Would a GTX 580 improve my current system?

Correct, but the discussion is around what happens when you have 2 cards in SLI. Some of the "gurus" reckon 2 cards in SLI only have the same bandwidth as one, whereas I say its doubled.

 

Well each card has the the same amount of bandwidth in SLi/Xfire. They can render faster as more bandwidth is available, meaning sharing the load. I understand what you are getting at, but the actual cards bandwidth still stays the same.

 

The irony was that you quoted a post where I had shown they you were wrong (again) about the texture fill rates and yet you said I didn't know what i was talking about when every single post has shown you know nought.


The texture fill rate tests the cards texture units (TMU)... dumbass.

http://www.gpureview.com/texture-fill-rate-article-375.html

 

Are you going to search around 7 year old web pages to answer why the colour fill tests are as they are too?

 

Thanks man, that's all I am saying, the head bangers in the thread don't get it though.

 

^ hah, thankyou for saying you are wrong? you'll use any cheap way of trying to make it look like people agree with you... he said it's not doubled...

 

The guy said you get extra effective bandwidth through work load sharing, which is what I am saying....... Whether he means effectively double or not he'll have to clarify.

http://techreport.com/articles.x/18521/5

Another article explaining the correlation between 3d marks colour fill and memory bandwidth. In hindsight it may be an even better indicator than texture fill as there is no correlation whatsoever between gpu power and the results in this test, only bandwidth seems to affect it..

http://ixbtlabs.com/articles3/video/gf100-2-p9.html

On a side note, now downclock your ram and run the same test as in your picture lets see if it makes a difference, and do the same with the colour fill test.

 

Almost everything has some correlation to memory bandwidth... we already know the colour/texture fillrate increases in sli as they are based on the gpu sub units i.e rop/pixel shaders/tmus, if this wasn't the case we wouldn't see twice the fps even at lower resolutions/framebuffers where memory bandwidth/size are not the bottleneck.


Enough is enough.. stop trolling.

 

I've done the same for the color fill test:

First the results at my cards standard clocks:

http://imageshack.us/photo/my-images/254/testst.png/

Now I've dropped my cards memory clocks by exactly 25%:

http://imageshack.us/photo/my-images/851/testlo.png/

Excuse the rough cropping...... Anybody good at maths?

I am, dropping the memory clocks by 25% drops the result score by exactly 25%

Amazingly enough, upping the clocks by just over 10% from standard increases color fill score by just over 10% too.....

http://imageshack.us/photo/my-images/543/testhi.png/

Next I've overclocked my GPU by just under 12% but left my memory clocks at standard:

http://imageshack.us/photo/my-images/3/testova.png/

You should notice the result is EXACTLY the same as the results in the first link where the same memory clocks was used, so increasing the GPU speed had NO effect on the score, so therefore this test is 100% affected by memory bandwidth.

At this point I point you back to this link:

http://techreport.com/articles.x/20088/3

If you doubt my results, try it yourself.

 

ahem you all are getting stupid, stay on topic and help NeoEnigma or gtfo

 

Ah, we were all stupid pages ago my friend, we're just wasting the hours away with pointless arguing on the net.....

 

I would not upgrade from 5870 2 GB card to GTX 580. Using a mild overclock on your card can get you to 6970 level of performance which is not that far away from GTX 580 performance level. The difference between 6970 and 580 is about 5 to 10 % overall depending on the game played.

The improvement would be too small for spending several hundred bucks in a "upgrade" you will not notice at all.

Recommendations on overclocking your cpu are quite right. I had a Q6600 2.4 GHz before my i7 and a Radeon 3850. Before overclocking the cpu i tried playing Crysis which ran at 18 to 28 fps. After overclocking the cpu to 3.2 GHz i hit 21 to 34 fps and the game became playable at high settings even with that low-end video card.

Overclocking your FSB to 400 to 425 MHz will bring some boost to overall system performance and gaming performance and will get you to 3.4 - 3.6 GHz.

In your case i would start in BIOS with FSB set to 400 MHz, all voltage settings on "Auto", memory divider set to 2.0B, FSB Bootstrap set to 400 MHz.

Chances are slim that it does not run at these settings. If it runs it will instantly get you to 3.4 GHz which should be enough to give a nice boost in performance.

 

uhmm no. a 5870 2gb is still a 5870 with more vram that doesnt come close to a 580. end of.

but by all means as we have said op should oc his cpu and test and then upgrade if he still needs 2. if he has the cash oc+upgrade.

 

Right, a 5870 doesnt come close to a 580 performance wise.

 

I disagree about the video cards. Benchmarks and seem to indicate that the 580 roflstomps the 5870. The extra 1GB on the 5870 doesn't really benefit it a whole lot.

My new CPU cooler should be arriving in the mail relatively soon. I'm going to pop that sucker on and if my temps are reasonable, I'll try the overclock.

I agree, and this is my plan. It also helps that I have $100 off my next newegg order due to them severely botching my previous one, so that makes the 580 a very nicely priced card.

 

5870 is a bit faster or at least the same as 6950. With OC to 900 MHz or more it is almost the same as 6970 which is 5 to 10% slower than GTX 580. Except for the tesselation performance that is correct.

This is an old test from last year when the 6970 was released:

http://www.computerbase.de/artikel/grafikkarten/2010/test-amd-radeon-hd-6970-und-hd-6950/27/#abschnitt_performancerating_nach_api

After 11 month now the performance of both card increased through driver updates and with a mild 100 MHz overclock the 5870 can still compete against the Radeon 6970 or the GTX 580. Looks even better when you have the 2 GB VRAM model.

I have the 5870 Eyefinity 6 model clocked to 900 MHz and it delivers absolutely excellent performance on every game i have on my 1920x1200 monitor. The Vapor-X has the better cooling system and it will overclock higher without any effort i think. Reports from german hardware forum say that people use to overclock the Vapor-X models at 960 to 1000 MHz for daily gaming use.

In my opinion it is defenitely not worth spending 500 $ for an average increase of 10 FPS. As you can see in this review and many others there is no game that the GTX 580 can handle and the 5870 not. A total waste of money. Better wait and save up for the next gen video cards coming in 2012.

 

The 5870 is generally a little sower than a 6950, with a few exception where the extra shader units provide it a little more performance, the 580 is atleast 20-25% faster on average and that's quite a lead so they they aren't really comparable, the 5870 sits in about the same range as the 6950/6970/570 at most resolutions/settings.

 

Unless you are into really competitive gaming like some kind of UT world championship or something like that 20% means nothing and i think it is really not important if you have 130 fps in certain games or 140 fps. That 20% difference to GTX 580 stock speed can be compensated with overclocking.

 

Yes well if you are getting 120-140 fps then you don't even need a 5807, a 5850/6850 would be more than sufficient. Higher-end cards really come into play at higher/multi-monitor resolutions with high IQ settings, at which point that 20% would be the difference between 50 and a 60 fps v-synced or better yet a very playable 35-40 fps and a barely moving image at 25-30 fps.

The GTX 580 can also be overclocked ofcourse so the differantial isn't going to change.

 

thats true a lot of people do not take into account that the 580 can be overclocked as well. my 772/1544/4008 is running a moderate 875/1750/4200. ram used to be 4400 dropped to 4300 and now 4200 without any fps difference.