Upgrade to Nvidia 580

[lake2k]

Hi I am thinking of maybe upgrading from my ati 5870 to a nvidia 580, i was looking at upgrading to a 480 few months back but decided to hold back as am thinking of going down the nvidia route again which i aint done since geforce 2 back in the day.

Few questions what are the benefits of features like cuda and physix that ati dont have

And if was ever thinking of using 3d vision would i need 2 of these cards

I would not be considering that anyway most likely cause would also have to replace my psu and from few issues this year realised cant use crossfire with 5870s or a 5970 with my psu which is 800 watt and it cant handle it on full load

So as well as extra features i would be looking for fastest single gpu setup

Not really to bothered about the upcoming 6900 range.

Any advice be helpful

Thanks

 

[Skyrocket]

CUDA is just a word for a programming language.
It dont mean squat.
Its like saying my card has C++ support....................
Physx is mostly a glorified havoc engine, when used right and correctly CAN be stunning. Too bad no game has used it well yet.
3D vision is ok but still a gimick at this stage in the day of small monitors....And costs some serious £. Some ppl say you might need SLI to make the most out of it but thats bs 580 gtx on its own will be fine.

580 OC vs 5870 OC you will see 50% FPS increase. simples. Good card.
Its on par with 5970 at stock.

 

[Mr Krugga]

580 OC vs 5870 OC you will see 50% FPS increase. simples. Good card.
Its on par with 5970 at stock.

I'm afraid there's not much truth in these statements.

Not that GTX580 is a bad card - it isn't. It certainly would suit the OP's needs as well. It just isn't as good as you colour it.

 

[RavenXXX2]

It's up there with a 5970, wins some and loses some as seen in this review, mild OC on the 580.

If we had to pick one of the systems from this or the original launch review as gamers, we would lean towards the GTX 580. Our reasoning? A solid gaming experience at 1920x1080, no reliance on driver profiles, low noise cooling with smooth fan speed increases and the great 3D/PhysX solution.

 

[Mr Krugga]

It's up there with a 5970, wins some and loses some as seen in this review, mild OC on the 580.

Is it just me or are you trolling a big one again? This quote has nothing to do with relative performance and earlier on in the Conclusion of the same review it says:

If we are looking purely at performance it is fair to say that the dual GPU systems are faster in a number of benchmarks, sometimes by a significant amount.

There are problems with multi-gpu setups and nobody is recommending a 5970 or Crossfire 6870s but stating that GTX580 is on pair with HD5970 is bull****.

Why are you people allowed to mislead others?

 

[RavenXXX2]

It's up there with a 5970.

 

[Mr Krugga]

It's up there with a 5970.

Whatever you say.

 

[kylew]

It's up there with a 5970 when crossfire isn't working properly.

Fixed so as not to be misleading.

 

[straxusii]

I'm afraid there's not much truth in these statements.

Not that GTX580 is a bad card - it isn't. It certainly would suit the OP's needs as well. It just isn't as good as you colour it.

I wanted to find out if it was worth me upgrading from 5870 to 580 so I took a series of gaming only benchmarks from Techpowerup and Anands, at 1920x1200 the result was 580 was 44% faster.

If it's anything like the 480 for o/c then I suspect it will overclock better than a 5870 so the 50% claim might not be as far off as you think.

 

[straxusii]

According to techpowerup at 1920x1200 the 5970 is 8% faster than GTX580, for a gain of only 8% I can't imagine anyone would choose 5970 compared to 580 given the potential for multi GPU issues.

580 is also a lot quieter 41dB compared to 49

 

[Mr Krugga]

I wanted to find out if it was worth me upgrading from 5870 to 580 so I took a series of gaming only benchmarks from Techpowerup and Anands, at 1920x1200 the result was 580 was 44% faster.

If it's anything like the 480 for o/c then I suspect it will overclock better than a 5870 so the 50% claim might not be as far off as you think.

Oh well, don't know how good your maths skills are but according to my Excel spreadsheet, GTX580 is about 36% faster at this resolution. Are you sure you included games only? OC is irrelevant when comparing cards that can both OC. And there's always a case of system setups and overclocking potential of individual cards thus making overclock comparison relavant under certain circumstances only.

According to techpowerup at 1920x1200 the 5970 is 8% faster than GTX580, for a gain of only 8% I can't imagine anyone would choose 5970 compared to 580 given the potential for multi GPU issues.

580 is also a lot quieter 41dB compared to 49

Nobody's recommending HD5970 because it's a crap card compared to GTX580. The latter isn't a good value either, still miles better at current prices (£375 for GTX580 and £415 for HD5970).

 

[CreAtiv3]

Mr Krugga you are a gentleman!
Thanks for putting that together, very useful for those looking to make the upgrade.

 

[Mr Krugga]

Mr Krugga you are a gentleman!
Thanks for putting that together, very useful for those looking to make the upgrade.

I'll add a few more scores from different reviews, might help some blokes in this Forum. Got Guru3D done already, might do HardOCP and bit-tech in a minute.

 

[CreAtiv3]

Would certainly be appreciated if you had the time. OcUK would have one resource for all the performance figures on the GTX580 then. Imagine it would cut down on some of the petty bickering that goes on here too.

Im upgrading from a 4870X2, which is said to perform similarly to a 5870 provided CF is working. Guess I should be looking for a ~40% increase going by those figures.

 

[straxusii]

Oh well, don't know how good your maths skills are but according to my Excel spreadsheet, GTX580 is about 36% faster at this resolution. Are you sure you included games only?

Here's the thread: http://forums.overclockers.co.uk/showthread.php?t=18205746

I didnt have any agenda with it so no point the various fan boys going off on one, it was purely for my own use to see if I thought it was worth it so no point saying ah but you didnt include XYZ, I included what I did for my own reasons as detailed.

 

[Mr Krugga]

 

[drunkenmaster]

Sorry but here lies the problem, your graph says xfire isn't working in Call of Juarez, it does work, it works fantastically, its 70% faster than a 480gtx, and 50% faster than a 580GTX in that review.

You only need to look at the 10-15 reviews they INCLUDED the 5970 results in the review, with the SAME drivers, then notice the 10-15 reviews they've excluded that SINGLE result dramatically bringing down the difference in performance.

I've linked numerous times to techpowerup's OWN results with the 5970 in that game, and the result, and where they stopped showing that result and the dramatic drop in difference they record between the cards as a result.

480gtx launch, vs 480gtx results 6 months later, across all games, 1-2% improvement AND loss in some games. 480gtx launch, the 5970 is 38% faster on average, 6 months later, its 14% faster, despite the same fps results in all games, just with several of the bigger 5970 results "removed".

Anandtech's results aren't completely representative as they only use the 5970 at higher resolution results, with all results combined it would show a massive lead even at 1920x1200, even higher lead lower down. Its lack of memory and several games at the extreme res/aa mean some results are miles ahead of the 1.5GB 480gtx, and some are miles behind, if these were averaged, or only "average" resolution used, lets be honest 99% of 5970/cf/sli 580/480 users are using a 1920x1200/1080 screen, as I am.

Hardware Canucks clearly cheat, I mean clearly because they claim they are using one driver, while using results identical in every detail to results 6 months earlier, they also use the single worst driver AMD have produced in the past 5 years, one where xfire sucks more than usual.

http://techpowerup.com/reviews/NVIDIA/GeForce_GTX_580/13.html

yet Far Cry 2 is barely 10% ahead, averaging across all resolutions maybe around 5%, yet one review has the 580gtx 20% ahead?

Hmm, even worse, look at the actual numbers, percentages, Techpower up has, in 20 tests, the 580gtx ahead in FIVE tests, never more than 10%, the 5970 is ahead in 15 tests, most between about 7% and 30%, theres no way in hell that averages just under 10% faster than the 580gtx.

For the 5 games that the 580gtx is ahead, you're looking at an average of 5% faster, for AMD's 5 best games, its averaging what, 25% faster. The 5970 is simply unquestionably the faster card.

Review sites are BS, their numbers are often complete bull, techpower up are flat out cheating with removing the best results, add in the 50% faster for the 5970 in call of Juarez and it will be averaging 25% faster or so, and on 80% of the planet, the 5970 costs less at the moment, its $100 cheaper in the states.

 

[^Andy^]

Bear in mind that HardOCP results are not representative (as for comparison) as they use higher resolution and AA levels.

Nice work and thanks for taking the time to compile that!

 

[Mr Krugga]

Sorry but here lies the problem, your graph says xfire isn't working in Call of Juarez, it does work, it works fantastically, its 70% faster than a 480gtx, and 50% faster than a 580GTX in that review.

Well not really, according to this review, it doesn't. I have no time to make analysis of the results, feel free to do it yourself if you have enough spare time.

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_580/9.html

You only need to look at the 10-15 reviews they INCLUDED the 5970 results in the review, with the SAME drivers, then notice the 10-15 reviews they've excluded that SINGLE result dramatically bringing down the difference in performance.

They did what they did, I won't fiddle around with the results because soon enough, I'll have a bunch of Nvidia trolls on my head. TechPowerUp failed miserably for me when W1zzard claimed to not have heard any info on GTX580 from Nvidia, not being invited to their presentation, not get a sample of the card, etc. and yet he managed to release this review before NDA was lifted.

I've linked numerous times to techpowerup's OWN results with the 5970 in that game, and the result, and where they stopped showing that result and the dramatic drop in difference they record between the cards as a result.

So they're using old results with other games and a made up pack for CoJ 2? Bunch of trolls

480gtx launch, vs 480gtx results 6 months later, across all games, 1-2% improvement AND loss in some games. 480gtx launch, the 5970 is 38% faster on average, 6 months later, its 14% faster, despite the same fps results in all games, just with several of the bigger 5970 results "removed".

Don't know what to think of it but anyone can do their own validation of the results. As long as several reviews show similar pattern, it's all good by me. I never based my opinions on a single source of information.

Anandtech's results aren't completely representative as they only use the 5970 at higher resolution results, with all results combined it would show a massive lead even at 1920x1200, even higher lead lower down. Its lack of memory and several games at the extreme res/aa mean some results are miles ahead of the 1.5GB 480gtx, and some are miles behind, if these were averaged, or only "average" resolution used, lets be honest 99% of 5970/cf/sli 580/480 users are using a 1920x1200/1080 screen, as I am.

Not sure what you mean as AnandTech seem to use 1920x1200 with 4xAA in that review. Which is what most of the enthusiast use these days, right?

Hardware Canucks clearly cheat, I mean clearly because they claim they are using one driver, while using results identical in every detail to results 6 months earlier, they also use the single worst driver AMD have produced in the past 5 years, one where xfire sucks more than usual.

They're a joke but I still included their results as it was linked in the GTX580 thread.

http://techpowerup.com/reviews/NVIDIA/GeForce_GTX_580/13.html

yet Far Cry 2 is barely 10% ahead, averaging across all resolutions maybe around 5%, yet one review has the 580gtx 20% ahead?

Different reviewers, different settings and setups. I don't mind if you investigate it further. It's 10% btw, 5% would be irrelevant as I included benchmarks at 1920x1200 only (with the exception of HardOCP review).

Hmm, even worse, look at the actual numbers, percentages, Techpower up has, in 20 tests, the 580gtx ahead in FIVE tests, never more than 10%, the 5970 is ahead in 15 tests, most between about 7% and 30%, theres no way in hell that averages just under 10% faster than the 580gtx.

Oh well, I just used a calc to check the numbers manually and it looks right. Show me your maths if you want to prove me wrong though. Bear in mind it's an average, not total of all scores.

For the 5 games that the 580gtx is ahead, you're looking at an average of 5% faster, for AMD's 5 best games, its averaging what, 25% faster. The 5970 is simply unquestionably the faster card.

I think you're getting overly excited over this, there are 3 games that GTX580 pulls ahead of HD5970 and it does it by 10-13%. Dunno about Call of Juarez 2 score thus it's not included in the average.

Review sites are BS, their numbers are often complete bull, techpower up are flat out cheating with removing the best results, add in the 50% faster for the 5970 in call of Juarez and it will be averaging 25% faster or so, and on 80% of the planet, the 5970 costs less at the moment, its $100 cheaper in the states.

Actually both cards can be had for $499 in the US, HD5970 for $479 after mail-in rebate but it's only one auction on ebay and one etailer selling so cheap in the limited promo. The price will raise once again soon as most retailers sell it for over $500.

Regarding BS, we could say that about most info posted over the internet, your maths for example is flawed as hell.

 

[Mr Krugga]

Nice work and thanks for taking the time to compile that!

Please, don't quote the entire table.

It seems that I made a mistake and put Hard OCP results for HD5870 2GB under 5970 columns. Fixed and added ***** results.