Radeon 3870 vs 7950 gx2

seachus

n00b
Joined
Nov 19, 2005
Messages
46
** Added 8800GT results and comments to conclusions **

Best buy last week had the Visiontek Radeon 3870 overclocked version for $140 out the door with no rebates so I picked one up. I liked the external exhaust aspect as it would help cool my chipset without having to crank up the noisiest fan in my system, the chipset fan.

So going into this I figured the 3870 and my 7950 GX2 were roughly the same speed. In TF2 I could run 2xAA and 16xAF with the Radeon where I ran no AA and 8xAF with the GX2, with both running between 30-60 fps @ 1920x1200 in crowded servers. Also a 7950GT (my older card) is about equal to a single 8600GT, where 2 8600GT's are about equal to a single 9600GT which is slightly faster than a 3870. So I was guessing the 7950 GX2 and 3870 are close with the Radeon being slightly faster.

Now let me preface this by saying the Radeon has better sound than my system does currently, however I have no way of tapping into the HDMI output to get the audio and video signal so it does me no good. Sucks cause I would love to use it but there is no way to get the sound to my non HDMI receiver/speaker setup. Also, as stated above the Radeon does help cool my system and the card itself is quieter than the two small fans on the GX2. And the 3870 has better image quality in general compared to 7 series cards like the GX2.

Question: Is the performance gain, if any, and other benefits = to the 140 bucks?

I did a little testing using 3dmark's various versions and some play time on both in TF2 and civ4 using XP 32bit. Ram shows up as 3.1 GB with the Radeon vs 2.75 with the GX2. Switching to vista x64 didn't really change the scores and being that I have no dx10 games yet I couldn't say if dx10 on the Radeon was superior to the same game with dx9 on the gx2.

Test system is system in sig with cpu @ 2.7 and the 4 gigs of ram @ 225. No additional oc applied to either card, although the Visiontek is factory oc'd slightly.

Below are my test notes.

How I am going to test the 3870 vs the 7950 GX2:

Reboot before you run the benchmark set, a benchmark set being 3
runs of a 3dmark. Let the system boot as normal with wireless and
smart guardian running in the background. Have this text file open,
run the set, write down the results after each run. Close the
benchmark and fully launch again for each test within a set, but do not reboot.

Card 1: 3870

3dMark03

Run 1: 16589
Run 2: 16601
Run 3: 16178

3dMark05

Run 1: 13345
Run 2: 13372
Run 3: 13385

3dMark06

Run 1: 8139
Run 2: 8130
Run 3: 8149


Card 2: 7950 GX2

3dMark03

Run 1: 26659
Run 2: 26626
Run 3: 26710

3dMark05

Run 1: 12654
Run 2: 12738
Run 3: 12733

3dMark06

Run 1: 7884
Run 2: 7904
Run 3: 7886


Card 3: 8800GT

3dMark03

Run 1: 31860
Run 2: 31886
Run 3: 31155

3dMark05


Run 1: 14491
Run 2: 14499
Run 3: 14507

3dMark06


Run 1: 9699
Run 2: 9730
Run 3: 9743

Conclusions: 3870 v 7950

1. GX2 is much faster on older stuff, 3870 is faster in newer stuff.
2. Higher memory bandwidth of 3870 (almost twice) allows better image quality at higher resolutions through the use of higher AA and AF settings
3. 3870 allows me to use dual monitors again (stupid SLI)
4. 3870 is fast enough in all old games
5. 3870 lowers chipset temps by 5 degrees C and allows quieter idle operation of computer.
6. GX2 is already here cost = nothing
7. TF2 plays exactly the same, AA at 1920 is hardly noticeable nor is the extra AF
8. The 3870's superior image quality is there, but I never really noticed it, maybe I'm too busy playing the games.
9. Civ4 still shimmers on the map, which I thought was a 7 series nvidia issue, but evidently its a game issue?
10. 8800gt is 20 bucks more than the Radeon...
11. New cards come out real soon...

Yeah so I think I'm going to return the card, 140 bucks for minor niceties and a slight performance boost is not worth it to me. I think I'm going to buy an 8800 gt and do the same set of tests...

Conclusions: 3870 v 7950 vs 8800GT

Returned Radeon, bought gt for 20 bucks more.

1. GT was the quietest of the 3 even though it was a single slot cooler.
2. Radeon cooled the chipset the best, but the gt was not far behind.
3. GT allowed the highest settings in all games 4xAA was possible in TF2 and Civ4
3b. TF@ at 4xAA 16xAF and max settings including very high texture detail gave a nice 40 - 80 fps performance. Radeon at 2xAA 16xAF was 30-60, 7950 @ 0xAA 8xAF is 30-60
4. Radeon allowed 2xAA in my games
5. 7950 GX2 has same resolution, but no AA and usually 8xAF instead of 16x
6. TF2 does look better at 4xAA vs no AA
7. Civ4 with 4xAA and smooth frame rates is nice, but still cpu lag is a bigger issue late in the game, graphics were never the limiting factor.
8. The difference in the games and such was sort of blah, thats nice but not really that different and definitely not worth the money.

When it comes down to it, my compy sits idle most of the time running @ 1.35 ghz. The Visiontek had a bug that didn't allow it to slow down its clocks to 2d running speeds so it was always running full blast at 800mhz even when idle at the desktop. This bug made the Radeon louder than it should have been, and allowed the GT to be the quieter card. Overall the new cards did not allow me to play anything better, just made things look a little nicer. Pretty is not worth 160 bucks.

Returned all cards and stuck back in my 7950 GX2 for another 6 months or so.
 
I agree with your assessment(besides for the 8800gt part), take it back and pickup a 4850 or 4870 when they come out. You will be a lot better off while spending only $70-90 dollars, given the rumored cost.
 
what about the power consumption & SLI efficiency in different games?

SLI efficiency never came into play, nor did I care. I tested older games and basic computer usage. To me if SLI didn't work then yeah it would perform bad, but it works in everything I use my computer for so SLI from a performance standpoint is a moot point. Now I *HATE* SLI because of its lack of multiple monitor support, but I only have 1 24 inch monitor at the moment so until I get a second equal or larger I am not putting my 19 back next to it.

As far as power consumption is concerned, I didn't hook a power meter up when I had the cards in the system, but my 7950 GT at full load was 230W, the 7950 GX2 at full load was 280W. I imagine the Radeon and the 8800GT both consume more, but both are close in total power consumption. Most review sites list the 8800GT as more efficient under load but overall power usage is similar between the Radeon and the 8800gt. My guess is my system was tipping 300W with the new cards. All cards have a single pci-e connector, and a max draw of 75 + 75 = 150W so the cards are not going to be that power hungry in the grand scheme of things. Again similar performance, similar power draw, yet another reason not to upgrade yet.
 
With respect 3D Mark 03/05/06 are near worthless benchmark and somewhat misleading in this situation.

The 7950 GX2 will show decent numbers against the 3870 and 8800GT however as widely reported its SLI performance in many games varies greatly. The 3870 will destroy the 7950GX2 in most newer games and the the gap is even wider with the 8800GT (which you kinda indicated).

Sure if you only play older games and the FPS and IQ are fine to you then you don't need a new card.

Bottom Line
Loose the 3D mark testing lol, if you had used Crysis, Mass Effect, AOC and fraps you would have seen a massive difference, and you would have not wasted the retailers time by returning all the cards (and effectively making them second had, I am sure they really appreciated that).
 
Back
Top