't achieve remarkable results temperature-wise.3) The results for Metro 2033 at 2560x1600 with all different configurations seem to indicate there might be some problems with both the game and the drivers (for both ATI and Nvidia).[/citation]
Yes... you "somehow missed them when reading". Despite popular belief on this thread the author does mention the power consumption of the 3-way SLI setup, try re-reading... carefully.
 

dragonsqrrl

Distinguished
Nov 19, 2009
1,280
0
19,290
[citation][nom]ajulian[/nom]And, Why didn´t you use the 2Gb version of the Radeon HD 5870, for me it´s pretty clear that the 1Gb Card will see its performance greatly hurted at 2560 x 1600 and even at 1920 x 1080 4X AA and 16X AF, would be more fair, don´t you think so?[/citation]
Did you read Tom's review of the 2GB HD5870 Eyefinity 6 edition? Despite the hopeful rumors of a "GTX 480 killer" prior to its release, the HD5870 Eyefinity 6 doesn't improve performance at the lower resolutions you've mentioned, in fact in many cases performance is lower then the 1GB HD5870 at comparable resolutions.
 

dragonsqrrl

Distinguished
Nov 19, 2009
1,280
0
19,290
[citation][nom]banthracis[/nom]For those wondering wondering about CPU limitations, I'll point you to an article Legion Harware did a few months ago using a 5970, which is pretty close to xfire 5870's. The summary is that with eye candy turned on, a stock speed i7-920 has basically the same FPS as a 4.0ghz OC i7-920. Given that much headroom, I doubt even the 3 way SLI 480's were CPU limited. http://www.legionhardware.com/arti [...] 970,8.htmlDone at 2560 x 1600[/citation]
http://www.guru3d.com/article/geforce-gtx-480-3way-sli-review/
...have a look, CPU bottlenecks abound.
 

descendency

Distinguished
Jul 20, 2008
582
0
18,990
Just curious (because the question is actually pertinent to the article), but can you actually play Crysis at 2560x1600 with everything turned on yet?
 

doron

Distinguished
Feb 15, 2009
553
0
19,010
The performance per dollar advantage vanishes if you're not building a completely new rig and need to upgrade PSU and case just to fit Nvidia's requirements for SLI. Also the huge power outage of 2x480 might explain the added performance, which is highly unjustified IMHO
 

gti88

Distinguished
Jul 11, 2009
242
0
18,680
I think, in Metro 2033, you don't get reasonable visual enhancement on ultra settings compared to medium, but perfomance drop is ridiculous.
Also, DX11 is next to useless in this game.
 

doron

Distinguished
Feb 15, 2009
553
0
19,010
Hmmm.. There was also an argument regarding card VRAM which could greatly affect higher resolution performance and which I personally agree with, maybe pit these gtx480 vs some 5870 2gb cards? They're also equally priced more or less, should be interesting ;)
 
No. Just...no. Over 1KW just to play some games? You would practically need a dedicated circuit to power it; add THAT cost to your analysis. I understand about spending money on one's hobbies, but I don't have a six-figure income, and a system like this isn't some dream; to me it represents a colossal waste.
 
So, Chris did you enable PhysX in Just Cause 2?

It wasn't mentioned in the settings, but it was made mention in the explanation of the poor performance, yet other testers (like Xbit) found the same thing with GPU water disabled.

Needs to be a little clearer IMO.

As for the sound thing, it's funny that now "it's not an issue" as long as you buy a new case built specifically for the graphics card, to which end, I would wonder why AMD wasn't contacted for their input on the best case for their sound and cooling?

It's laughable that the way to handle this issue is by changing the tests, since a CoolerMaster Centurion case wasn't good enough in the first nVidia 'recommended retest'. I recommend people reduce sound levels lower and get liquid cooling. ;)

I wonder how many people felt it was the case's fault for the bad HD2900 and FX5800 sound & heat issues? :heink:
 

thegreathuntingdolphin

Distinguished
Nov 13, 2009
256
0
18,780
No. Just...no. Over 1KW just to play some games? You would practically need a dedicated circuit to power it;

You bring up an excellent point in the circuit. Most people realize it will cost more in power; however, that much power on one circuit can cause problems. If you have a bunch of stuff connected to the circuit in an older house or apartment, you might trip the breakers. I had a Q6600 overclocked with 2 overclocked GTX 260s being stressed tested, another computer stress testing (weak PC, maybe 250 watts), 2 monitors, and a space heater (that was running at a KW and was upstairs) and after 30 seconds of stress testing the whole circuit (1/3 of my apartment) went out.

Something for people to consider...
 

lashton

Distinguished
Mar 5, 2006
607
0
18,990
[citation][nom]dreamphantom_1977[/nom]You can can tell toms is full of ati fanboys. You throw evidence out that "CLEARLY" shows gtx 480 is faster then ati 5870 and they still say ati is better and you post evidence showing gtx 480 is more powerful, and the fanboys just pretend like the evidence isn't there. Then when they are cornered, they throw out heat and noise and power consumption, as if running any high end rig is really energy efficient (rofl), and everyone knows if you have the money to spend on tri sli, it's gonna end up water cooled. Even toms writers are ati fanboys, as every other site clearly shows gtx 480 is the winner, and they base there best buy on non AA numbers, when everyone knows a true test of a graphics card is what it can run "MAX" settings in a non bottlenecked system, which obviously they taylored to give ati an advantage. Not even mentioning that for ati to physx you still need to purchase a nvidia card. (See my last post, that what probably thumbs down to -4 and not showing now because it shows the truth that even in toms articles is there bias towards ati even though it is underpowered). The only win I would give ati would be in the dual gpu single card win. Yeah, it's the champ right now, but you still don't have physX. In the charts, it clearly shows nvidia is the winner. I'm getting so sick of posting anything about nvidia, and always thumbs down. I could just post a fricken pic of nvidia and no comments and it would get 10 thumbs down. The only win I give ati is availability. We are gamers, not tree huggers. Seriously, they are talking sli and crossfire with top end cards. Energy efficiency isn't a huge concern with hardcore gamers. If it was, nobody would even be buying high end cards. There wouldn't even be a market for it, but there is. Also, all the cards run hot and loud, thats why most people water cool. Obviously it's not in the news about gtx 480's burning up more then any other cards, so it's not a problem. And yes, the system setup is bottlenecking the cards. So sick of everyone bashing posts just because it's "nvidia". Seriously, I just about had it with toms forums. It's B.S. Go ahead, give me thumbs down, I don't care, cuz it just proves my point.I've never seen any other forum's that are so blindsided to benchmarks and so one sided. It's rediculous, I mean most of us are fanboys, but it's rediculous that the same comments keep coming up after every post about nvidia and every time you post you get negative after negative if you post anything positive about nvidia. Yeah, there cards run hot, and aren't as efficient as ati's, but they are great cards and offer a lot for the money. Look at the charts. But you don't see me or other nvidia fans giving negatives on every ati post.[/citation]
ati fanboys what about nvidia fanboys saying "the GTX480 is the fastest single GOU card" well maybe but it still is not the fastest Video card, thats like saying my FTO is the fastest non turbo car, there are faster cars out there, who the hell wants 1kW of consumption FFS! you say high end rigs dont really care about energy consuption, well i dont wanna use 1kW (1000 watts) of energy on a system that is noisier than my TV, and beleieve me real Gamers care about this crap all you are doing is covering up the obvious short falls of Fermi, your idea on physx is mute people dont really give a flying f**k about physx, notice how the benchie used 3 GTX480s and only 2 5870, use the same money and put 4 5870s against 3 480GTX and then we see nvidia get thier ass handed to them, besides just weeks away in the rumour super clocked 5870 with 2 GB GDDR5 RAM with 1 GB each using thier OWN memory controller, 512bit effective through pout, i see the GTX480 going down
 
Status
Not open for further replies.