ATI 4870 vs. nV GTX 260

[thilanliyan]

http://www.xbitlabs.com/articl...y/force3d-hd-4870.html

 

[Mwing]

If not EVGA step up program, I would have chosen 4870 instead of 260GTX

 

[Wreckage]

So, the GeForce GTX 260 may be considered a formal leader in this competition

No surprise really. It's faster in most games even at stock speeds. Although several companies offer factory overclocked models with a full warranty.

 

[WT]

Looks like it runs out of RAM above 1920x1200. I personally am not interested in the 4870 512s. The 1gb cards with better cooling and higher clocks will do just fine.

 

[Ares202]

Who actually runs above 1920x1200 anyway

the 4870 only runs out of memory on 1-2 games crysis is one of them that is no suprise and the GTX280 cant handle that res anyway <10fps is not good at all

people who have 30" monitors will buy something like the 4870x2 or gtx280 sli so i dont see the point of saying that 512mb is not enough memory when for 95% of people it is

 

[Creig]

Actually, it ran out of physical memory in at least three of the tests. I think most people already knew that for ultra-high resolutions, a card with only 512mb would not be sufficient which makes me wonder why they didn't hold off publishing this review until they could obtain one for benching. The 1GB 4870's are due out soon, so it will be interesting to see if xbit updates this review once they get ahold of a sample.

edit - They used Catalyst 8.6 driver? 8.7 has been out since July 21st! That makes no sense at all.

 

[Qbah]

So let's do a summary, 1920x1200:

BF2142:
- HD4870: min 67 avg 87 <- extremely playable
- GTX260: min 77 avg 92 <- extremely playable

BioShock:
- HD4870: min 44 avg 91 <- extremely playable
- GTX260: min 42 avg 80 <- extremely playable

Call of Juarez:
- HD4870: min 11 avg 26 <- not playable
- GTX260: min 14 avg 19 <- not playable

Call of Duty 4:
- HD4870: min 37 avg 69 <- extremely playable
- GTX260: min 37 avg 52 <- extremely playable

Crysis:
- HD4870: min 10 avg 21 <- not playable
- GTX260: min 10 avg 18 <- not playable

Enemy Territory: Quake Wars:
- HD4870: min ?? avg 67 <- extremely playable
- GTX260: min ?? avg 76 <- extremely playable

Half-Life 2: Ep2:
- HD4870: min ?? avg 70 <- extremely playable
- GTX260: min ?? avg 69 <- extremely playable

S.T.A.L.K.E.R:
- HD4870: min 16 avg 40 <- playable
- GTX260: min 35 avg 47 <- extremely playable

Lost Planet: Extreme Condition:
- HD4870: min 21 avg 25 <- borderline playable
- GTX260: min 11 avg 15 <- not playable

Tomb Raider: Legend:
- HD4870: min 36 avg 67 <- extremely playable
- GTX260: min 42 avg 87 <- extremely playable

Hellgate: London
- HD4870: min 22 avg 46 <- playable
- GTX260: min 36 avg 45 <- extremely playable

Oblivion:
- HD4870: min 36 avg 68 <- extremely playable
- GTX260: min 30 avg 67 <- extremely playable

CoH: Opposing Fronts:
- HD4870: min 22 avg 43 <- playable
- GTX260: min 8 avg 54 <- playable

C&C: Kane's Wrath:
- HD4870: min 26 avg 30 <- extremely playable
- GTX260: min 24 avg 30 <- extremely playable

World in Conflict:
- HD4870: min 12 avg 28 <- playable
- GTX260: min 10 avg 25 <- playable

So... That's dead even in my eyes. The only deal breaker would be Lost Planet, which is a shitty coded DX10 port. Just run it in DX9, have 99% of the effects in DX10 and play at butter smooth framerates! That's what I did anyway. Game's cool though CoJ and Crysis don't run well on either, so saying a card wins in one of those games is just silly. WiC is a RTS, lower framerates don't break the gameplay here.

Though they don't mention if they ran each test a few times, as a one-time frame dip 'cause the drive was reading something else or w/e other reason could explain the lower min frames with high avg sometimes.

In the end, pick whatever you feel better with. I like monthly driver updates, rock solid operation and was always partial to ATi due to having *ALWAYS* problems with my nVidia cards (don't ask).

EDIT: Forgot to mention one thing. Why do all of the reviews play Crysis DX10 if they don't use Very High anyway? DX10 Medium for example. Hell, DX9 High-tweaked runs *extremely* well on my setup at 1920x1200. At DX10 I get like 10-14FPS. Going DX10 -> DX9 Medium or High doesn't change the graphics at all, however gives you like double framerates!

 

[Extelleron]

Originally posted by: Wreckage

So, the GeForce GTX 260 may be considered a formal leader in this competition

No surprise really. It's faster in most games even at stock speeds. Although several companies offer factory overclocked models with a full warranty.

The GTX 260 is not faster in most games. This review shows that very clearly as have others. In a number of games the 4870 is a competitor for the GTX 280. When comparing the GTX 260 to the HD 4870, at the same price, the choice is clear. The only advantage the GTX 260 seems to have is overclocking. Given that the HD 4870 can overclock too, overclocking is not going to give the GTX 260 much of an advantage in performance.

 

[sourthings]

Originally posted by: Extelleron

Originally posted by: Wreckage

So, the GeForce GTX 260 may be considered a formal leader in this competition

No surprise really. It's faster in most games even at stock speeds. Although several companies offer factory overclocked models with a full warranty.

The GTX 260 is not faster in most games. This review shows that very clearly as have others. In a number of games the 4870 is a competitor for the GTX 280. When comparing the GTX 260 to the HD 4870, at the same price, the choice is clear. The only advantage the GTX 260 seems to have is overclocking. Given that the HD 4870 can overclock too, overclocking is not going to give the GTX 260 much of an advantage in performance.

Yeah, talk about complete reversal of the truth. The 4870 is faster in most games than the GTX 260. Except when the GTX 260 is OCed, and everyone fails to mention that if you overclock the 4870, again it pulls ahead.

 

[apoppin]

Originally posted by: Creig


edit - They used Catalyst 8.6 driver? 8.7 has been out since July 21st! That makes no sense at all.

http://www.xbitlabs.com/articl...d-hd-4870_6.html#sect0

ATI Catalyst 8.6 for ATI Radeon HD

*invalid*

welcome to June

:thumbsdown:


Here is their conclusion - which is good for 2 months ago:

As you can see, the Radeon HD 4870 and GeForce GTX 260 both deliver comfortable performance at 2560x1600 with 4x FSAA in seven out of the 15 tests. In the remaining eight games the $299 cards are either barely short of the comfortable level or downright slow. Note also that it is only in rare cases that one of these cards enjoys an overwhelming advantage over the other.

So, if you want to have smooth gameplay in every game in this display mode, you should consider purchasing a multi-GPU solution. Practice suggests that the GeForce GTX 280 is not far faster than the junior model in many applications and thus cannot satisfy a fastidious gamer.

Choosing between the GeForce GTX 260 and Radeon HD 4870 for playing at 2560x1600 with 4x FSAA, the former is better in five applications while the latter, in three. So, the GeForce GTX 260 may be considered a formal leader in this competition although your choice should be based on the particular games you are intending to play.

If you play at 25x16 - still true for is in Late August - you need SLi or Xfire. No single card will suffice with max details and 4xAA[period]

and i am ReTesting ALL my benches with Cat 8.8
- why be lazy?

 

[geoffry]

Xbit has had bizarre vid card reviews lately.

Their 4850 CF was WAAAAAY off.

Even their non CF ATI numbers were quite a bit lower than everywhere else in some cases.

Ergo for this generation I don't pay attention to what they've posted.

Its a shame, I love their site, its very in depth, especially for monitors.

 

[Keysplayr]

Originally posted by: Creig
Actually, it ran out of physical memory in at least three of the tests. I think most people already knew that for ultra-high resolutions, a card with only 512mb would not be sufficient which makes me wonder why they didn't hold off publishing this review until they could obtain one for benching. The 1GB 4870's are due out soon, so it will be interesting to see if xbit updates this review once they get ahold of a sample.

edit - They used Catalyst 8.6 driver? 8.7 has been out since July 21st! That makes no sense at all.

Yeah, they also used Forceware 177.41 released in June as well. 177.83 is out now.
Makes no sense.

 

[geoffry]

Originally posted by: keysplayr2003

Originally posted by: Creig
Actually, it ran out of physical memory in at least three of the tests. I think most people already knew that for ultra-high resolutions, a card with only 512mb would not be sufficient which makes me wonder why they didn't hold off publishing this review until they could obtain one for benching. The 1GB 4870's are due out soon, so it will be interesting to see if xbit updates this review once they get ahold of a sample.

edit - They used Catalyst 8.6 driver? 8.7 has been out since July 21st! That makes no sense at all.

Yeah, they also used Forceware 177.41 released in June as well. 177.83 is out now.
Makes no sense.

Wow, I didn't realize they had ancient NVDA drivers as well.

Wonder why they dropped the ball this gen.

 

[apoppin]

Originally posted by: keysplayr2003

Originally posted by: Creig
Actually, it ran out of physical memory in at least three of the tests. I think most people already knew that for ultra-high resolutions, a card with only 512mb would not be sufficient which makes me wonder why they didn't hold off publishing this review until they could obtain one for benching. The 1GB 4870's are due out soon, so it will be interesting to see if xbit updates this review once they get ahold of a sample.

edit - They used Catalyst 8.6 driver? 8.7 has been out since July 21st! That makes no sense at all.

Yeah, they also used Forceware 177.41 released in June as well. 177.83 is out now.
Makes no sense.

i can understand 8.7 .. with 8.8 being released today

but to test with drivers that are over 1 month old is irresponsible imo

it does not take THAT LONG to retest with the latest. Setting up the article and the testbed is what takes so damn long; a retest is relatively easy [although it may change their conclusions]

did they even use the Hot-Fix 8.6's [doubtful - although they say "for ATi Radeon HD" ]


What awful *irrelevant* testing and editorial IRresponsibility to have Aug 19th's date on it. Worse than useless for any current conclusions - we now have the same thing as a launch article -
they are just 2 months late.
!

 

[Wreckage]

Originally posted by: Extelleron

Originally posted by: Wreckage

So, the GeForce GTX 260 may be considered a formal leader in this competition

No surprise really. It's faster in most games even at stock speeds. Although several companies offer factory overclocked models with a full warranty.

The GTX 260 is not faster in most games. This review shows that very clearly as have others. In a number of games the 4870 is a competitor for the GTX 280. When comparing the GTX 260 to the HD 4870, at the same price, the choice is clear. The only advantage the GTX 260 seems to have is overclocking. Given that the HD 4870 can overclock too, overclocking is not going to give the GTX 260 much of an advantage in performance.

You clearly did not read the review or their conclusion. As they clearly give the edge to the GTX260.

The 4870 is also terrible to overclock as it already runs very hot and you will void the warranty.

Not to mention it's memory limitation in comparison.

So that's 3 strikes against it.

 

[chizow]

Originally posted by: sourthings

Originally posted by: Extelleron

Originally posted by: Wreckage

So, the GeForce GTX 260 may be considered a formal leader in this competition

No surprise really. It's faster in most games even at stock speeds. Although several companies offer factory overclocked models with a full warranty.

The GTX 260 is not faster in most games. This review shows that very clearly as have others. In a number of games the 4870 is a competitor for the GTX 280. When comparing the GTX 260 to the HD 4870, at the same price, the choice is clear. The only advantage the GTX 260 seems to have is overclocking. Given that the HD 4870 can overclock too, overclocking is not going to give the GTX 260 much of an advantage in performance.

Yeah, talk about complete reversal of the truth. The 4870 is faster in most games than the GTX 260. Except when the GTX 260 is OCed, and everyone fails to mention that if you overclock the 4870, again it pulls ahead.

Except the 4870 isn't going to overclock nearly as much as the GTX 260. Going by the highest OC'd variants as an indication of attainable OC speeds, you have the GTX 260 FTW @ 666MHz, which is a 15.6% increase over stock. Diamond's XOC or PowerColor's PCS 4870 come in at 800MHz, a 6.6% increase over stock. And that's before considering speeds above the BIOS limited 780/790MHz clocks on the 4870 are impossible without a BIOS flash and changing the default fan profile speeds.

But ya I'd agree that Xbit's reviews have been questionable lately. Even on release drivers, I haven't seen so many 4850CF results that don't scale out of a single benchmark suite.

 

[CP5670]

This review was probably written quite some time ago but only published recently. That is the only way to explain their use of old drivers. Maybe the delay with their reviews is due to the Russian-English translation?

 

[geoffry]

Originally posted by: CP5670
This review was probably written quite some time ago but only published recently. That is the only way to explain their use of old drivers. Maybe the delay with their reviews is due to the Russian-English translation?

They don't get early samples like Anand and other American, Asian and European review sites. I believe they buy their cards at launch and start work then, hence the delayed reviews.

 

[chizow]

Originally posted by: CP5670
This review was probably written quite some time ago but only published recently. That is the only way to explain their use of old drivers. Maybe the delay with their reviews is due to the Russian-English translation?

Its possible for sure, but I think it also has to do with them dropping to 2nd/3rd tier reviewer status and waiting for production parts from board partners rather than receiving engineering samples and reference boards from ATI/NV directly. It was pretty clear who ATI considered 1st tier based on who got the 10-12 4870X2 ES for the preview. Its also very clear who NV considers top tier with the reference cards going to reviewers who often supplement it with a 2nd or 3rd from board partners for early SLI comparisons.

This situation also emphasizes the pressures and competition in the reviewer industry and how quickly things can change when sites like Xbit and Digit-Life are months late on reviews and relying on board partner generosity for their reviews. Its certainly why I didn't give AT a hard time breaking their proclaimed soft/paper launch embargo.

 

[Elcs]

Originally posted by: chizow
Except the 4870 isn't going to overclock nearly as much as the GTX 260. Going by the highest OC'd variants as an indication of attainable OC speeds, you have the GTX 260 FTW @ 666MHz, which is a 15.6% increase over stock. Diamond's XOC or PowerColor's PCS 4870 come in at 800MHz, a 6.6% increase over stock. And that's before considering speeds above the BIOS limited 780/790MHz clocks on the 4870 are impossible without a BIOS flash and changing the default fan profile speeds.

But ya I'd agree that Xbit's reviews have been questionable lately. Even on release drivers, I haven't seen so many 4850CF results that don't scale out of a single benchmark suite.

Just to note: HD4870's can be overclocked past their 790/1100 CCC locks by AMD's GPU Clock Toolso therefore BIOS flashes are not required.

Fan profile fixes are not a requirement either I do not think but a definite temperature preference. The fan is set silent by default and starts ramping up at around 80-82C in my short experience with the stock fan before my Thermalright HR-03 GT.

Im hoping to find an advanced fan speed profile maker for my HD4870 as I have spliced a 120mm Scythe Kama PWM fan to run off the header on the card. Very brief testing revealed that the fan powers up but I am not running it yet until I receive the replacement for my Slipstream 1200 rpm fan which was a ticker.

 

[chizow]

Originally posted by: Elcs

Originally posted by: chizow
Except the 4870 isn't going to overclock nearly as much as the GTX 260. Going by the highest OC'd variants as an indication of attainable OC speeds, you have the GTX 260 FTW @ 666MHz, which is a 15.6% increase over stock. Diamond's XOC or PowerColor's PCS 4870 come in at 800MHz, a 6.6% increase over stock. And that's before considering speeds above the BIOS limited 780/790MHz clocks on the 4870 are impossible without a BIOS flash and changing the default fan profile speeds.

But ya I'd agree that Xbit's reviews have been questionable lately. Even on release drivers, I haven't seen so many 4850CF results that don't scale out of a single benchmark suite.

Just to note: HD4870's can be overclocked past their 790/1100 CCC locks by AMD's GPU Clock Toolso therefore BIOS flashes are not required.

Fan profile fixes are not a requirement either I do not think but a definite temperature preference. The fan is set silent by default and starts ramping up at around 80-82C in my short experience with the stock fan before my Thermalright HR-03 GT.

Im hoping to find an advanced fan speed profile maker for my HD4870 as I have spliced a 120mm Scythe Kama PWM fan to run off the header on the card. Very brief testing revealed that the fan powers up but I am not running it yet until I receive the replacement for my Slipstream 1200 rpm fan which was a ticker.

Are you sure about AMD GPU Tool breaking the BIOS limit? Judging from this thread and Fattysharp's reply and others here and places like XtremeSystems, Rage3D etc. it seems thats not the case.

I've also read complaints about load temps on the 4850 and 4870 specificially because the fan does not ramp up as they should. Seems like its hit or miss I suppose based on BIOS, but if you were overclocking to 800MHz and beyond on a 4870 its not something I'd leave to chance.

 

[Munky]

Who cares if the gtx260 is faster at 2560x1600, when it's still unplayable in majority of games at those settings, and it loses at playable 1920x1200 rez?

 

[sourthings]

If you want to turn up your 4870 fan speed: http://www.xtremesystems.org/F...howthread.php?t=192932

Bought a 4870 for my girlfriend's pc recently after being so satisfied with X2. Set the fan to 38% with this fix, it's quiet and idles at 45, loads at 72 or so, this is at 790/1100.

A link to someone who compared frame rate returns with overclocking/underclocking on a 4870 in Crysis :http://www.hardforum.com/showt....php?t=1319500&page=10

It's also worth mentioning that most people who buy a card in the performance range of a 4870/GTX 260 up to GTX280 and up to 4870X2, are definitely planning on using AA. And the GTX 260's performance with AA enabled tanks compared to a 4870 using 8X AA which shows a very small performance hit enabled on the 4870, whereas the 260 takes a big hit in performance.

 

[Creig]

Originally posted by: chizow
Except the 4870 isn't going to overclock nearly as much as the GTX 260. Going by the highest OC'd variants as an indication of attainable OC speeds, you have the GTX 260 FTW @ 666MHz, which is a 15.6% increase over stock. Diamond's XOC or PowerColor's PCS 4870 come in at 800MHz, a 6.6% increase over stock. And that's before considering speeds above the BIOS limited 780/790MHz clocks on the 4870 are impossible without a BIOS flash and changing the default fan profile speeds.

A 4870 doesn't have to overclock as much as a GTX260. In the majority of benches, a 4870 at stock is already faster than a GTX260 and even trades victories with the GTX280. When the 1GB 4870's are released, it will make the 4870 even faster overall since it won't drop performance at 2560x1600 in certain games as the xbit review clearly shows with the 512MB 4870.

 

[Creig]

Originally posted by: keysplayr2003

Originally posted by: Creig
edit - They used Catalyst 8.6 driver? 8.7 has been out since July 21st! That makes no sense at all.

Yeah, they also used Forceware 177.41 released in June as well. 177.83 is out now.
Makes no sense.

What's the point of publishing benchmarks that are so old that newer drivers are already available? So we can see how well the cards used to perform?

A lot can change from driver to driver. It's nice of them to go through the effort of benching the cards and writing the review, but as apoppin said, the hard part is getting the testing platforms assembled. They should have checked for newer available drivers and rebenched once they discovered they were using older drivers for all the cards.