1.5gb vs 3gb 580

dietrich · 3 Jul 2011 at 05:47

If you were to choose guys between these two 580's what would you pick? The msi 1.5 lightning or the gainward 3gb phantom.

harmony · 3 Jul 2011 at 06:09

Keep in mind that the MSI Lightning is a long card and you'll need to make sure your case is compatible.

The 3GB Phantom is taking more than 2 slots.

3GB cards are designated for SLI under high resolution. A single GTX580 with 3GB vram is pointless in most cases.

Due to the nature that the 3GB cards are meant to setup SLI, I only recommend 3 versions of 3GB GTX580s:

a) MSI GTX580 Lightning Xtreme Edition
b) EVGA reference 3GB GTX580
c) Palit dual-slot 3GB GTX580

Mr. May · 3 Jul 2011 at 11:21

harmony said:
Keep in mind that the MSI Lightning is a long card and you'll need to make sure your case is compatible.

The 3GB Phantom is taking more than 2 slots.

3GB cards are designated for SLI under high resolution. A single GTX580 with 3GB vram is pointless in most cases.

Due to the nature that the 3GB cards are meant to setup SLI, I only recommend 3 versions of 3GB GTX580s:

a) MSI GTX580 Lightning Xtreme Edition
b) EVGA reference 3GB GTX580
c) Palit dual-slot 3GB GTX580

+1 couldnt off put it better myself

dietrich · 3 Jul 2011 at 12:05

Do you think 1.5gb vram is enough for 3D gaming and a 27" 120hz monitor? like this one.

chaparral · 3 Jul 2011 at 12:09

dietrich said:
Do you think 1.5gb vram is enough for 3D gaming and a 27" 120hz monitor? like this one.

that a very low Resolution for a 27" monitor (it even lower resolution then my 24")

Rroff · 3 Jul 2011 at 12:12

You can't get much higher res at 120Hz before running into the bandwidth limits of current connection standards.

1920x1080 requires just under 9Gbit/s - dual link DVI maxes out at 9.9, HDMI 1.3 only managed 10.

rounddodger · 3 Jul 2011 at 12:33

For upto 1080p 1.5GB is enough.

chaotic_uk · 3 Jul 2011 at 14:31

rounddodger said:
For upto 1080p 1.5GB is enough.

not if he is going for 3D ?

Icm76 · 3 Jul 2011 at 14:48

[H]ard|OCP recently published a really nice GTX 580 SLi 3GB vs 1.5GB vs 2GB AMD comparison: MSI N580GTX Lightning XE 3GB SLI Video Card Review

See the this page of the review and scroll down to the results for Metro 2033 @ 2560×1600 with 4×MSAA & DoF enabled. One of the stand out results, on average framerate 3GB is 4.2 times faster than 1.5GB. Of course the gap is much smaller in other tests, and AMD win on overall value vs performance.

Meaker · 3 Jul 2011 at 21:12

Hmz that looks bugged to me since at 3 times the resolution the same limitation at decent details did not work.

Duff-Man · 3 Jul 2011 at 21:21

Meaker said:
Hmz that looks bugged to me since at 3 times the resolution the same limitation at decent details did not work.

The 2560 resolution test has depth of field enabled - that thing eats memory and GPU clock cycles like mad. It seems like the VRAM requirement in that benchmark is between 1.5 and 2Gb.

Also, 5760*1200 is 1.7 times higher resolution than 2560*1600, not three times.

Icm76 · 3 Jul 2011 at 23:40

[H] reviews need to be read carefully, under the 'highest playable settings' table they run 4×MSAA on the 3GB, but AAA on the 1.5GB.

Rroff · 3 Jul 2011 at 23:42

[H] Reviews are meaningless imo due to their testing methods, even apples to apples aren't the most enlightening, the rest are easily skewed to show anything you want and very obviously selectively picked sometimes.

Icm76 · 3 Jul 2011 at 23:50

Rroff said:
[H] Reviews are meaningless

Only if reading and comprehension are advanced skills to you

If you need reviews to hold your hand and just point to the biggest number then sure, stick to the predictable canned benchmarks on other sites.

Rroff · 3 Jul 2011 at 23:56

Icm76 said:
Only if reading and comprehension are advanced skills to you If you need reviews to hold your hand and just point to the biggest number then sure, stick to the predictable canned benchmarks on other sites.

You've obviously never read any of my other posts regarding reviews.

Icm76 · 4 Jul 2011 at 00:54

I take it you're not keen on the canned benchmarks either then? I like hardocp reviews precisely because they go outside of the norm, sometimes they frustrate because they don't test enough at mainstream res (Kyle is obviously obsessed with multi-screen gaming) and I wouldn't use the results exclusively to choose which graphics card to buy, but the tests aren't meaningless. They provide a useful alternative that can expose strengths and weaknesses not shown in other reviews, and if they cause controversy that should be a good thing as it prompts others to look more deeply at specific issues.

harmony · 4 Jul 2011 at 01:20

Rroff said:
[H] Reviews are meaningless imo due to their testing methods, even apples to apples aren't the most enlightening, the rest are easily skewed to show anything you want and very obviously selectively picked sometimes.

Hardocp is one of the best review site for enthusiasts. For example, when an enthusiast is setting up a new rig for triple 30" gaming, only Hardocp review would tell him to use GTX580 3GB versions in SLI, because most of the other reviews fail to have the ball to test certain games with max settings (e.g. Metro 2033). After reading reviews from Hardocp the enthusiast should know that he'd insist buying the 3GB versions in SLI even if it's out of stock.

Randomface74 · 4 Jul 2011 at 01:28

I would go for the 1.5 Phantom

Rroff · 4 Jul 2011 at 01:45

Icm76 said:
I take it you're not keen on the canned benchmarks either then? I like hardocp reviews precisely because they go outside of the norm, sometimes they frustrate because they don't test enough at mainstream res (Kyle is obviously obsessed with multi-screen gaming) and I wouldn't use the results exclusively to choose which graphics card to buy, but the tests aren't meaningless. They provide a useful alternative that can expose strengths and weaknesses not shown in other reviews, and if they cause controversy that should be a good thing as it prompts others to look more deeply at specific issues.

For a long time I've been an advocate of more information than just canned benchmarks, especially I want to see benchmarks that show a percentage of time below 30fps and 60fps rather than a min and average figure (average figure is good to).

While I applaud the effort to present information outside the normal box of canned benchmarks I'm of the opinion that the non-apples to apples benchmarks are often quite subjective in regards to settings and way to easy for potential scenarios to be glossed over or missed that can completely change the outcomes (theres just way too many variables to conclusively test for in any sensible amount of time).

The apples to apples info is more useful but I've seen it contradict the other bits to often.

Meaker · 4 Jul 2011 at 10:54

The non-apples to apples should just be a table with certain criteria to achieve the highest resolution time spent below a certain FPS or number of dips below, that kind of thing.

Then multiple graphs over multiple resolutions at the same settings.