NVIDIA GeForce GTX 580 1536 MB

[CDdude55]

damn man seriously, its not even 15% faster, damn man the naming for these cards is getting really screwed up, wtf amd and nvidia

at least it uses less power

It really should of been named ''GTX 485'' or something of the sort, as it's still based on the same architecture and process, just with more efficacy added in.

 

[AFQ]

congrats. you are the first person to EVER complain about that. i've had people complain about pretty much everything else, except for vantage. vantage is gay, wont use. maybe 3dmark 11 is better

lol, any idea when 3DMark 11 will release? You must know better than us.

and can you please tell how you bypassed the voltage limitation?

 

[douglatins]

Performance wise its only 15% better, but it has better thermals and load

 

[W1zzard]

and can you please tell how you bypassed the voltage limitation?

you mean the power limiter? i just measure power at a high sample rate that is faster than nvidia's system so i can measure before the limiter can kick in.

technically it's not a voltage limitation. voltage is pretty much constant on the 12v line but current (amps) change.

power (watts) = voltage * current, thats why nvidia also measures voltage, multiplies (in the driver) and gets the power draw by that formula

 

[AFQ]

you mean the power limiter? i just measure power at a high sample rate that is faster than nvidia's system so i can measure before the limiter can kick in.

technically it's not a voltage limitation. voltage is pretty much constant on the 12v line but current (amps) change.

power (watts) = voltage * current, thats why nvidia also measures voltage, multiplies (in the driver) and gets the power draw by that formula

Ah...thanks for clarification.

 

[mdm-adph]

Still uses more power than a 5970 while being slower. Fail.

In order to stay within the 300 W power limit, NVIDIA has added a power draw limitation system to their card. When either Furmark or OCCT are detected running by the driver, three sensors measure the inrush current and voltage on all 12 V lines (PCI-E slot, 6-pin, 8-pin) to calculate power. As soon as the power draw exceeds a predefined limit, the card will automatically clock down and restore clocks as soon as the overcurrent situation has gone away.

Why, those tricky devils.

 

[qubit]

as mentioned before, nvidia's response to the feedback in that thread was that they tried to get me a board on the weekend and managed to achieve the impossible!

I'm sorry, I didn't see your previous response and didn't know.

Anyway, that's mighty awesome! You must have burned the midnight oil to get this review done by the deadline? Kudos dude.

 

[btarunr]

Still uses more power than a 5970 while being slower. Fail.



Why, those tricky devils.

But we managed to put those devils to sleep (disabled that throttling logic). Check out the "Maximum" graph for Furmark power draw with throttling logic disabled.

 

[mdm-adph]

But we managed to put those devils to sleep (disabled that throttling logic). Check out the "Maximum" graph for Furmark power draw with throttling logic disabled.

Aye, will check it out. It just always freaks me out when a company directly targets reviewers and benchmarking programs in that way, though. Too close to home.

 

[douglatins]

the card throtles itself on Furmark? Lame

 

[btarunr]

Their contention is that since Furmark/OCCT don't reflect "real-world usage" (hear hear, like racing doesn't reflect real-world usage of "cars" i.e. driving from office to home), they decided to play nanny with a supposedly enthusiast-grade product. It's kind of like fitting the speed-governor device used in school buses in a Porsche 911.

 

[mdm-adph]

Their contention is that since Furmark/OCCT don't reflect "real-world usage" (hear hear, like racing doesn't reflect real-world usage of "cars" i.e. driving from office to home), they decided to play nanny with a supposedly enthusiast-grade product. It's kind of like fitting the speed-governor device used in school buses in a Porsche 911.

If it's not "real-world usage," then they should ignore it. Nobody is going to see the Furmark results besides us enthusiasts anyway.

Doing something like this makes it look like they're trying to hide something. Since the limiter only kicks in during either Furmark or OCCT, couldn't more power be conceivably drawn during some really, really, really heavy gaming? I mean something really intensive?

 

[ToTTenTranz]

Card is good, pricing isn't so bad either.
But GTX 580? So from now on, nVidia is going to change the family number denominator twice a year? By 2013 we'll have GTX 1180?


I guess 2010 is by far the year with the crappiest naming conventions for graphics cards, ever.

 

[Hayder_Master]

glad to see your review w1z , dig it guys

 

[jfgwapo]

Nice review, as always.

Nice performance gain vs 480, good power reduction too.

 

[wahdangun]

wow, its just fully enabled GTX 580,


btw where is bastenegia ?? where is 128 TMUS he claimed ?

 

[Imsochobo]

Looks good actually. On par with the 5970, better power consumption than the 480.. Nice.

Let's not forget that.

a.\ its almost a year since 5970 came out.
b.\ not beating 5970 in power usage, per/watt

But its great news!
Nvidia have had lack of products that compete with with AMD.
460 can compete, and 580 are their strongest ones.
But they're fighting a whole very good lineup from amd though.
maybe my 5850 will go under launch price now ? instead of over...

 

[Animalpak]

finally my new directx11 video card is out.

 

[trickson]

Thanks for the nice review W1zzard !
Man they are pumping out video cards like mad lately . I just got crossfire with 5770's . I feel like I need to ask bill gates for some help just to keep up with all the new tech coming out every week it seems like some thing new is coming out to make every thing I have obsolete !

 

[alexsubri]

great review W1zz, at least its almost in par with 5970

 

[N3M3515]

The price is nothing we can do about it, after all it is the fastest SINGLE gpu card.
I'm disappointed anyway, as i was especting at least 20% average increase over GTX480.
On single card side of things, still 5970 rules, performance, power comsumption and price wise.
This card is on average 24% faster than 5870, so, all the 6970 has to do is be 30% faster(than the 5870) and take the single GPU performance crown from GTX580, which i think is very probable.
Then, 5970 undefeated, passes its crown to 6990.

 

[newtekie1]

It really should of been named ''GTX 485'' or something of the sort, as it's still based on the same architecture and process, just with more efficacy added in.

Again, this comes down to ATi moving to the next generation first with their cards that are still based on the same architecture and process. The same reason nVdia had to move to the 9800 series when they had planned to release everything under the 8800 series. In marketting it would seem like nVidia is a generation behind if ATi is roling out their "next gen" cards and nVidia is still putting out GTX400 series cards.

Their contention is that since Furmark/OCCT don't reflect "real-world usage" (hear hear, like racing doesn't reflect real-world usage of "cars" i.e. driving from office to home), they decided to play nanny with a supposedly enthusiast-grade product. It's kind of like fitting the speed-governor device used in school buses in a Porsche 911.

While I somewhat agree, I also see that ATi has done the same thing with previous cards. Except they did it too late as an afterthought when people started killing hardware with Furmark. Personally, I don't think Furmark should be used for anything more than testing stability(and now it can't even do that). So use Kombuster instead. Power use under Furmark is completely un-realistic. Using your anology, it is like judging the fuel mileage of a race car based on 100% full throttle the entire race, it just doesn't happen. That is why when I look at power consumption, I always look at the Peak graph, because that is real world.

Still uses more power than a 5970 while being slower. Fail.



Why, those tricky devils.

Yes, but it is an unnoticeable difference, and for almost $100 less. I'll take it.

 

[largon]

Too little too late.
I reckon Cayman will be one tough adversary for GF110.

 

[Imsochobo]

Again, this comes down to ATi moving to the next generation first with their cards that are still based on the same architecture and process. The same reason nVdia had to move to the 9800 series when they had planned to release everything under the 8800 series. In marketting it would seem like nVidia is a generation behind if ATi is roling out their "next gen" cards and nVidia is still putting out GTX400 series cards.

Where is ati the same arch from 5000 to 6000
Its rather diffrent, how can a 1120 shader card trumph my 1536 shader card.

 

[johnnyfiive]

BTW guys, I think minimum frame rates are being overlooked... by a lot of you guys.
For each person saying the card is inferior to the 5970, that its too late, that its not really worth the $559 price tag... you're wrong. It beats the 480, by a decent mount, same with the 5870. It performs as well in most cases, as the 5970.., and the kicker, it has a equal or sometimes better minimum frame rate as the 5970, has CUDA, and much improved tessellation performance.

So what am I missing that the majority of you guys are bashing on?

http://www.hardwarecanucks.com/foru...s/37789-nvidia-geforce-gtx-580-review-11.html

Look at those minimum frame rates fellas. I love the reviews on TPU, but minimum frame rate really should be added to reviews. It makes a big difference. Max frame rate is nice to know.. but its not nearly as important as the minimum fps rate. Think about it, if you have a 60hz monitor, is having 120 max, 55 min better than having 110 max, 65 min? The 580 is awesome for being single GPU.

And you can't tell me this isn't impressive...

As for it being more expensive than a 5970... some 5970's are $600, and its OLD architecture. It's just two 5850's basically. For being a single GPU, improving Fermi architecture, improving thermals and power consumption... this is a great card for $559.