GTX 480: Your verdict

Your verdict on the GTX 480

  • Total voters
    553
Status
Not open for further replies.
alg7_munif said:
If GTX 470 gets 4, what would get 1 for you? My verdict for GTX 470 would be 1 since it loses in every single way.
Click to expand...
I am not the guy, but I will answer for him: another rebranded card with software DX11 emulation would justify 1.
 
vjcsmoke said:
The performance is comparable. It just loses out on the noise, price, and power consumption front.
So if it pushes the FPS at a comparable rate, the tech isn't that bad, it's just not as good a VALUE as the 5850.
And it certainly wouldn't be considered an upgrade over a 5850 since it came out 7 months later and is still NOT appreciably faster.
For it to earn a 1 it would have to be considerably slower. For me a 1 would mean that i wouldn't even use it for anything but a very expensive paperweight.
Click to expand...

Well I guess that the scale doesn't match for both of us. :eek: I would consider 6 and above being positives while 5 and lower are negatives.
Xzyrus said:
Watching a video is no way accurate to judge sound since it'll depend a lot on how loud I crank my speakers. I'd rather just trust sound measurements from the same site (assuming they keep the same methodology).
Click to expand...
As long as you keep the volume at the same level, you can compare the cards relative to each other. Btw dB scale is a logarithmic scale and cannot be compared directly.
 
Fermi is fucking horrible and that makes me sad.

Marginally faster in some games, slower in others, very hot, shit loads of power usage, Cost more, needs 2 cards to run multimonitor (not to mention after market coolers to keep from frying them), and very late to the game.

According to [H] they ran 93c under full load. WHAT!? That is unacceptable.

These should have blown the 5000 series out of the water at this far into the game.

Nvidia, you guys SUCK this time around. Please go back to the board and release some worthy hardware.

Hell, ATI could sit on their refresh now and suck up all the earnings from current hardware that sells for way above MSRP. Sad day
 
McCartney said:
"Moving on, we have the GTX 470 to discuss. It’s not NVIDIA’s headliner so it’s easy to get lost in the shuffle. With a price right between the 5850 and 5870, it delivers performance right where you’d expect it to be. At 5-10% slower than the 5870 on average, it’s actually a straightforward value proposition: you get 90-95% of the performance for around 87% of the price. It’s not a huge bargain, but it’s competitively priced against the 5870. Against the 5850 this is less true where it’s a mere 2-8% faster, but this isn’t unusual for cards above $300 – the best values are rarely found there. The 5850 is the bargain hunter’s card, otherwise if you can spend more pick a price and you’ll find your card. Just keep in mind that the GTX 470 is still going to be louder/hotter than any 5800 series card, so there are tradeoffs to make, and we imagine most people would err towards the side of the cooler Radeon cards."

Why are you lying munif? These metrics by anandtech CLEARLY show that the card is better for price/performance. I am going to report you for spreading propoganda. This isn't fair. If you conduct these polls, do so without bias. I will not allow you to misrepresent your favorite brand's competition because you feel you have the right.
Click to expand...

And the results from [H] CLEARLY show that the 470 is slower in each game tested than the 5850. Anandtech uses canned benchmarks, [H] uses actual gameplay. So great, the 470 can bench faster than the 5850, but the 5850 is faster playing games while using a shit ton less power.
 
7.5/8

Preordering my 480 and getting ready to Wc it. This is why I have a Corsair 1k psu. Why not put it to good use. Noise? Haha my current 9800GTX laughs at noise, Heat? WC time.

Still not as good as I would think but still a very good card and we will see more of it later on.
 
I voted a 5, it's mediocre at best compared to ATI and ATI's offering has been out for several months now. Nvidia has had plenty of time to do what they could to get ahead, and they have not done so. Their offering is just barely competitive with ATI's offerings.

Nvidia's product just isn't as compelling as ATI's when you factor in the price to performance ratios, seeing as the 5850 smashes the 470, rendering the 470 irrelevant in this phase of the card game.

Tap 2 blue mana, summon Fail
 
kllrnohj said:
And the results from [H] CLEARLY show that the 470 is slower in each game tested than the 5850. Anandtech uses canned benchmarks, [H] uses actual gameplay. So great, the 470 can bench faster than the 5850, but the 5850 is faster playing games while using a shit ton less power.
Click to expand...

And real gameplay guarantees the amount of polygons rendered for each video card are identical, right? WRONG.

You cannot optimize hotpaths on a compiler for a video card due to "canned benchmarks", that is an absolutely FALSE argument.

You're WRONG, pal. A "canned" benchmark guarantees the same control group for the 2 variables being tested. Do you honestly think testing without a control group is a valid, scientific way to show that one card is better than another?

No, it isn't, and it never was. I highly suggest you prepare your next argument as I'm currently doing compiler optimization with Machine Learning research, so I'll be happy to sit you down and tell you the difference.

The amount of uneducation on any forum astounds me, and I will not allow these intangible results pollute the power of a video card. If you can show me how exactly, using llvm or the likes, the compiler optimization is done for "canned benchmarks", I will eat my own words.

Until then, I am not going to tolerate people bashing a video card that, when given the IDENTICAL CONTROL GROUP TO THE ATI OFFERING, beats it.

If you want to talk methodology, that's fine. I think the way H reviews is different, and very good. But do not let it fool you into thinking that it is some true way of comparing cards.

Comparison needs to have the same environment and testbed for both variables we're testing. Unfortunately, "real world gameplay" could mean that one cutscene experiences many more polygons than the other depending upon actions taken. That's like saying "The set of infinitely uncountable real numbers shows that this borel algebra has a larger measure than this sigma algebra". Sure it does, but it's also an infinitely uncountable probability space, which we define an outer measure on.

Anyways.
 
Xzyrus said:
Watching a video is no way accurate to judge sound since it'll depend a lot on how loud I crank my speakers. I'd rather just trust sound measurements from the same site (assuming they keep the same methodology).
Click to expand...

Unfortunately, test from different sites can't be compared. Different methods, ambient acoustics, and such.

After comparing [H]'s 5870 noise level to my own, and adjusting my speaker volume accordingly, both of the GTX4x0 cards are louder/more annoying when under load. Hopefully, you can take manual control of the fan and not burn up the card.
 
SicKlown42012 said:
Unfortunately, test from different sites can't be compared. Different methods, ambient acoustics, and such.

After comparing [H]'s 5870 noise level to my own, and adjusting my speaker volume accordingly, both of the GTX4x0 cards are louder/more annoying when under load. Hopefully, you can take manual control of the fan and not burn up the card.
Click to expand...

Nah I was comparing Guru3D's reviews of 480/470 to 5870 and 5850. But yea I went back and checked again. I was wrong and Guru3D seems to be the only place saying they aren't ridiculously loud.

I don't know what I was thinking earlier lol. It is comparable if I leave my speakers the same volume and play both vids simultaneously and one after the other.
 
I gave it a 7. In the grand scheme of things, only Nvidia could release a card this fast. Yes it is late, and is terribly hot/loud, but comeon, I haven't seen PowerVR, S3, Intel, or any other GPU manufacturer (ok PowerVR is IP only) have a release this good in their entire history (again, caveat, PowerVR had a pretty good Kyro launch, but still had buggy drivers).

Nvidia has great drivers, great performance, but really bad power use characteristics. You can argue that its relative performance isn't good enough (I agree with that), and that it uses far far too much power, but the average GPU player releases stuff far far far worse than this (so a 5 isn't appropriate).
 
McCartney said:
And real gameplay guarantees the amount of polygons rendered for each video card are identical, right? WRONG.

You cannot optimize hotpaths on a compiler for a video card due to "canned benchmarks", that is an absolutely FALSE argument.

You're WRONG, pal. A "canned" benchmark guarantees the same control group for the 2 variables being tested. Do you honestly think testing without a control group is a valid, scientific way to show that one card is better than another?

No, it isn't, and it never was. I highly suggest you prepare your next argument as I'm currently doing compiler optimization with Machine Learning research, so I'll be happy to sit you down and tell you the difference.

The amount of uneducation on any forum astounds me, and I will not allow these intangible results pollute the power of a video card. If you can show me how exactly, using llvm or the likes, the compiler optimization is done for "canned benchmarks", I will eat my own words.

Until then, I am not going to tolerate people bashing a video card that, when given the IDENTICAL CONTROL GROUP TO THE ATI OFFERING, beats it.

If you want to talk methodology, that's fine. I think the way H reviews is different, and very good. But do not let it fool you into thinking that it is some true way of comparing cards.

Comparison needs to have the same environment and testbed for both variables we're testing. Unfortunately, "real world gameplay" could mean that one cutscene experiences many more polygons than the other depending upon actions taken. That's like saying "The set of infinitely uncountable real numbers shows that this borel algebra has a larger measure than this sigma algebra". Sure it does, but it's also an infinitely uncountable probability space, which we define an outer measure on.

Anyways.
Click to expand...

Yes, you absolutely can optimize for canned benchmarks. Not only can you, but both nvidia and ATI got busted doing *EXACTLY* that. That is why [H] when to real gameplay in the first place. You may be doing compiler optimizations, but I'm guessing you've never done driver optimizations of DX/OGL. Heck, it sounds like you've never developed DX/OGL either, otherwise you'd know that only the shaders get compiled and that games use a lot more than just shaders. Canned benchmarks also don't stress the rest of the system the same, meaning the controlled variables lead to an invalid test.

And it doesn't matter if the runs aren't completely identical, that is why we have averages. Differences in play will average out between the two cards over time. Hell, the full FRAPS run through is shown so you can compare on more than just average FPS. Is it scientific? No. Is it more accurate? YES

And [H] doesn't benchmark cutscenes - anandtech does, though :rolleyes:
 
5: Mediocre.

With all the extra transistors (50% more) and memory bandwidth (50% more), the 480 should be getting a much larger (than 10-15%) performance jump over the 5870.

It is a bit faster, but it is much bigger/hotter and more expensive to produce.

Overall, not a very impressive cycle for NVidia.
 
McCartney said:
And real gameplay guarantees the amount of polygons rendered for each video card are identical, right? WRONG.

You cannot optimize hotpaths on a compiler for a video card due to "canned benchmarks", that is an absolutely FALSE argument.

You're WRONG, pal. A "canned" benchmark guarantees the same control group for the 2 variables being tested. Do you honestly think testing without a control group is a valid, scientific way to show that one card is better than another?

No, it isn't, and it never was. I highly suggest you prepare your next argument as I'm currently doing compiler optimization with Machine Learning research, so I'll be happy to sit you down and tell you the difference.

The amount of uneducation on any forum astounds me, and I will not allow these intangible results pollute the power of a video card. If you can show me how exactly, using llvm or the likes, the compiler optimization is done for "canned benchmarks", I will eat my own words.

Until then, I am not going to tolerate people bashing a video card that, when given the IDENTICAL CONTROL GROUP TO THE ATI OFFERING, beats it.

If you want to talk methodology, that's fine. I think the way H reviews is different, and very good. But do not let it fool you into thinking that it is some true way of comparing cards.

Comparison needs to have the same environment and testbed for both variables we're testing. Unfortunately, "real world gameplay" could mean that one cutscene experiences many more polygons than the other depending upon actions taken. That's like saying "The set of infinitely uncountable real numbers shows that this borel algebra has a larger measure than this sigma algebra". Sure it does, but it's also an infinitely uncountable probability space, which we define an outer measure on.

Anyways.
Click to expand...

Wow, you put the GTX 480 at 10, right up there with the 8800GTX?
 
nvidia needs to drop their prices. You would have to give me better performance at the same $$$ for being a power hog.

480 $400
470 $300
 
McCartney said:
You're WRONG, pal. A "canned" benchmark guarantees the same control group for the 2 variables being tested. Do you honestly think testing without a control group is a valid, scientific way to show that one card is better than another?
Click to expand...
But the tests are controlled. If they are not controlled, the shapes of the frame rate graphs would not match to such an astonishing degree.

The [H] review methodology offers less control than running a canned benchmark, but neither is perfect. Since you work on compilers, you should know that even though computers are designed to be deterministic, performance tests are never perfectly controlled. But perfect control is never necessary -- all you need is sufficient control to eliminate any systematic bias. If you see any possible systematic bias in the [H] review, I am sure Kyle and Brent would like to hear from you.

You should take it easy. You sounded so upset -- like someone had insulted your first crush.
 
7, although I considered 6 too. Nice card, but not awesome. Looks like I'll be waiting for a 5850 to come up at a decent price (and not at "oops, sorry wrong price....cancelled" Amazon)...unless nV can squeeze quite a bit more from GTX470 before that happens.
 
I voted 1.

I'm totally underwhelmed.

The icing on the cake is that there aren't any cards for sale on launch day.

friggin paper launch...
 
Snowdog said:
5: Mediocre.

With all the extra transistors (50% more) and memory bandwidth (50% more), the 480 should be getting a much larger (than 10-15%) performance jump over the 5870.

It is a bit faster, but it is much bigger/hotter and more expensive to produce.

Overall, not a very impressive cycle for NVidia.
Click to expand...

GTX480 does not have 50% more bandwidth, it is closer to 15%! While it does have a 50% wider bus, the memory is only operating at ~75% of the frequency. But even so, I had hoped for a little bit more. The 470 is what disappointed me, since that is what I was considering as a 5850 alternative.
 
McCartney said:
And real gameplay guarantees the amount of polygons rendered for each video card are identical, right? WRONG.
Click to expand...

No, but that is what I'm going to be using the card for. I couldn't give a fuck how much better it may or may not run a benchmark if that doesn't translate to actual gameplay performance.

Obligatory car analogy:
How do you determine which car is faster around a given track? You hand it over to your tame racing driver (ie The Stig) and they drive it around the course several times. Using a computer to control the car and set it along the exact same path at optimal speeds might give a more precise measurement of how fast it could possibly manage it, but it is not a real-world scenario.
 
The upside to these units is that they could replace a space hater. So there's another nice major feature that they have.
Click to expand...
If you live somewhere like Alaska or Greenland, sure. If not, then I guess the most obvious question is, "What about when summer comes round?"
 
I give it 10. Fastest single card on the market. What else is there to discuss?
 
A sure buy for me. I choose performance over the heat issues as I have ways around that. There is nothing I can do about the power consumption but I'm sure that's not going to be a problem since I plan to only purchase one GTX 480. 8 for me :).
 
I give it 5 for the following reasons:
No true Eyefinity, too loud, too much heat, heavy power requirement and not enough performance.
With some luck, the GTX 580 is only 6 months away...
 
cybereality said:
I give it 10. Fastest single card on the market. What else is there to discuss?
Click to expand...

/agree... 15-20%+ more performance for 20% more price ($400 vs 500) 5870 vs. 480, at most, what's there to discuss? Some driver flaws but it's out in a couple weeks which hopefully will see fixes regardless... in some cases it is beating out the 5870 by 30% or more, not a bad showing.

Price/performance never scales linearly at the top of the heap. All I see are a lot of fanATIcs changing their complaints at random to try to put down the Fermi, because of their bright red goggles.

RadXge said:
I give it 5 for the following reasons:
No true Eyefinity, too loud, too much heat, heavy power requirement and not enough performance.
With some luck, the GTX 580 is only 6 months away...
Click to expand...

Not enough performance? What is a 5870 then, WAY too little?
 
meanwhile every F@H gauge for these cards is radically different, I see everything from "it'll probably do 13k ppd" to "it could easily do 17k ppd." This all using a non-released beta client from Stanford because the current client won't even work with them.

I've got Final Fantasy XIII eating up all of my gaming time for a good few months, plus vacation and the Detroit Electronic Music Festival in the near timeline, so F@H is pretty much the only thing my videocard is going to be used for until... August? No solid backward compatibility and no solid F@H numbers to go off of... bummer. This is a card they were boasting the expanded compute abilities of, right? Bah.
 
GoldenTiger said:
/agree... 15-20%+ more performance for 20% more price ($400 vs 500) 5870 vs. 480, at most, what's there to discuss? Some driver flaws but it's out in a couple weeks which hopefully will see fixes regardless... in some cases it is beating out the 5870 by 30% or more, not a bad showing.

Price/performance never scales linearly at the top of the heap. All I see are a lot of fanATIcs changing their complaints at random to try to put down the Fermi, because of their bright red goggles.
Click to expand...

According to the thread YOU STARTED its 7-15% faster. Just decide to make up the 15-20%+ number or what?
 
MrXROK said:
nvidia needs to drop their prices. You would have to give me better performance at the same $$$ for being a power hog.

480 $400
470 $300
Click to expand...

That would be a best case scenario, but I just can't see it happening. Nvidia knows that they can sell those cards at that premium price due to either customer loyalty(fanboys) or ignorance(those who think AMD is a second rate chip maker, GPU wise).
 
kllrnohj said:
According to the thread YOU STARTED its 7-15% faster. Just decide to make up the 15-20%+ number or what?
Click to expand...

15%+, some games show a gain of 30-40%, some show 10-15%. The thread I started said around 15%. No one buys a $500 video card to run without AA ;). Had I omitted the non-AA results, you simply would have whined that I "cherry-picked" results ;).

SicKlown42012 said:
That would be a best case scenario, but I just can't see it happening. Nvidia knows that they can sell those cards at that premium price due to either customer loyalty(fanboys) or ignorance(those who think AMD is a second rate chip maker, GPU wise).
Click to expand...

Or outright performance, cost be damned ;)... or people who have had bad experiences with ATI chips. It isn't all red and green ya know ;).
 
SicKlown42012 said:
That would be a best case scenario, but I just can't see it happening. Nvidia knows that they can sell those cards at that premium price due to either customer loyalty(fanboys) or ignorance(those who think AMD is a second rate chip maker, GPU wise).
Click to expand...

I think they priced the cards correctly. Of course if sales figures are really off, they'll drop prices, but I think their marginal price/marginal benefit analysis was correct for consumers.
 
I was nice and gave it an 8.

I see Newegg has a SAPPHIRE 100282-3SR Radeon HD 5850 for $279 (in stock and for this weekend only), to me the 5850 is the Fermi killer.
 
Last edited:
GoldenTiger said:
15%+, some games show a gain of 30-40%, some show 10-15%. The thread I started said around 15%. No one buys a $500 video card to run without AA ;). Had I omitted the non-AA results, you simply would have whined that I "cherry-picked" results ;).
Click to expand...

And yet it only average 15% faster overall WITH AA. That does not translate to 15-20%+, it translates to 15% if you only look at with AA (you're so called "cherry-picked" situation).

And it was mainly Far Cry 2 that showed huge gains. MOST games showed 10-20% gains, and BC2 showed even less of a gain. No one buys a $500 video card to run old games, and if you narrow it to only DX11 games then Fermi's lead rapidly shrinks.
 
I am not disappointed with the performance, nor the price to performance ratio, I am seriously pissed about the heat and power consumption.

I am not going to buy this gpu, but with this heat and power consumption OEMs are not going to buy this gpu either. Furthermore I am going to have to deal with "the problems" when idiot people buy nvidia for their friends told them ATI sucks and suddenly they have problems with overheating, noise, and having to buy a very nice psu, for the crap psus can't even come close to powering this beast.
 
I hope Nvidia improves the framerate with better drivers. Ati had longer time to get their drivers in a row. Nvidia deserves same amount of time before judgement. I give it 6 (Heat is big concern). May give card a short lifespan. I thought shrinking the die size should bring down the heat. My fans already loud because I have (3) 3000 RPM fans to push air. Results in 60 degree Celcius at 60 percent fan speed. I have 3 gtx 280's. I find 2 gen Nvidia usually better and cooler than first gen. I am waiting for second gen. To run these cards will take good case cooling.
 
Give it a 6. Ten or nine will have to be a GPU that you can say innovative or "card skip a generation." 8-7 would have been a card that outperform the competition while still having the same power/heat/noise in the same range. 6-5 is where the card meets it goal. This card was design to be the fastest single GPU card, dam the power / heat /noise that it takes to get there. 4-3 underachieve. 2-1 would have been FAIL on everything.
 
kllrnohj said:
And yet it only average 15% faster overall WITH AA. That does not translate to 15-20%+, it translates to 15% if you only look at with AA (you're so called "cherry-picked" situation).
Click to expand...

I don't think you've realized it yet, he cherry picks his numbers to fuel his agenda.

He's posted on other forums about how its 25-30% faster then when people call him out, he twists his words or downright ignores it.
 
Status
Not open for further replies.
Back
Top