• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

5970 under performing!

S

SuperBrandon

n00b
Joined
Mar 26, 2010
Messages
10
I got my XFX 5970 Black Edition in today.

I'm thinking I want to go back to 2x5850's, and here's why:

Im using Call of Pripyat benchmark and here's my results:

Code: 
[B]2x5850 Completely Stock (725 Mhz / 4000 Mhz)[/B]

BENCH | MIN FPS | AVG FPS | MAX FPS
-----------------------------------
Day   | 37.4    | 141.5   | 195.5
-----------------------------------
Night | 39.8    | 130.6   | 201.5
-----------------------------------
Rain  | 72.8    | 144.4   | 214.9
-----------------------------------
SunSh | 41.5    | 71.9    | 109.0



[B]Completely Stock 5970 (725 Mhz / 4000 Mhz)[/B]

BENCH | MIN FPS | AVG FPS | MAX FPS
-----------------------------------
Day   | 30.8    | 134.2   | 194.9
-----------------------------------
Night | 51.7    | 124.0   | 195.2
-----------------------------------
Rain  | 55.1    | 139.3   | 209.6
-----------------------------------
SunSh | 42.3    | 77.4    | 114.6



[B]OC & OV 5970 (920 Mhz / 4500 Mhz)[/B]


BENCH | MIN FPS | AVG FPS | MAX FPS
-----------------------------------
Day   | 28.8    | 136.5   | 200.4
-----------------------------------
Night | 44.7    | 127.3   | 198.6
-----------------------------------
Rain  | 58.5    | 143.6   | 210.9
-----------------------------------
SunSh | 45.7    | 78.1    | 116.8


I have very minimal gains in some areas, and surprising losses in others!

Does this seem right to anyone else? Any ideas?
 
that is very weird. only thing would be if you had task running in back ground with 5970 and 2x5850 nothing. but the oc scores really puzzling since not getting much improvement at all with oc and ov to past 5870 levels on core clock and almost 5870 levels on mem. what cpu your using that the only other thing i can think about. might be cpu limited
 
Well I cant find where to modify my signature but here's the specs:

i7 930 @ 4.4ghz
12 GB G.Skill 2000 mhz ram @ 9-9-9-24 (@1600mhz)
evga x58 mobo
PPC&C 850w PSU
XFX 5970 Black Edition (920MHz/1125MHz)


CPU limited, wha?? lol j/k

this sucks though! highly disappointed in my $730 impulse buy of doom!
 
well definitely not cpu limited. tried doing a fresh install with the drivers? as for sig click on user cp then edit signature.
 
yeah this is on a fresh install of the drivers. im running out of ideas. i just ran 3DMark06 and scores 27306... i pulled 27980 with my 2x5850 =\
 
What resolution and settings are you using? you won't notice much improvement with a 5970 over your previous setup unless you were becoming GPU bound, which unless you're running 1920x1200 or higher with AA and max settings you probably weren't.
 
i would have to agree with violator. look at the benchmarks, sunsh test was about 1/2 as many FPS as the other tests. in this test the 5970 scored better. also wasn't the 5970 slightly lower clock then then a 5870. even your chart shows the 5970 at the same clock speed as the 5850 so it is not going to preform like 2x5870.

edit: tehquick makes a good point about the OC test. that is a hefty OC with little gains. i would do a little more testing before you make a decision.
 
Last edited:
no matter what your going to see some frame rate loss running a 5970 vs 2 5850's on crossfire.. since thats the nature of the beast with a dual gpu card running a transparent crossfire bridge on a single card vs running 2 different cards and a crossfire bridge..

you may want to do some more testing with other games because it could just come down to a lack of crossfire support/optimization in the game limiting the setups.. ya might want to use a single card vs the 5970 just to see how well the games scaling in crossfire..
 
You should definately try out more games, but also download MSI's Afterburner Utility. The monitoring portion of it will give you a good idea by giving you a graph of your real-time temps, GPU clocks, and so on. I'm not sure if this utility works with dual GPU cards, but this is what I use. If your GPU is overheating and throttling down, you'll see it in the graph with this utility. You'll also see a real-time plot of your fan speeds, etc...

Also test more games just to be sure. I also noticed that your motherboard uses an x58 chipset, which I believe means that you maintain two full 16x PCIx slots when you crossfire a card. This means that each card has access to a full 16x speed versus most motherboards that go to 8x when crossfired. As an above poster stated, it may also have to do with bandwidth available as well since your 5970 is sharing a single slot instead of each GPU having its own 16x slot.
 
Look like it may be in fact throttling. Here is a screenshot of afterburner near the end of the COP benchmark. It has a nice dip in it. But throttling at around 95 C ? I thought the threshold was a little higher then that.

afterburner.png
 
What Afterburner doesn't show is the VRM temps, which can also throttle the card; GPU-Z can show you these (called VDDC temps in GPU-Z.)
 
I upgraded the BIOS on this 5970 to support manual fan settings. I cranked it up to 100% and ran CoP on all max settings to compare with someone's 2x480GTX rig. Here's the results:

Code: 
Call of Pripyat Bench - All Max Settings
----------------------------------------

2x480GTX in SLI

BENCH | MIN FPS | AVG FPS | MAX FPS
-----------------------------------
Day   | 25.2    | 66.8    | 89.3
-----------------------------------
Night | 42.0    | 67.1    | 89.0
-----------------------------------
Rain  | 39.9    | 69.8    | 89.9
-----------------------------------
SunSh | 32.8    | 53.2    | 74.8



OC & OV 5970 (920 MHz / 4500 MHz)

BENCH | MIN FPS | AVG FPS | MAX FPS
-----------------------------------
Day   | 30.1    | 68.9    | 97.6
-----------------------------------
Night | 33.4    | 57.9    | 81.0
-----------------------------------
Rain  | 37.2    | 61.1    | 82.4
-----------------------------------
SunSh | 26.3    | 44.2    | 67.3

Look right?
 
SuperBrandon said:
I upgraded the BIOS on this 5970 to support manual fan settings. I cranked it up to 100% and ran CoP on all max settings to compare with someone's 2x480GTX rig. Here's the results:

Code: 
Call of Pripyat Bench - All Max Settings
----------------------------------------

2x480GTX in SLI

BENCH | MIN FPS | AVG FPS | MAX FPS
-----------------------------------
Day   | 25.2    | 66.8    | 89.3
-----------------------------------
Night | 42.0    | 67.1    | 89.0
-----------------------------------
Rain  | 39.9    | 69.8    | 89.9
-----------------------------------
SunSh | 32.8    | 53.2    | 74.8



OC & OV 5970 (920 MHz / 4500 MHz)

BENCH | MIN FPS | AVG FPS | MAX FPS
-----------------------------------
Day   | 30.1    | 68.9    | 97.6
-----------------------------------
Night | 33.4    | 57.9    | 81.0
-----------------------------------
Rain  | 37.2    | 61.1    | 82.4
-----------------------------------
SunSh | 26.3    | 44.2    | 67.3

Look right?
Click to expand...

The results now look even worse though if you compare them against what you were getting with 2x5850s. Unless that first benchmark is done with completely different settings? You still haven't mentioned your resolution btw or if you're using AA.
 
something is wrong with those results in the first post m8, no way losses like that. how many loops?? maybe the OC was unstable?? maybe the GPU was throttling due to heat??
 
Okay I reran the same benchmark settings I had in my very first post here. With manual fan speed set to 100% for the entire bench, I completely smashed those scores. Here they are:

Code: 
BENCH | MIN FPS | AVG FPS | MAX FPS
-----------------------------------
Day   | 33.1    | 162.0   | 235.8
-----------------------------------
Night | 63.1    | 153.0   | 231.1
-----------------------------------
Rain  | 63.8    | 171.5   | 250.9
-----------------------------------
SunSh | 50.9    | 95.0    | 141.8

Also I reran 3DMark06 with 100% fan the entire run and got 28231. My two 5850 setup scored 27964. Still a little dissappointed by the 3DMark score, expected much better gain.

Well, it simply looks like heating is an issue with this card for me. I'll need to get it on water cooling to keep it on the top end of things.

However, I'm still thinking that perhaps I should have went with 2x5870. What is everyone elses thought on this?
 
The problem I see here is that you had to set the fan speed at 100% to keep things moving along. Something just doesn't seem right for a brand new stock card as i've always thought that auto fan settings should be able to keep temps in line with a stock card. If you bought it brand new, you might be able to exchange it for another one if possible at whichever store. I know that those XFX 5970's have a decent overclock, but I don't recall anyone else having your type of problem. In fact, I have seen a few sigs at this forum with people running really good overclocks on their 5970 with the stock cooler. Maybe someone else more experienced could chime in though as i'm still trying to relearn what i've been missing out during my 5 year hiatus from PC gaming.

SuperBrandon said:
Okay I reran the same benchmark settings I had in my very first post here. With manual fan speed set to 100% for the entire bench, I completely smashed those scores. Here they are:

Code: 
BENCH | MIN FPS | AVG FPS | MAX FPS
-----------------------------------
Day   | 33.1    | 162.0   | 235.8
-----------------------------------
Night | 63.1    | 153.0   | 231.1
-----------------------------------
Rain  | 63.8    | 171.5   | 250.9
-----------------------------------
SunSh | 50.9    | 95.0    | 141.8

Also I reran 3DMark06 with 100% fan the entire run and got 28231. My two 5850 setup scored 27964. Still a little dissappointed by the 3DMark score, expected much better gain.

Well, it simply looks like heating is an issue with this card for me. I'll need to get it on water cooling to keep it on the top end of things.

However, I'm still thinking that perhaps I should have went with 2x5870. What is everyone elses thought on this?
Click to expand...
 
This may sound stupid,but ive seen it happen were a psu wasn't up to par and it slowed the gpu down..and no it didn't crash,just low fps...just a thought..
 
amdgamer said:
The problem I see here is that you had to set the fan speed at 100% to keep things moving along. Something just doesn't seem right for a brand new stock card as i've always thought that auto fan settings should be able to keep temps in line with a stock card. If you bought it brand new, you might be able to exchange it for another one if possible at whichever store. I know that those XFX 5970's have a decent overclock, but I don't recall anyone else having your type of problem. In fact, I have seen a few sigs at this forum with people running really good overclocks on their 5970 with the stock cooler. Maybe someone else more experienced could chime in though as i'm still trying to relearn what i've been missing out during my 5 year hiatus from PC gaming.
Click to expand...

Humm might be the heatsink need reseating. that would explain the throttling at least. try running the card on stock with 100% fan. if the scores are different then i would definitely say reseat heatsink. If you don't want to risk warranty rma the card.
 
came across this today. seems that 2 5850 cf outperform a 5970 in certain games. don't know why or how but they do. see here. that might explain so of your score at first. though i would still check to see if heatsink is seated properly. that would explain the oc throttling since so many other got there to work with need 100% fan.
 
SuperBrandon said:
I upgraded the BIOS on this 5970 to support manual fan settings. I cranked it up to 100% and ran CoP on all max settings to compare with someone's 2x480GTX rig. Here's the results:

Code: 
Call of Pripyat Bench - All Max Settings
----------------------------------------

2x480GTX in SLI

BENCH | MIN FPS | AVG FPS | MAX FPS
-----------------------------------
Day   | 25.2    | 66.8    | 89.3
-----------------------------------
Night | 42.0    | 67.1    | 89.0
-----------------------------------
Rain  | 39.9    | 69.8    | 89.9
-----------------------------------
SunSh | 32.8    | 53.2    | 74.8



OC & OV 5970 (920 MHz / 4500 MHz)

BENCH | MIN FPS | AVG FPS | MAX FPS
-----------------------------------
Day   | 30.1    | 68.9    | 97.6
-----------------------------------
Night | 33.4    | 57.9    | 81.0
-----------------------------------
Rain  | 37.2    | 61.1    | 82.4
-----------------------------------
SunSh | 26.3    | 44.2    | 67.3

Look right?
Click to expand...

Seems reasonable. It looks like the GTX480 SLI is getting about 5-10% more performance, but, costs around $300 more. So, yes I'd say your getting pretty reasonable speeds for your hardware /w its overclock and what you paid for it.
 
coldstorm said:
came across this today. seems that 2 5850 cf outperform a 5970 in certain games. don't know why or how but they do. see here. that might explain so of your score at first. though i would still check to see if heatsink is seated properly. that would explain the oc throttling since so many other got there to work with need 100% fan.
Click to expand...

The only benchmarks i've seen where two 5850's in crossfire have outperformed a 5970 is in a few games when the test motherboard is using the Intel x58 chipset where each video card has access to a dedicated 16x pcix lane instead of dropping to 8x/8x in crossfire mode. With a 5970, two GPU's have to share a single 16x pcix lane. Apparantly the difference in bandwidth makes a difference in a few games. I believe Far Cry 2 might be another game where a crossfire 5850 can outperform a 5970. However he is running a black edition 5970 which has a pretty good overclock from the factory.
 
You must log in or register to reply here.
Back
Top