gtx 470 2 and 3 way sli review.

[happy medium]

It seems a pair of gtx 470's = 2 5870's for 100$ less.

The 2 gtx 470's only consume 47 more watts then the crossfired 5870 also.
Not bad.

http://www.guru3d.com/article/geforce-gtx-470-2-3-way-sli-review/4

 

[Lonyo]

How utterly useless. No minimum FPS numbers _anywhere_.

And I wouldn't be inclined to put much worth in those power numbers either. Apparently a single GTX480 uses 8w more than a single GTX470, but when you have 2 of each in SLI, suddenly there's a 135w difference!
Not sure which number is off, but certainly something looks like it might be off.

Also all the charts were against the HD5850, unless you were looking at the bar chart at the bottom for HD5870 vs GTX470.

 

[happy medium]

How utterly useless. No minimum FPS numbers _anywhere_.

And I wouldn't be inclined to put much worth in those power numbers either. Apparently a single GTX480 uses 8w more than a single GTX470, but when you have 2 of each in SLI, suddenly there's a 135w difference!
Not sure which number is off, but certainly something looks like it might be off.

Also all the charts were against the HD5850, unless you were looking at the bar chart at the bottom for HD5870 vs GTX470.

Minimum fps are gtx strong points, that should not be a problem.

100$ less for equal or better performance is nice though.

 

[Lonyo]

Minimum fps are gtx strong points, that should not be a problem.

100$ less for equal or better performance is nice though.

Minimum FPS is the strong point of the GTX480. The GTX470 starts falling behind in minimum FPS in comparison.

Also it's 2 wins NV, 4 draws, 2 wins ATI with the HD5870 vs GTX470 (at 1920x1200 as shown in the charts below).

 

[Skurge]

Its disappointing the 5870CF results are for 1920x1200, the 5850CF closes the gap at 2560x1600

So I wouldn't say 470SLI beats or equals 5870CF. At lower Resolutions maybe, at the top end, I don't think so.

EDIT: I never trust GURU3d reviews, ever!!

 

[RussianSensation]

It seems a pair of gtx 470's equal or beat 2 5870's for 100$ less.

Curiously I do not see the numbers for 5870 CF setup other than for Modern Warfare?

As far back as I can remember, SLI generally scales better than CF. So I wouldn't be suprised if GTX470s beat out 5870s.

 

[happy medium]

Its disappointing the 5870CF results are for 1920x1200, the 5850CF closes the gap at 2560x1600

So I wouldn't say 470SLI beats or equals 5870CF. At lower Resolutions maybe, at the top end, I don't think so.

EDIT: I never trust GURU3d reviews, ever!!

Well most people have a 22 inch monitor, but in these forums most have a 24 inch monitor @ 1900x1080.

Very few care about 2500x1600 res.
In more demanding games the 58xx series fails at 2500x1600 ultra high settings due to there lack of ram also.

You have nothing nice to say about sli'ed gtx 470's ha? hmmmm typical thread crapping fanboy.

 

[happy medium]

Minimum FPS is the strong point of the GTX480. The GTX470 starts falling behind in minimum FPS in comparison.

Also it's 2 wins NV, 4 draws, 2 wins ATI with the HD5870 vs GTX470 (at 1920x1200 as shown in the charts below).

For 100$ less.

 

[Qbah]

Already thoroughly discussed here:

GTX 470 SLI or 5970?

EDIT: On second thought, this link is HD5970 vs GTX470 SLi, yours is different.

Anyway, I'd go green if presented with those two choices. Speed is almost the same, nVidia offers a bit more features.

However, the price argument is only valid for one market - US. Here in Europe (pretty much anywhere else actually) they cost the same. Still, a draw is a draw based on those numbers, hence I'd go green.

 

[MrK6]

Yeah, I think Guru3D was bought by NVIDIA a few years ago...

But in all seriousness, even looking at other reviews around the net, you'll find that there isn't much difference among the high end cards, especially once they're put into multi-GPU setups. That's why I've been saying for awhile that 5850CF is the way to go, and anything else shows such a diminishing return on investment I'd call it a waste of money. Will GTX 480 SLI get you faster performance than 5850CF? Yes. Is it much different? No, and certainly not $500 different in my eyes. This rings especially true once you throw overclocking into the picture. For example, in a lot of benchmarks my overclocked to 900/1250 trade blows with a GTX 480 SLI setup. Yes, you can overclock the GTX 480's too, but then I could push my 5850's more as well. In the end, the GTX 480's only end up being 10-15% faster for a lot more money.

Now there are niche reasons for getting some higher end hardware (GTX 470 for CUDA development, 5870 2GB for Eyefinity goodness, etc.), but overall the cards of the high-end segment don't do enough to differentiate themselves.

My $0.02.

EDIT: I should have probably included some sources: http://www.hardocp.com/article/2010/05/10/galaxy_geforce_gtx_470_gc_sli_review/1

This review was done this week, and features slightly overclocked Galaxy GTX 470's. Using the "Apples-to-Apples" comparisons, the HD5850 and GTX470's pretty much trade blows. The Metro 2033 CF results are odd, as my cards scale well with the game, and previous reviews don't have this issue, so I guess that was a driver bug?

 

[Skurge]

Well most people have a 22 inch monitor, but in these forums most have a 24 inch monitor @ 1900x1080.

Very few care about 2500x1600 res.
In more demanding games the 58xx series fails at 2500x1600 ultra high settings due to there lack of ram also.

You have nothing nice to say about sli'ed gtx 470's ha? hmmmm typical thread crapping fanboy.

So If i don't agree with you, im a fanboy? Gotcha

You are so blinded by your own fanboyness you failed to see I didnt say anything bad about it either, You always claim 5800series run out of ram at 2560x1600 and you havnt shown that yet.

I wouldn't conclude one is faster over the other from ONE guru3d review. show me a constant set of benches from a number of sites (Just like I did with 5970CF vs gtx480SLI) and then ill agree with you.

 

[SlowSpyder]

Well most people have a 22 inch monitor, but in these forums most have a 24 inch monitor @ 1900x1080.

Very few care about 2500x1600 res.
In more demanding games the 58xx series fails at 2500x1600 ultra high settings due to there lack of ram also.

You have nothing nice to say about sli'ed gtx 470's ha? hmmmm typical thread crapping fanboy.

But most people who are buying two GTX470's or two 58x0's have more than a 22" monitor.

If the GTX470 SLI gives you roughly the same performance as two 5870's (and that looks to be debatable depending on the review site you trust and resolution/game) than it's not a bad deal at all.

But for an multi-GPU enthusiast rig I'd still consider overclocking and probably get two 5850's.

 

[Skurge]

Yeah, I think Guru3D was bought by NVIDIA a few years ago...

But in all seriousness, even looking at other reviews around the net, you'll find that there isn't much difference among the high end cards, especially once they're put into multi-GPU setups. That's why I've been saying for awhile that 5850CF is the way to go, and anything else shows such a diminishing return on investment I'd call it a waste of money. Will GTX 480 SLI get you faster performance than 5850CF? Yes. Is it much different? No, and certainly not $500 different in my eyes. This rings especially true once you throw overclocking into the picture. For example, in a lot of benchmarks my overclocked to 900/1250 trade blows with a GTX 480 SLI setup. Yes, you can overclock the GTX 480's too, but then I could push my 5850's more as well. In the end, the GTX 480's only end up being 10-15% faster for a lot more money.

Now there are niche reasons for getting some higher end hardware (GTX 470 for CUDA development, 5870 2GB for Eyefinity goodness, etc.), but overall the cards of the high-end segment don't do enough to differentiate themselves.

My $0.02.

agree with this 100%

 

[JAG87]

The only conclusion you need to draw from that review is that in 2-way SLI > Crossfire. That is why 470 < 5870 but 2x470 > 2x5870.

From what I've read though, crossfire has another gear in 3-way and scales better than 3-way SLI. And 3-way solutions with Fermi are just not conceivable in terms of power consumption, while 3-way Crossfire is very conceivable.

 

[happy medium]

The only conclusion you need to draw from that review is that in 2-way SLI > Crossfire. That is why 470 < 5870 but 2x470 > 2x5870.

From what I've read though, crossfire has another gear in 3-way and scales better than 3-way SLI. And 3-way solutions with Fermi are just not conceivable in terms of power consumption, while 3-way Crossfire is very conceivable.

I think you might run into a cpu bottleneck with 3 of these high end cards, at least with most games
Wouldn't you think?

I agree with your scaling comment.

 

[happy medium]

So If i don't agree with you, im a fanboy? Gotcha

You are so blinded by your own fanboyness you failed to see I didnt say anything bad about it either, You always claim 5800series run out of ram at 2560x1600 and you havnt shown that yet.

I wouldn't conclude one is faster over the other from ONE guru3d review. show me a constant set of benches from a number of sites (Just like I did with 5970CF vs gtx480SLI) and then ill agree with you.

I didn't want to get in a who's faster argument.
My point was for 100$ less you can get = performance to sli'd 5870's.

@MRK6
As far as overclocking goes gtx 470's with a slight voltage bump net you a 20% overclock but even better you get all 20% of that overclock in performance.

 

[Skurge]

I didn't want to get in a who's faster argument.
My point was for 100$ less you can get = performance to sli'd 5870's.

@MRK6
As far as overclocking goes gtx 470's with a slight voltage bump net you a 20% overclock but even better you get all 20% of that overclock in performance.

You said equals or beats, not equal or trades blows. Agreed that in the US they might be better value, but you do this in almost every gtx thread you make. You say that particular gtx card IS faster when your links are inconclusive.

 

[happy medium]

You said equals or beats, not equal or trades blows. Agreed that in the US they might be better value, but you do this in almost every gtx thread you make. You say that particular gtx card IS faster when your links are inconclusive.

I edited original post.

 

[Skurge]

I edited original post.

Your far more reasonable than most people on this thread. I'll give you that.

 

[MrK6]

@MRK6
As far as overclocking goes gtx 470's with a slight voltage bump net you a 20% overclock but even better you get all 20% of that overclock in performance.

No, not really. Fermi does tend to scale better with an overclock than the Cypress chips, but it's not 100%. Like I said, the costs (monetary or otherwise) don't outweigh the benefits IMO.

 

[happy medium]

No, not really. Fermi does tend to scale better with an overclock than the Cypress chips, but it's not 100&#37;. Like I said, the costs (monetary or otherwise) don't outweigh the benefits IMO.

See my last post in the other gtx 470 thread..

copy paste it for ya.

The overclocked gtx 470 scored 133 fps (106 fps stock) and the gtx 480 got 135 fps. 24% overclock and 24% more performance. That great overclocking scaling.

A overclocked gtx 470 at this speed will eat a 5850 overclocked for breakfast.

http://www.guru3d.com/article/kfa2-galaxy-geforce-gtx-470-review/20

 

[tincart]

You have nothing nice to say about sli'ed gtx 470's ha? hmmmm typical thread crapping fanboy.

Translation: I find this impressive and everyone else must find it impressive too!

Agree with a previous comment above, SLI has and continues to scale very well.

 

[MrK6]

See my last post in the other gtx 470 thread..

copy paste it for ya.

The overclocked gtx 470 scored 133 fps (106 fps stock) and the gtx 480 got 135 fps. 24&#37; overclock and 24% more performance. That great overclocking scaling.

Actually it's a 28.64% overclock and you're getting 25.4% return on performance (and that's counting the core only, mind you; the RAM overclock will synergize this, so it's probably even less). It's good scaling in this particular game, but I don't know if it's universal.

A overclocked gtx 470 at this speed will eat a 5850 overclocked for breakfast.

http://www.guru3d.com/article/kfa2-galaxy-geforce-gtx-470-review/20

Considering I've shown my 5850(s) at 900/1250 to be equal to or faster than a GTX480(s), and this GTX 470 still can't outperform a GTX480, no, it won't. Also remember that my 5850's can be pushed to 1025MHz+, how much more headroom do you think that GTX 470 has, if any?

 

[Skurge]

Actually it's a 28.64% overclock and you're getting 25.4% return on performance (and that's counting the core only, mind you; the RAM overclock will synergize this, so it's probably even less). It's good scaling in this particular game, but I don't know if it's universal.

Considering I've shown my 5850(s) at 900/1250 to be equal to or faster than a GTX480(s), and this GTX 470 still can't outperform a GTX480, no, it won't. Also remember that my 5850's can be pushed to 1025MHz+, how much more headroom do you think that GTX 470 has, if any?

are those stock cooled 5850s?

 

[MrK6]

are those stock cooled 5850s?

Yep, reference Sapphire models; I haven't even changed the TIM. For technical details, I keep them at 900/1250 because I only had to bump the voltage to 1.1125V (stock is 1.0875V) and they're 100&#37; stable on the default fan profile (which is between 27-32% in games). If I up the voltage to 1.3V and the fan speed to 50-60%, they'll do 1GHz. If I crank the volts and fans even higher, they'll do 1.025GHz+, but I've never really kept them up there for anything beyond benchmarking, and I run them daily at 900/1250 because I enjoy the silence.