How do we account for the review sites, listing the 4870 faster than the 280?

T

TalonMan

[H]ard|Gawd
Joined
Feb 16, 2007
Messages
1,351
How do we account for some review sites, listing the 4870 faster than the 280, or real close in performance?

In this review: http://www.techreport.com/articles.x/14990/9

It is odd the review has for COD4 at 2560x1600 showing:

GX2 = 57.6 FPS, and the 280 = 51.2 FPS,

And in this video COD4 at 2560x1600 : http://ca.youtube.com/watch?v=3GHyE9xJZJE

we see that in actual game play shows:

GX2 = 40 to 80 FPS, and the 280 = 50 to 169 FPS, and is running faster the entire way...

So if we know a GX2 is faster than a 4870 in the reviews, and we can see a 280 is faster than a GX2 in the video, where does that leave the 4870 vs the 280?

The non-video review gives an inaccurate conclusion on the better card.

They must of had a funky 280?

They do have the 4870 at 44.4 FPS, and in Crossfire at 82 FPS. That might be correct, but the 280 is faster than that in the video for sure.

As far as the [H] goes...

The [H] has a 280's max settings in Crysis at 1920x1200 No AA 16x AF:
Texture Quality:High
Objects Quality: Medium
Shadows Quality: High
Physics Quality High
Shaders Quality: High
Volumetric Effects Quality: High
Game Effects Quality: High
Postprocessing Quality: High
Particles Quality: High
Water Quality: High

Min=17 Max=46 AVG=31.1

For the 4870 they have max settings in Crysis at 1920x1200 No AA 16x AF:
Texture Quality:High
Objects Quality: Medium
Shadows Quality: Medium
Physics Quality High
Shaders Quality: High
Volumetric Effects Quality: Medium
Game Effects Quality: Medium
Postprocessing Quality: Medium
Particles Quality: Medium
Water Quality: High

Min=11 Max=41 AVG=27.2


That does not look like the 280's performance to me...



In Assassins Creed, the [H] has for the 280 2560x1600 4x AA, AF Enabled:
Shadows: 3/3
Multisampling 3/3
Post FX: On
Graphic Quality: 4/4
Level of Detail: 4/4

Min=25 Max=53 Avg=37

For the 4870 in Assassins Creed, the [H] has for the 280 2560x1600 (Only 2xAA), AF Enabled:
Shadows: 3/3
Multisampling 2/3
Post FX: On
Graphic Quality: 4/4
Level of Detail: 4/4

Min=23 Max=52 Avg=38.2

This one is closer, but I am not sure how much extra work it puts on the GPU going from 2x AA, to 4x AA, and using the lower Multisampling setting.

Closer to 280 performance, but still not there...


Age of Conan, the 280 rules going up to 2560 x 1600 No AA, 16x AF. the 4870 can't hang there.

At 1920 x 1200 the 280 can use higher in game settings, and run 8x CSAA and 16x AF. The 4870 could not.

Dosen't look like the performance you get from a 280 again...



In COD4 as I posted above, a 280 is faster than a GX2 in the video by far:
http://ca.youtube.com/watch?v=3GHyE9xJZJE

We also know a GX2 beats a 4870 pritty bad in COD4 from the reviews.

I would say the 4870 is much more behind the 280's performance in this game also...


I would love for the [H] to run some 4870's in Crossfire, against a single 280.

That would be a fun report to read.

The 4870 -vs- the 280 sure has had so many conflicting reports...

The video looks real to me...

I think the 280 will beat the GX2 99% of the time.

I also think a GX2 beats a 4870 99% of the time.

If we were to just look at their COD4 performance graph:
http://anandtech.com/video/showdoc.aspx?i=3341&p=14

And were to pick the winner based only on their data between the GX2 ant the 280.
280=90.0, GX2=107.9
We would pick the GX2...

Now how do we reconcile that, with the video of the 280 putting a convincing spankie-doodle on the GX2 in that very game? I can't...

It makes me think if their COD4 numbers are out of wack, what am I to think about all the others. A 280 beats a GX2 in COD4 period.

When we see something like the [H] review using max playable settings, the 280 shines, just like in the video against the GX2 running real games.

I think the 4870 is closer to the 260's performance myself, not the 280.

I also noticed that the [H] has the same max playable settings for a 260, and a 4870 in Crysis:

For the 4870 they have max settings in Crysis at 1920x1200 No AA 16x AF:
Texture Quality:High
Objects Quality: Medium
Shadows Quality: Medium
Physics Quality High
Shaders Quality: High
Volumetric Effects Quality: Medium
Game Effects Quality: Medium
Postprocessing Quality: Medium
Particles Quality: Medium
Water Quality: High

Min=11 Max=41 AVG=27.2


For the 260 they have max settings in Crysis at 1920x1200 No AA 16x AF:
Texture Quality:High
Objects Quality: Medium
Shadows Quality: Medium
Physics Quality High
Shaders Quality: High
Volumetric Effects Quality: Medium
Game Effects Quality: Medium
Postprocessing Quality: Medium
Particles Quality: Medium
Water Quality: High

Min=18 Max=46 AVG=30.1

The 260 is still a wee bit faster...

More reason to think a 4870 and a 260 are close in performance, and that the review sites that have a 4870 listed faster than a 280 are bogus...
 
This was DeadSkull's responce:

DeadSkull said:
How do you know? Do you personally own a gtx 280 and 4870cf rigs?

Higher fps for the 4870 vs 280 in cod4 testings could easily be explained by the cpus the review sites were using in the test rigs. Just wait and let me compile and arrange all the cpus and the fps for cod4.

First off we have anandtech.
- Intel Core 2 Extreme QX9770 @ 3.20GHz
- 4 x 1GB Corsair DDR3-1333 7-7-7-20

fps 4870
- 82.4 1920 x 1200 4xaa
gtx 280
- 90.9

Techreport
- Core 2 Extreme QX9650 3.0GHz
- 2 x Corsair TWIN2X20488500C5D
DDR2 SDRAM at 800MHz

fps 4870
- 69.1 1920 x 1200 4xaa 16aa. Notice the 10 fps drop b/c of slower cpu and slower ram.
fps 280
- 68.8 same settings. Hmmm so both cards are bottlenecked here. Alright whatever I'll just keep listing sites.

guru3d
- Core 2 Duo X6800 Extreme (Conroe)
- 2048 MB (2x1024MB) DDR2 CAS4 @ 1142 MHz Dominator Corsair
fps 4870
- 60(!!!) fps for 1920 x 1200 4xaa 16aa all settings maxed out. Note older slower dual core x6800 is only 2.9 ghz and only sports 4mbs total cache results in further 10 fps drop. So comparing to the anandtech 3.2 quad guru3d review gets 20 fps less because of cpu bottleneck. So I hope everyone sees a pattern here. BTW they didnt have a 280 in comparison but I just threw in fps for the 4870 because it was relevant.


hardware canucks.
testing rig. Now these guys really push it cpu wise.
- Processor: Intel Core 2 Quad Extreme QX9770 @ 3.852Ghz
- Memory: G.Skill 2x 2GB DDR2-1000 @ 1052Mhz DDR
Site didn't have 1920 x 1200 data but I'll just link it so all of you can see how the 4870 is on average 10 fps behind the 280 in most res.



In the end with if we take out the cpu bottleneck out of the equation and hopefully take out the ram limitation too, 4870 is a little bit behind the 280 in cod4 performance but not by a wide margin. Both in anandtech review and the hardware canucks review where the latest 45 nm quads were clocked 3.2 and above the delta inbetween the 4870 and the 280 at most was 10 fps so in the end in cod4 280 is a bit faster but not enough to guarantee its x2 pricetag over the hd4870. Hope that helps clarify whatever issues you guys had with various review fps numbers.
Click to expand...

I still am not sure the differences in system specs really explain this... :D
 
it is obvious you thought long and hard about this one and it is simple really, no one said the 4870 is faster than the GTX280. Quote from the Techreport conclusion "In practical terms, what all of this means is that the Radeon HD 4870, a $299 product, competes closely with the GeForce GTX 260, a $399 card based on a chip twice the size."

Quote from Anandtech conclusion "in our testing we definitely saw this $300 part perform at the level of NVIDIA's $400 GT200 variant, the GTX 260. This fact clearly sets the 4870 in a performance class beyond its price."

You are drawing conclusions off of a few games in which the 4870 was listed faster. Read the entire review and if not, read the conclusion. They all said what you say, the 4870 competes with the GTX260.
 
That's good to hear...

When you look at Race Grid: http://www.techreport.com/articles.x/14990/13

At 1920x1200:
They have the GX2 Min= 48, Max= 85.5
The 4870 at Min= 62, Max= 80
And the 280 at Min= 56, Max= 67.6

From this it looks like the GX2 is the faster card, but I would bet the 280 smokes it...


I guess we do need to see (2) 3870's in crossfire, against (1) 280.

I think it will be a better race than most, especially in a game that supports PhysX.
 
The different reviews are just opionions to be accepted or disposed of. I do tend to consider these sites as most reliable to me:

1. HardOCP
2. TheTechReport
3. Guruof3D

I have found going to different sites gives me a trend to follow to make a safe choice on a upgrade.
 
The 280 IS faster. in a few situations the 4870 comes close. Mostly the 4870 competes with the 260.

as others have said - for $300 - the 4870 is an amazing "value"

heck, i was *set* on a GT280 .. but $650 killed it for me. i didn't think at all when i pulled the trigger on my 4870 for $309. it is in my PC and i think it is a solid value. A step-up over my 8800GTX, but not "amazingly faster"

i am benchmarking it now with the hotfix drivers. My scores seem low with my 2900xt Crossfire edging out my HD4870 - so something may be wrong. i am installing 8.6 today and repeating my own benchmarking. But i am certain 280 will be faster[period]
 
My impressions are similar. By pretty much all reviews, my 8800GTS 512 sli (basically 9800gtx) should be faster than the GTX280. But the GTX280 feels a lot smoother in games and nothing seems to be able to phase it.

For example, crysis at 1920x1200 very high...there are plenty of parts that are just unplayable with the SLI. But with the GTX280, it stays playable all the way through although it's generally lower than I'd like...it even handles the ice levels and last boss fight.
 
Your 8800GTS 512 SLI is like the GX2 the then...

Looks good on paper, but the 280 will run faster, and allow for higher in game quality settings. :)

No microstutter to boot! :cool:
 
The thing is, the ATI fanboys are cherry picking only the benchmarks where the 4870 just so happens to beat the GTX 280 and claiming it's the winner. Now that's just based solely on performance. When price is taken into consideration, for something that is 10-13% slower, but costs less than half the price, then yes, the 4870 just beats the snot out of the GTX 280. The 4870's competitor is the GTX 260, where it's closer to most of the time, rather than the off chance that it comes close to the GTX 280 performance. But again, at the price point the GTX 260 comes in, it's pointless when the 4870 beats it handily.
 
I've seen a lot of posts where people are excited for the 4870... But I haven't read anyone claiming it's faster than the 280, fanboys or otherwise. Everyone just says that it's better than the 280 because it gives amazing performance at an amazing price. If the 280 were $400 instead of the 260, this would be a different game.

But we know nVidia probably can't drop the price like that, soooo... I want a 4870 =D
 
ya, no one claims the 4870 is outright faster than the gtx 280. the only times i've seen it claimed for certain benches/games are at like 2560x1600 8xAA, where some sites have given numbers that it is. Given that the 4800 series seems to take little hit for 8xAA (MSAA), it seems plausible, but i don't have gtx 280 or a 4870 to test, and in any case i don't have a 30" monitor either.
 
I'd say the fact that the cards are even comparable is a major win for ATI. For $300 you get a card that performs in between the GTX260 and the GTX280 but costs several hundred dollars less.
 
its like any other generation of cards, some will do better in some areas then others. I would think that they will be some areas that the 4870 will be faster, but by and large no. on the question of the better value........
 
I take them into account probably, if not many have complained at the credibility of the site and reviewer...
 
The GTX 260/280s suffer under 4x-8x AA, whereas the 4850/70 don't lose a lot from even 8x MSAA... techreport's review used AA and AF in all conditions, hence a more realistic gaming view with settings people actually play with. Other sites used no AA or no AF, or neither, creating more favorable bench settings, that don't pan out in actuality. The [H], TechReport's, and Guru3d's reviews tend to be around what you can expect, and in this case the TechReport seems to be the most correct (I have owned both a 280 and a 4870 (the 4870 died, but is being refunded) and noticed essentially equal performance in my own benching).
 
GoldenTiger said:
The GTX 260/280s suffer under 4x-8x AA, whereas the 4850/70 don't lose a lot from even 8x MSAA... techreport's review used AA and AF in all conditions, hence a more realistic gaming view with settings people actually play with. Other sites used no AA or no AF, or neither, creating more favorable bench settings, that don't pan out in actuality. The [H], TechReport's, and Guru3d's reviews tend to be around what you can expect, and in this case the TechReport seems to be the most correct (I have owned both a 280 and a 4870 (the 4870 died, but is being refunded) and noticed essentially equal performance in my own benching).
Click to expand...

It's strange that so many 4870s are dying, I wonder what ati is going to do about it. Or the qimonda memory perhaps can't take the extra pressure from oc.
I agree with you on that you would want to get at least som level of AA and AF in games.
 
Nuker said:
It's strange that so many 4870s are dying, I wonder what ati is going to do about it. Or the qimonda memory perhaps can't take the extra pressure from oc.
I agree with you on that you would want to get at least som level of AA and AF in games.
Click to expand...

To be fair, my 4870 was flashed with a modified BIOS, even though it was flashed back, and died. I thankfully had the very good luck to have a seller that would take it back anyway, so I have another one on order from amazon this time (Asus) that I'll never flash, in the past I'd only done software tweaking for video cards due to the risk, and this one time I tried it I got burnt.

I *don't* think it's the memory because I ran it at 790core/4400ram for hours just fine with Catalyst Control Center's overdrive feature, then started tweaking with flashes which it died from shortly after to try to pump the core higher. Upon further reading it seems MOST of the people with dead 4870s did flash, I've only seen one person insisting they didn't do anything (even overclocking) and had four (two in crossfire per rig) die on him which his vendor said may have been a bad batch as they'd had other reports from that same shipment.

Hopefully there isn't an actual issue... $300 high-end cards are good stuff :).
 
so true,

not reading correctly the reviews, but it is the best bang for the buck,

sam
 
Nuker said:
It's strange that so many 4870s are dying, I wonder what ati is going to do about it. Or the qimonda memory perhaps can't take the extra pressure from oc.
I agree with you on that you would want to get at least som level of AA and AF in games.
Click to expand...

it can't be the DDr5., imo

i got mine from 900 to 1100Mhz without trying; CCC settled on it with a core boost of only +30 mhz to 780 :p

it *appears* that too many users used an improperly modified BIOS - which seems to have been developed for it by someone that did not even have the card when he wrote it [or so i am reading at B3D]

it appears that O/C'ing is officially encouraged by Sapphire - it mentions "higher OCs" right on the box. But i doubt they will accept a GPU for RMA that has another BIOS.
 
GoldenTiger said:
The GTX 260/280s suffer under 4x-8x AA, whereas the 4850/70 don't lose a lot from even 8x MSAA... techreport's review used AA and AF in all conditions, hence a more realistic gaming view with settings people actually play with. Other sites used no AA or no AF, or neither, creating more favorable bench settings, that don't pan out in actuality. The [H], TechReport's, and Guru3d's reviews tend to be around what you can expect, and in this case the TechReport seems to be the most correct (I have owned both a 280 and a 4870 (the 4870 died, but is being refunded) and noticed essentially equal performance in my own benching).
Click to expand...


Don't mess so much with your replacement card. ;)

The OP fails to realize that it is only at certain setting and games in which the 4870 can beat a 280. In COD4 and at 1680x1050 the 4870 beats the 280. That is confirmed by two reviews I looked at. Up the res though and the 280 beats the 4870. In another thread I simply claimed the 4870 beats the 280 in some games and I claimed that in some other games it gives 90% performance of the 280. Those claims are correct for the resolutions and game settings I play at. That doesn't make me an ATI fanboi, the fact is that both of my PC's have 8800GT's in them right now so it is impossible to claim I am an ATI fanboi. Yea, the OP was directing his comments at me from what I can tell and he has taken what I said out of context.
 
anime4u said:
The different reviews are just opionions to be accepted or disposed of. I do tend to consider these sites as most reliable to me:

1. HardOCP
2. TheTechReport
3. Guruof3D

I have found going to different sites gives me a trend to follow to make a safe choice on a upgrade.
Click to expand...

You forgot PcPer :D
 
pcper didn't even bother to compare 4870 vs 260 or 280 in their latest review.

btw talon, thanks for putting up my mini comparison.
 
You must log in or register to reply here.
Back
Top