4850CF performance in Warhead and ClearSky

[twjr]

I got my two 4850's back when they were released and it seemed like every man and his dog were getting a couple. I have been running them in Crossfire ever since i got them and have been pretty happy with them thus far.

They took everything in their stride (except Crysis) at max settings at 1650x1080. Lately I have been playing a lot of Warhead and Clear Sky. Warhead was just playable at enthusiast settings, albeit with the odd lag. Clear Sky on the other hand was only playable at high settings until the latest drivers which allowed maximum settings (though the somewhat unpredictable slowdowns still happen).

With the warming weather (I live in the Southern Hemisphere) i thought that i should probably give a little more airflow to my cards, so put another fan under my bottom card. By the way is 95 C safe? So for the first time since I got the cards I took one out. While I only had one card in I thought I might as well see the difference that the extra card made so fired up a couple of games. Well it turns out less than I thought.

I tried Warhead first and to my curiosity it seems that the extra card makes no difference. While I didn't take any benchmarks, according to FRAPS and my memory, it seems that there is probably only 2-5 frames difference between the single and dual cards on the levels I tested.

Clear Sky was not as cut and dry. While CF appeared to have higher max FPS it was also less consistent and appeared less smooth than a single card. Also I have noticed a large number of slowdowns to around 5 FPS with CF that didn't seem to occur with a single card.

Just to see if i could see this anywhere else I tried Oblivion. Unfortunately due to vsync being on the FPS pretty much held around 60 which pretty much just says that one 4850 can max out this game.

And no to the point of this story. I have come up with a few possible ideas behind this.
1 Poor scaling
2 CPU limited
3 I have botched something setting up Crossfire

Sorry for the rather long rant, this just irks me and I was wondering if anyone else had some ideas?

Oh and by the way my CPU is a Q6600 @ 3.0

 

[Hauk]

Warhead on enthusiast is pretty ambitious, even at 1680x1050. You may want to drop to high settings. DX10 and 2 or 4xAA with high settings should run nicely. Can't comment on Clear Sky. It's always good to confirm CF is working by checking with GPU-Z as well. Should show enabled in both CCC and GPU-Z screen. As far as scaling, it's hit and miss. I ran a 4850 CF rig and it seemed my second card wasn't doing much work. At full load it's GPU was 10c cooler than primary card GPU.

As far as temps, it's never a good idea to operate at the high end of a tolerance spectrum. Sounds like you've not made any manual fan speed adjustments? The latest CCC drivers have a manal fan speed adjustment. You should be running both fans at minimum 35%. Watch your idle and load temps drop dramatically. How about case ventillation? Those things are heaters inside your case. How is airflow, specifically exaust? How many, what size and speed fans?

On cpu, I had a Q6600 and Q9550 at 3.2, but have seen decent gains with an E8600 at 4.2. How cpu bound you are, maybe someone can provide actual data. You probably are a bit, but proper eye candy settings and insuring CF is working are important here. If mobo and cooling allow, you may try 3.2+ on your Q6600.

 

[apoppin]

First of all, FC2, Clear Sky does scale pretty well:

http://forums.anandtech.com/me...=2243797&enterthread=y
[2nd post]

Warhead is basically the same engine as Crysis and it does not scale as well. Not to mention, you are a bit CPU limited at 3.0Ghz; especially for CF

.. and 95C is pretty near the extreme end of warm; i'd play with fan speeds and case cooling.
:sun:

i am SO glad Winter gaming is beginning for me

 

[twjr]

Ok well if Warhead has poor scaling that makes me feel somewhat better though I was seeing nearly no scaling at all.

After doing some more testing between one and two cards I have noticed that Stalker Clear Sky does scale reasonably well. However, the problem is not the lack of fps in CF but random drops in speed to about 5 fps before returning to normal (is amazingly annoying if happens during a firefight). These happen somewhat at random not just when there are lots of lighting effects as I've read elsewhere. I played a whole day with a single card and was getting about 2/3 fps of CF but not a single slowdown happened. I just updated to the latest patch which has some graphical improvements so will see what difference that has.

About my temps. I suspect I've got a stuffed (warm running) card (55 C idle to 95 load). I have done fan speed mods. I run at about 5% higher fan speed than the temps, so about 60% at idle on this card. I have got my case open so ventilation isn't a problem and currently have one of the ridiculously high RPM fans from a T-Bird cooler venting air out of the open PCI slots in the bottom of my case. I have also replaced to stock thermal paste with AS5 which made no difference. With the same fan settings my other card runs about 20 C cooler. I just realised that I was running a slight OC 650 core 1000 memory. But returning to stock settings made no difference. The only thing I think that could be wrong is that I might have accidentally changed a voltage setting when adjusting the idle speed settings in the BIOS. Is there any way to check that?

I'll try pushing a bit more OC, say to 3.3, and see what that does too. Though I'm having a couple of odd things happen here too, but I'll head over to the CPU forum for that.

 

[apoppin]

You have GOT to realize that STALKER always had this weirdness with the drops and the this is the same engine - and DX10 really slows it down
- at times

Wait for a couple of new CS patches and also Cat 8.11 or 8.12 .. by then you will be done with the game .. but it will run better

 

[twjr]

But why with CF and not a single card? Or is that just part of it's weirdness?

Also I just ran my fan for 3 hours on full and it only got down to 48C so something is not quite right. And I knew those fans spun fast but 9500 RPM :shocked:

 

[apoppin]

Originally posted by: twjr
But why with CF and not a single card? Or is that just part of it's weirdness?

Also I just ran my fan for 3 hours on full and it only got down to 48C so something is not quite right. And I knew those fans spun fast but 9500 RPM :shocked:

the CF'd cards generally have the same minimum as the single card; but it is just for a moment, so i don't think they will ever completely solve it
.. and with 8-10 stock cats, my X3 does not run at all with Stalker, although X2 is fine and scales good also

don't run your fan all the time .. especially not on high

 

[twjr]

Originally posted by: apoppin

Originally posted by: twjr
But why with CF and not a single card? Or is that just part of it's weirdness?

Also I just ran my fan for 3 hours on full and it only got down to 48C so something is not quite right. And I knew those fans spun fast but 9500 RPM :shocked:

the CF'd cards generally have the same minimum as the single card; but it is just for a moment, so i don't think they will ever completely solve it
.. and with 8-10 stock cats, my X3 does not run at all with Stalker, although X2 is fine and scales good also

don't run your fan all the time .. especially not on high

My CF set up has a much lower minimum due to the lags. I thought there may be some CF specific problem which is why I don't get the lags with only one card.

What are safe temps though. I mean I don't like seeing my temps almost reaching boiling. Is it possible that my chip just runs hot?

 

[apoppin]

there probably is a CF specific problem; considering the immaturity of the driver and the newness of the Clear Sky - these are new DX10.1 effects for their engine - i say, expect issues. i was slightly kidding, but by the time you are done with the game, it will be sorted out - to STALKER's consistency of performance [which although not great, is leaps over when it first came out bug-ridden]. Remember i reported that my CFx3 wouldn't run with STALKER and Cat 8-10; it was fine with 8.9
- but of course 8-9 predates Clear Sky and you should use 8-10 or the hotfix

 

[twjr]

In terms of the game I can't believe that they are able to release with so many bugs. I mean the save game corruptions alone are pretty bad. Too bad like any other product that you can't return it even if it is defective.

I played it through with 8.9 and the performance was really quite sub-par. The latest hotfix gave a huge increase in performance and stability so like you said it will only get better. At least the game has some replayability.