GTX480 / GTX470 Reviews and Discussion

[notty22]

When the AA/AF is cranked up , it takes a 5970 to beet Fermi, and IQ is why enthusiasts buy these cards.

 

[OvrClkr]

I just pulled the trigger on the Gigabyte 470 (Newegg), I should be getting the card between Thur/Fri of this week. Hopefully I will have some numbers for you guys by Friday evening after work.

 

[builderbobftw]

Since when does the 4890 compete with the GTX 285?? AFAIK the 4890 is on par with the GTX 275 give or take.

Read the reviews builder 😉

http://www.techreport.com/articles.x/16681/13

The 4890 gets real close to the 285, Dominates thwe 275.

And now with new 10.3 drivers, 4890 is MUCH faster than the GTX 285.

They were withinn a few % even on old drivers.

 

[AMW1011]

This thread makes me hate life...

Look here:
http://www.overclockersclub.com/reviews/xfx5850/6.htm

A 5850 overclocked to 1GHz on the core, a 38% overclock, gains 12-18% extra performance.

A GTX 470 is 10% faster stock for stock. Even at 800 core, a 180MHz overclock, vs a 1GHz 5850 the GTX 470 should still be faster. You guys are acting like the clock speed percentage is directly related to the performance increase.

A 5850 should be able to make up some ground when OVERVOLTED, something that is very dangerous and will destroy your cards life span.

Oh and people comparing this to Fermi's heat need to learn how this all works, temperature is far less dangerous than voltage. If you keep a GTX 4xx card under 100c you are fine, the max temp is 110c where it starts throttling, because nVidia used a compound in their dies to allow it to survive more heat. This is why a CPU will fry at 80c, but most high end cards load at 80c+, its the design. Voltage will kill anything, regardless of temperature. I'm not saying a 5850 @ 1.3v will die out with in a year, but it certainly doesn't have a higher life expectancy than a GTX 470 at stock volts kept in the temperature range. Look how many GTX 295s, 8800 GTs, and 9800 GX2s, 3870s, 3850, 3870 X2s, 8800 GTXs, 8800 Ultras, 1950 XTs still survive to this day. Compare that to highly overvolted cards.

Oh and many 5850s can't do 1GHz on the core, about half the people I've seen overvolt their cards can't hit 1GHz on the core.

 

[builderbobftw]

When the AA/AF is cranked up , it takes a 5970 to beet Fermi, and IQ is why enthusiasts buy these cards.
http://images.tweaktown.com/content/3/2/3230_33.png

To Beet Fermi?

 

[AMW1011]

deleted

 

[notty22]

To Beet Fermi?

http://www.fox.com/areyousmarter/pdf/contestant-app.pdf

Theres your prize. You would do wonderful there.

Charlie over at SA is trying to make a big hullabaloo about the open gl drivers. Nvidia released them
and has got Charlie all worked up with revision 2 and 2a of his story. This is about the biggest non-issue EVER.

 

[RealityRush]

607 out of 725 is about 84%

If 5850 clocks to 1100 with voltmods - 470 would have to reach ...... wait for it ...... 924Mhz. Now that is something impossible on air I can guarantee.

Notice this : 480GTX was clocked at 800 on 100% fan reaching 83C. This is only a mild 14% overclock when you consider its stock is 700.

Now how much you think you will get with voltmod 😀. if you just up the voltage with a notch and still run at 800 you will get more heat as more voltage is pumping in, so you will be already in 85C or more.

I really doubt if you will be able to get to 850 even with voltmod, you will just get 100C on 100% fan and I bet you cant game at these settings for like 2-3-4h. It might run tests but try gaming a whole day, you will melt some case metal there

So no, I dont think OC Fermi = OC 5000. ATI have much more headroom.

However I dont think lot of people overclock, let alone so agressive so this probably matters to a very small number of enthusiasts.

What matters is stock vs stock and 470 beats the 5850 hence the price is a little more. Same for 480.

1100MHz my arse. Most of the 5850's I've seen OC to 1GHz on air with upped voltage.

At stock speeds, a 470 is 118MHz slower than a 5850, and it is still almost 10% faster. An OC'd 5850 would have to be like 300MHz or more faster than a 470 just to break even with it. If you can OC a 470 to even 800MHz on air, a 5850 would still lose to it OC'd on air.

And normally I don't like to agree with notty as he seems pretty polarized, but he's right, nVidia ends up handling AA much better than the ATI cards right now, and tessellation. So a 5850 would have to be OC'd even more when DX11 and AA comes into play to match a 470, a LOT more.

I'm not hating on the 5850, it is a great card... but it isn't that much better at OCing than a 470 on air.

You guys are acting like it can OC light years ahead when it really can't performance wise. Strictly MHz wise.. yes it can, but in the end performance is what matters and a 5850 is not going to beat a 470.

 

[AMW1011]

Exactly reality, you hit the nail on the head.

 

[OvrClkr]

The 4890 gets real close to the 285, Dominates thwe 275.

And now with new 10.3 drivers, 4890 is MUCH faster than the GTX 285.

They were withinn a few % even on old drivers.

Show me a site that puts the 4890 ahead of the 285 "OVERALL" ...

In a few games I do see the 4890 ahead but overall most sites put the GTX 285 clearly ahead.

Same goes for Fermi, the only game where AMD is clearly ahead is in Crysis, and that is a 3 year old game that is gaining irrelevancy vs. more modern games.

 

[OvrClkr]

Here is an interesting chart :

http://img39.imageshack.us/img39/2112/gfxcomparerev2smallpng.jpg

 

[SV_Bubbles]

From the umpteen million benchmarks I've come across, the 275 has been about 5-10 percent faster. Double in some situations though looses minutely on certain games.
At first the 4890 was very close and then over the months the 275/285 pulled away a bit.

 

[RealityRush]

A guy on Overclock.net is at 815/1675/1895 on a 470! His benchs are higher than a stock 480. Another said his 480 got to 925/1850/2000 but is too hot without more cooling. Pretty insane. Both with voltage tweaks.

http://www.evga.com/forums/tm.aspx?m=299325

I didn't verify the validity of these, but just for info.

Some guy on there say he got 840MHz core on a 470 on air without voltage tweaking as well, interesting...

EVGA needs to hurry up and release some voltage modding software asap, I want to see what these cards can do with higher voltages >.<

 

[TheGreatGrapeApe]

Here is an interesting chart :
http://img39.imageshack.us/img39/2112/gfxcomparerev2smallpng.jpg

Posted on 'page 2' by SV_bubbles;
http://www.tomshardware.com/forum/page-284770_33_99.html#t2110084

[:jaydeejohn]

Along the lines of the discussion at hand a more interesting article by Xbit on voltage, overclocking and power consumption;
http://www.xbitlabs.com/articles/video/display/gpu-power-consumption-2010_4.html#sect0

That's without a serious hardware mod, just the software voltage tweak and stock air. 1Ghz.

Anywhoo, it's an endless discussion as to which OC's with what on what, which is one of the reason long ago the discussion of OCs in comparison for entire lines was removed since they differ so much even within a line.

 

[TheGreatGrapeApe]

But that aside, are you seriously suggesting that the links you provided point towards irrefutable evidence of shady pay offs?... Are they wrong to say that the massive exageration is unjustified?

I didn't say payoffs, but you asked if they were ignoring the issue and actively trying to bury it. Those are examples of them sloughing it off, and actively going to sites to have them re-do their investigations, many of which actually mention that nVidia contacted them after their reviews. Do you honestly think this was a spontaneous 're-evaluation' by reviewers, essentially for the first time not related to IQ issues? [:grahamlv:3]

C'mon it's pretty straight forward, nV asked, they were accommodated. Did money change hands? I doubt it, but by the same token, they actively went out of their way, especially with the Czech review which twice had nV's involvement in the request and promotion. So it pretty much fits the request you asked for very simply.

You seen to have alot of issues with the way other people do their thing, be it criticism of a corporate entity, or your idea of what a moderator should do, and seem fine with criticizing them in turn.
Essentially expressing that people should treat the paid engineers and executive as if they are fragile children in need of ego stoking/stroking, or else they would withhold their gifts to mankind (and thus becoming bankrupt and being bought for 10c on the dollar for others can profit from their genius instead). You seem to lose focus that the products and companies must work to meet customers needs, not vice versa.
If engineers or the executive leadership team prefer praise and hugs instead of compensation, then they can give away their product and compensation, and people will give them a tip of the hat for the effort. However most would prefer the money, and as such the consumer and shareholders have every right to criticize poor results. Especially when the leadership engages in over-inflating their promises at the expense of the competition (nV also did this for the FX5800, ATi did this for the HD2900). Similar to these two products they get criticized for not delivering what the company/PR promises, and conversely like I said those that perform as expected or outperform (like the R9700 and GF8800) get praised. Success is rewarded and praised (an even emulated), failure is criticized and often punished by sales or shareholders, that's the way of market, and it helps ensure that things move forward not stagnate. It's not like companies are going to stop trying simply because of one or two successes or failures, if they do stop they will simply be passed by others who will get rewarded for their successes while those standing still with get punished for lack thereof.

That's good but right back at you there, you would have lead others to think it were a fair comparison by bringing it up as an argument where it was irrelevant. Opengl acceleration so far has been mostly for cosmetic and interface responsiveness enhancements while the cuda implementations are rumored to increase performance by more than a full order of magnitude. Allowe me to remain unimpressed that opengl enhancements were even mentioned.

Once again, you're misreading what I wrote attributing other factors just like the 'pay-off' comment. I'm not comparing the built in OpenGL acceleration, nor even the better OpenGL plug-ins, however I'm also not going to say they have support for something they don't. That's what BOTH sets of cards currently have access to and I'm not telling anyone which path to go, but I'll make sure they know what is and isn't the case, not make up support for something that's not there. We already had that a few times in the past with things like AVIVO which promised more than they delivered.

Also, I don't have brand loyalty per se, I simply happen to favor whoever supports innovation and all attempts to further branch off today's boring status quo and meander somewhere that'll end up much more interesting.

So are you saying now that that development is one sided between the two options?
If the answer is as you've portrayed it here by the selective promotion, then you're really not looking at the whole picture from all sides, and also ignoring Adobe's own statements about the future development of CS.

Now onto this side-bar issue which really has no place, but since you bring it up, to avoid it would be to pretend it has merit.

Here, several people have expressed discontent at the pace, meaningless content and general amount of belligerence in posts looking at the board through ati-colored glasses in this one thread alone yet, not a word was said about it.

Words have been said and actions taken about the pacing and the meaningless content, a few times by myself, however should I simply regulate all the meaningless content then your erroneous posts would simply have been deleted now wouldn't they? 😛
As for the 'ati-coloured' glasses, how is that any different than the counterbalancing nVidia-coloured glasses? Seems both side pervade the forum and the threads therin (just look at the HD5870 thread or others). If you don't see that, that's likely due to your own discolouration.

I don't mean to question the way you do your job here but why are you ignoring them?

You do mean to do so and in so doing try to use it to discredit the other comments, or else you wouldn't mention it with regards to our discussion, which has nothing to do with moderation. There are other moderators here, including the OP, yet you decide to comment on it in this reply and the last one simply as a tool, not as genuine concern or interest. I don't ignore them, and you will see sofar I have probably involved myself more than other moderators in trying to keep people on topic, however I am not their nanny, and I limit it to the topic and staying within my bounds of their acceptable conduct. All of which has nothing to do with this discussion, other than for something for you to criticize in an attempt to tie it to the other points in the discussion.

Should I cut elie off for trying to support his point, or Izzy for trying to support his? Would that be taking sides?

These discussions have weight and merit, as long as they remain somewhat civil and not devolve into yes / no without support, then they expose more information from all sides trying to support their position, which bings more information to everyone. And between the two side of the discussion likely lies the 'truth', or something resembling it, somewhere in the middle.

But like I said before I don't see how that moderation role has anything to do with the discussion, and it's usually brought up when people don't have anything else intelligent to say and resort to that instead.

Seems at least that should've warranted it's own separate post if it was a true concern and not simply an attempt to shape the conversation.

 

[mindless728]

At stock speeds, a 470 is 118MHz slower than a 5850, and it is still almost 10% faster. An OC'd 5850 would have to be like 300MHz or more faster than a 470 just to break even with it. If you can OC a 470 to even 800MHz on air, a 5850 would still lose to it OC'd on air.

actually it would be a percentage

(clock speed ratio)*(increased performance) = (percentage clock speed 5850 needs to equal GTX 470)
(725/607)(1.1) = 1.314

so GTX470 at stock would be equivalent to the 5850 at 798 MHz

and GTX470 @800MHz, i would be a 5850 @1051MHz

 

[Guest]

197.45 WHQL up to 30% SLI Crysis! 13% Single GPU [:mohsentux:9]

Looks like nVidia are doing their best to improve drivers as they couldn't clock the cards higher.

@mindless728: You 4got the memory oc on both cards. 1300 aka 5200 for the 5850 up from 1000, 470 idn whats the value here.

For this round I chose ATi as they were the first and had good fps/$ since launch, sure now with Fermis drivers, it is degrading but it was solid for 8 months which is great fro a video card. My next upgrade will be in 2011, Fermi refresh Vs ATi 6000, there should be preview topics here so we can argue again [:studiocjf]

 

[scirishman76]

Here is my take on it I took a large compilation of game benchmarks run @1920X1200 with 4Xaa(just games no unigine or 3dmark) as this is where I will run.At those settings the GTX480 came out over 20% ahead overall.Also it seems like close to 40% of new games and upcoming games use PhysX and many upcoming games are supposed to use tessellation.The 480 has PhysX and beats the 5870 in tessellation by more than double.I definitely plan on playing Metro 2033 and the 5870 can't even play it with tessellation and 4xmsaa enabled where the 480 can produce playable rates.

Also despite what everyone thinks it seems the 480 O.C.'s better also and sees more of an improvement per overclock percentage.5870 averages 900-925 vs 850 core and 1300 vs 1200 memory that add up to about 9% O.C. on the core and 8% or so on the memory.The GTX 480 is said to OC to 800-825 vs 700 core and a whopping 5000 memory vs 3690 stock a 14% core O.C. and a 26% memory O.C..I know what you guys are gonna say that voltage modded 5870's are hitting over 1000 however already have voltage modded 480's hitting 875 which is almost exactly the same ratio if not even more in favor of the 480.

I am not as concerned with power consumption as I am with performance so I don't consider it a real issue for me.Heat may be a problem for some as it heats up the case the GPU itself is in no real danger this GPU will not see any damage until it gets over 120C and it is set to throttle back power at 110C.Most of the reviewers used the auto fan speed and were hitting 90C sometimes but reviews where they set the fan to 65%-70% at idle and 85% after a few minutes if load said they were only seeing 78 to 83C which is in the ballpark of alot of GPU's. Most users that actually have one also say it's not nearly as loud as reviews are making it out to be(most are saying about as loud as a gtx260)and 80 to 85% is the sweet spot as it doesn't seem real loud at this speed.I have 3X140mm fans(one right on the gpu)and 3X120mm fans so I think I will be fine.

All I can think if is how so many hd4000 series users talked about how much money they saved.Then within 6 months most of the serious ones bout a dedicated PhysX card(usually 8800gt)and ended spending way more money and using way more power versus if they had just gone Nvidia in the first place.Now many 5870 users are adding a gts250 to run dedicated PhysX at which point it becomes more expensive and draws about the same power as a gtx480.

What FPS benchmarks don't tell you is that in games distant views almost always look better on 480 a fair amount of games use PhysX and will look better on the 480 and in tessellation games it looks way better.

So will I pay $100 more for 20% better FPS better overclocking and better looking game play HELL YES!

 

[Guest]

Well physx isnt made for every one. If you find it is great then why not. The 480 Cooler is better than the 470 and @ 100% fan it should be enough but overvolted idn.

Personally, I never will let my GPU or any electric component go over 100C. Fermi throttle @ 105C

@notty, Far Cry 2 has always favored the GTX cards , even a 470 beat a 5870 in this game.

@scirishman76 The 480 @ 875 was under LN2 and not mention if hard modded, I doubt the air cooler will keep up when overvolted.

Anyway, considering the Driver boosts, thers no need for OC for the GTX cards right now.

 

[scirishman76]

Well physx isnt made for every one. If you find it is great then why not. The 480 Cooler is better than the 470 and @ 100% fan it should be enough but overvolted idn.

Personally, I never will let my GPU or any electric component go over 100C. Fermi throttle @ 105C

One of the main points that didn't come through real strong is that many reviewers did many tests with no aa enabled.The gtx 480 pulls further ahead with AA around 20%.Now honestly who buys a card like the 5870 or gtx 480 and doesn't use AA?

 

[welshmousepk]

All I can think if is how so many hd4000 series users talked about how much money they saved.Then within 6 months most of the serious ones bout a dedicated PhysX card

you lost all credibility with that stupid statement.

the fermi cards are good, i still maintain that they are not worth the price premium and power/heat issues, but i certainly do not think either are bad cards, and would not put anyone off buying one (assuming they have the cash and good case ventilation)

but Physx was, and still is, an absolute gimmick. its totally worthless, and nobody in their right mind would buy a dedicated physx card. many people use old nvidia cards they have lying around for that job, since they are worth so little when sold on. But to buy one so you can get some superfluous additions to a couple of games (and a few more crappy games that no one plays) would be pure idiocy.

there was much truth in your post, but it was also extremely biased, and the Physx stuff pushed it into fanboy territory.

 

[Guest]

Ya I know, when AA 8xx+ is applied the gap is 20% between 5870 and 480. I dont use AA as my monitor is 21inch@1080p, textures are so tiny to see them lol The 400 has improved AA mechanism that was weak in previous generation.

oh not again phyx lol, it has been said 999x time, OpenCL and DirectCompute that came with DX11 will replace PhysX as they are faster and free open source to developpers and interchangeable between all GPU makers DX11 compliant, much easier to code than PhysX.

 

[OvrClkr]

This is a very interesting thread about both the 470 and 480.

http://www.overclock.net/nvidia/707613-new-830mhz-benches-gtx-470-unboxing.html

 

[scirishman76]

you lost all credibility with that stupid statement.

the fermi cards are good, i still maintain that they are not worth the price premium and power/heat issues, but i certainly do not think either are bad cards, and would not put anyone off buying one (assuming they have the cash and good case ventilation)

but Physx was, and still is, an absolute gimmick. its totally worthless, and nobody in their right mind would buy a dedicated physx card. many people use old nvidia cards they have lying around for that job, since they are worth so little when sold on. But to buy one so you can get some superfluous additions to a couple of games (and a few more crappy games that no one plays) would be pure idiocy.

there was much truth in your post, but it was also extremely biased, and the Physx stuff pushed it into fanboy territory.

Go to Newegg look at the 5870-5970 reviews you would be surprised to find how many people bought an Nvidia card to run for dedicated physX I'm not making it up.If its worthless then alot of people wasting money.

 

[Guest]

It is worth it only if the card your getting is <99$ and the user really want the extra effects, paying more than that is not justified.

Asus OverVoltage for the GTX Cards but the limit seems low.

Nice find. Now he needs the new Drivers to boost his scores 😀