Comparison Revisited, Methods Refined: Nvidia GeForce GTX 480/470 vs. ATI Radeon HD 5870/5850

[ikenlob]

Guest said:
I think people are forgetting that the performance gap gets larger in the GTX480's favor when 2 are ran in SLI compared to 2 x 5870s ran in Crossfire. Many people with $400 - $500 dollars have enough money and usually do by a 2nd high-end card for a multi-GPU configuration. In this case which is more so in the likes of reality, 2 x 480s literally eat 2 x 5870s for breakfast.

Dwell on that thought for awhile.

I bought two 7800gtx in sli a few years ago, so new I had to wait a couple months on them. That cost me >$1000. More than a year later I let Steam survey my system and I was in the top 3% for GPU configuration. A year later, there were dozens of single cards that outpaced my two. I'd guess a lot more people SLI nowadays, but I am affiliated with a few clans, probably close to 150 gamers and don't know one single person with two high end cards in sli. For 99% of all gamers, why spend >$1000 when you can play everything on the market for $150 - $300?

Same argument for tesselation, 3D and such. Great when it makes financial sense or if you have nothing better to spend your money on. That is the point that Nvidia is missing.

 

[princeton]

Well this is the last time i trust a techspot review on gpu's. Just cause 2. Well you guys have the gtx 470 getting smoked. Well every other site I've seen with the gtx 470 vs hd 5850 just cause 2 benchmark has the opposite. Here are some examples.

http://www.guru3d.com/article/geforce-gtx-470-480-review/25
http://www.overclockersclub.com/reviews/inno3d_gtx470_hawk/11.htm

Look at this. On OC club the stock gtx 470 although it has different fans still runs at stock clocks. Is beating a factory overclocked 5850. How interesting.

Lets try another game.

Bad company 2. Now your review had it being beaten by the hd 5850 and by a noticeable bit as well.

http://www.bit-tech.net/hardware/graphics/2010/03/28/nvidia-geforce-gtx-470-1-1280mb-review/9
http://www.guru3d.com/article/geforce-gtx-470-480-review/16
http://www.hardwarecanucks.com/foru...s/30321-nvidia-geforce-gtx-470-review-14.html
Wow. Look at that. The GTX 470 isn't being beaten at all. Infact it's winning by its minimum framerate in every case. Even on hardware canucks the hd 5850 beats it by average but the gtx 470 has 3 more fps on the minimums.

I know everyone benchmarks different and I am NO nvidia fanboy. Infact the gtx 260 216 was the first card I've every bought and I bought it after reading reviews, specifically ones on this very site. And I was planning on getting an hd 5850 card but after seeing the benchmarks for the gtx 470 on every site EXCEPT this very one as better than the hd 5850. I wouldn't base an nvidia vs ati purchase off this site for this generation of cards.

 

[TomSEA]

princeton - you missed the whole point of this particular comparison. TechSpot ran these comparisons using Fraps over a period of time in real game play. Not just the usual bank of frame-rate testing software that everyone else in your links did.

This was a better representation of real-life performance.

 

[Kibaruk]

Would you please do one of this reviews, but taking into consideration heat and power, which is what really makes the difference between this.

 

[Guest]

How many electrical engineers do we have posting on here? How many of you non-engineers can conclusively say that the ninety degrees temps are "bad" while the ATI 80 degree temps are "good". Nobody complains that a Honda engine combustion temperature reaches 2300K, which is more than a 'typical' engine, because the actual degree doesn't doesn't matter as long as it doesn't exceed the temp it was designed for.

Nvidia's cards were designed to operate at a particular temp and who are you laymen to claim that the temp they've designed to is 'bad'. The cards have only been available for 3 months and you're claiming they won't last the year because they happen to run 10 degrees hotter than ATI's cards?

At best, that's piss pour logic. At worst it's a bunch of people who know nothing about how a board is designed, built and tested talking out their ***.

 

[Guest]

..Because high temps cause all sorts of hardware failures? Comparing it do an engine is pretty stupid since those things are built to withstand that kind of thing, whereas computers are more delicate when it comes to heat. It's why when I overclock my 4870 too high my computer crashes - I have a perfectly nice 650w power supply so that ain't the problem. However, I don't have any case fans and need to get a new case full stop, which is the reason my temps are higher than others, not the card itself. And out of curiosity, where do you get this 80C ATI temps from? From what I've seen it's much lower than that in the benchmarks.

 

[Guest]

I see that these were done with retail cards. In the first review, they appear to be generic. Could that make the difference we see here? There were some rumors floating around about a BIOS change between the launch and actual appearance on the retail market, purportedly to address the fan levels, but who knows what else may have been changed.

 

[LNCPapa]

Not only that princeton, but look at the driver versions and processors (one of them) used - they could explain some of the differences you see as well.

 

[princeton]

LNCPapa said:
Not only that princeton, but look at the driver versions and processors (one of them) used - they could explain some of the differences you see as well.

At 1900x1200 in most games CPU won't have much of an impact. Now when your playing at 1900x1200 you are probably running an sli/CF setup or a dual gpu card. Dual gpus means the cpu will need to be better to feed info to both cards. But since these are single card benchmarks I doubt it would have much impac. Regarding driver versions. I have a gtx 260 and I can say I gained significant increase in fps in crysis warhead. Around 4-5fps actually. My friend micheal runs a gtx 480 and he has also reported increased FPS in warhead and metro 2033 to name a couple.

 

[princeton]

TomSEA said:
princeton - you missed the whole point of this particular comparison. TechSpot ran these comparisons using Fraps over a period of time in real game play. Not just the usual bank of frame-rate testing software that everyone else in your links did.

This was a better representation of real-life performance.

FRAPS we recorded a 1-minute passage: Crysis warhead

Yah 1min is suuuure an accurate representation of actual gameplay. It doesn't even say what level it was. For all we know it could be any level in the entire game and as we all know crysis and warheads levels tend to produce different levels of fps. You want a representation of real life performance. Take a 10run through adapt or perish with every card. Then come say "representation of real life performance."

 

[TomSEA]

Princeton - what part of those other reviewers using the same generic testing software and didn't do it using a real-time performance test are you not getting? Even an assessment for one minute of actual game play is better than some 3DMark or StoneGiant b.s. "who the hell knows how they got it" benchmarking figures

 

[dividebyzero]

NIce review as far as it went.
Interesting to see how the usual benchmark suite was culled down to seven games...
The games left out:

Batman:Arkham Asylum
CoD: MW2
Far Cry 2
HAWX
Company of Heroes:Opposing Front (this one getting a little long in the tooth anyhow)
World in Conflict: Soviet Assault

mmmmm....well I suppose certain tech (satire) sites will have one less target in their sights.

BTW: Nice to see the HD 5870's benchmark scores finally move on Resident Evil 5. The original 5870 review (23 Sept 2009) listed the fps as 108, 96, 70 ( for 1680, 1920, 2560 resoluitions) using the release Catalyst 9.9 driver, while under the Catalyst 10.4 driver in the iChill GTX 480 review (26 Mar 2010) the benchmark fps's were an identical 108. 96, 70. The Far Cry 2 benchmarks I noted were also identical.

 

[princeton]

TomSEA said:
Princeton - what part of those other reviewers using the same generic testing software and didn't do it using a real-time performance test are you not getting? Even an assessment for one minute of actual game play is better than some 3DMark or StoneGiant b.s. "who the hell knows how they got it" benchmarking figures

So your implying that every review for gpu's ever made that has used a benchmark tool is useless? Well then I guess we can put these on the list of useless review sites.

tomshardware
bit-tech
hadwarecanucks
bjorn3d
guru3d
overclockers club
overclockers.com
overclock.net

ect. I understand that the benchmarks here MAY be better representations of actual gameplay. The point still stands that the nvidia cards are more powerful. If they weren't then they wouldn't get higher scores in benchmarks. I would also like to point out that people whining about the operating temperatures obviously don't understand that there is no universal "safe" component temperature. I know people who have ran core i7 cpus at 100 degrees and they didn't crash or die. A motherboard northbridge tends to get very hot. These nvidia gpus are designed to run at higher operating temperatures. I would also like to note that these people seem to think that after all these years of outstanding service and products by nvidia that they would just decide to put out a product they know wont last very long.

 

[grvalderrama]

It was quite a mistake not to consider the heat thing by Nvidia's staff. Heat is the nightmare of all microdevices developers... I must say AMD did it really good.

Cheers!

 

[Guest]

Nvidia has had a long long time to create a product to beat the 5870 but what they brought out was about the same at a higher cost and poorer efficiency. Fail.

 

[Guest]

I bought my gtx 470 msi for $270 dollars new. Can't complain!

 

[princeton]

Guest said:
I bought my gtx 470 msi for $270 dollars new. Can't complain!

/Puts gun to your head. WHERE WHERE WHERE WHERE WHERE!!!!!!

 

[Guest]

@ princeton - ohh no TechSpot is not trusted by an *****ic Nvidia fan boy that throws links around to try and make his point. I have seen plenty of tests where ATI is faster in JC2. It comes down to the settings used and TS used maximum quality settings. Google GTX 470 Just Cause 2 performance and on the first page there are a few reviews that use the same settings and find the same results.

Awesome review guys. Thanks for posting the truth!

 

[vipervoid1]

Nice review...
and thx for the review too...

Well, I'm planning to get myself a GTX480 or HD5870...
Now I decided ,that I will get myself a HD 5870...

 

[Julio Franco]

Thank you everyone for your feedback. It's not uncommon for this kind of review to generate some controversy, as mentioned in the article we were surprised by the results ourselves.

Rest assured we are not trying to convince or please anyone other than our loyal readers that want facts. We have described our methods that in our opinion represent a closer picture of real gameplay performance for the games we tested.

Finally, for this article we focused on performance and not temperatures or power consumption, which were measured on our original review of the GeForce GTX 480. You can use those values as reference, we don't believe newer drivers have had any effect on those.

 

[Guest]

Nice review, I like those line plots better because they highlight places where the gaming is slow. Base on my personal experience, it is not how many 100+fps the graphics card can do to have good enjoyable gaming experience but rather, it is how well the graphics card can do to cope with the minimum fps.

Seems to me that the 5870 virtually the same as the gtx480 with the exception of Metro2033. The GTX470 is painfully slow in almost every benchmark compared to the 5850. That card, to me, is a worst out of the four.

I know Nvidia has got CUDA and Physx. However, the man responsible for CUDA has moved to AMD promoting the development of AMD FUSION APU. I think it is clear that the tide is shifting. The way I see it, the turtle (AMD) has overtaken the hare (Nvidia, who has been napping for too long).

 

[princeton]

Guest said:
@ princeton - ohh no TechSpot is not trusted by an *****ic Nvidia fan boy that throws links around to try and make his point. I have seen plenty of tests where ATI is faster in JC2. It comes down to the settings used and TS used maximum quality settings. Google GTX 470 Just Cause 2 performance and on the first page there are a few reviews that use the same settings and find the same results.

Awesome review guys. Thanks for posting the truth!

As I already outright proved I am no nvidia fanboy. Infact I have only owned one gaming gpu in my life so there is no possible way I could be biased. But nice try trying to make yourself look intelligent.

 

[OneArmedScissor]

So your implying that every review for gpu's ever made that has used a benchmark tool is useless?

I can't say I've ever played a benchmark.

I can say I have seen a lot of extremely lazy reviews on so called "PC enthusiast" websites, and you named some of the most serious offenders.

And for the record, you can't replicate a 10 minute run of a game. I mean seriously, you believe benchmarks are worthwhile, but then you ask for that? Where is the logic?

 

[red1776]

As I already outright prooved I am no nvidia fanboy. Infact I have only owned one gaming gpu in my life so there is no possible way I could be biased.

This makes the top ten of most ridiculous statements ever made on TS.

And for the record, you can't replicate a 10 minute run of a game. I mean seriously, you believe benchmarks are worthwhile, but then you ask for that? Where is the logic?

+1

Why don't you just purchase your GTX 470 and be happy? instead of trying to convince everyone that it's superior? .....BTW...how is the benefit package over at NVidia?

 

[dividebyzero]

Seems to me that the 5870 virtually the same as the gtx480 with the exception of Metro2033. The GTX470 is painfully slow in almost every benchmark compared to the 5850. That card, to me, is a worst out of the four.)

In this review yes.
But like any other review it’s in the numbers…
If , for arguments sake, you included HAWX (which up until this review was part of Techspots benchmarking suite), you might see this
And if you included CoD:MW2, a fairly popular game I believe (which up until this review was part of Techspots benchmarking suite), you might see this
Likewise for a full electric chair review; “Sims 3”, a title that heavily favours AMD cards could have been included at the expense of say, “Metro 2033”

A fairly simple metric of ascertaining how "good" or "bad" these cards are would be to look at the newegg reviews for both the GTX480/470, the HD 5850 and HD 5870 as well as the incidences of “buyers remorse” in the online auction arena

I know Nvidia has got CUDA and Physx. However, the man responsible for CUDA has moved to AMD promoting the development of AMD FUSION APU. I think it is clear that the tide is shifting. The way I see it, the turtle (AMD) has overtaken the hare (Nvidia, who has been napping for too long).

"The man responsible for CUDA".....oh dear! Don't your arms get tired waving the red flag?
The man in question is Manju Hegde. He was a vice president of PhysX and CUDA MARKETING. While he was the CEO of Ageia (admin, not tech) when nVidia bought the company, he certainly didn't scale the same heights at nVidia. The main man for CUDA is Ian Buck (PhD Stanford I believe).
Feel free to feel the excitement of AMD acquiring one of nVidia's PR guys.

Somehow I doubt that an APU sporting HD5450 level graphics is going to seriously hinder enthusiast graphics cards development for the foreseeable future, and considering ATI graphics are the only thing standing between AMD and a sea of red ink maybe that's just as well.
You might also consider that part of AMD's appeal has been as the plucky underdog...something thats a little hard to maintain when you're heading for parity or better in market share. You might do well to take note of the noticable groundswell of a reversal of opinions in light of the many ill informed bandwagon jumpers that permeate the public domain :wave:

Oh, and before you get to carried away in the euphoria of nvidia's impending demise, here's a tidbit regarding Nebulae- the worlds second biggest computer. You'll note that in addition to the 9,280 Intel Xeon X5650 CPU's it also runs 4,640 nVidia C2050 Tesla cards (workstation GTX 470). If certain tech "journalists" are right with their 5,000-8,000 total GTX4xx cards being produced, then nVidia might need to run off another batch fairly soon.:haha:

This makes the top ten of most ridiculous statements ever made on TS.

I'll see your ridiculous nVidia fanboy...and raise you a nonsensical AMD fanboy.....watcha got?, I got a full house, kneejerk reactionaries over tards!:haha: