Catalyst 8.8 Released

[Bateluer]

https://a248.e.akamai.net/f/67..._88_release_notes.html">Text</a>



We now have two topics on this:

http://forums.anandtech.com/me...=2219508&enterthread=y

Since the title edit of this one came after the other thread was posted, I?m locking this one.

Video Mod BFG10K.

 

[Modular]

When are these due out?

 

[Bateluer]

Originally posted by: Modular
When are these due out?

This month.

 

[ronnn]

Hope this is finally true, as lack of power play is a major fault of the 4800's.

 

[Sylvanas]

This has been talked about on Rage3d for a while- it appears this driver is going to drop power draw (and in turn, temps) quite a bit.

 

[idiotekniQues]

lookin forward to installing these when they come out

 

[Leon]

The problem with power draw is related to highly clocked DDR5 - drop the frequency to, say 200, and power and temperature decreases dramatically. Unfortunately, this cannot be automated at the moment - changing the memory frequency results in annoying screen flash. Hopefully AMD will fix this issue ... in the meantime, use AMD GPU clock tool and create separate profiles.

Leon

 

[QuixoticOne]

404 Page not found at the link for me.

Anyway I don't quite understand about powerplay not working on the 48xx boards. I have a 4850, and I'm under the impression that in my case the factory BIOS simply lists 500MHz at some nominal voltage as being the only available clock speed for the thing at any of the powerplay settings, so even with a "working" powerplay *driver* it puts it into an lower clock rate / voltage in the idle/2D state that just happens (because of the BIOS table) to be the same settings it would use for 3D, so it doesn't save much any power. Similarly the fan speed table in the BIOS just isn't programmed to increase the RPM vs temperature much so it doesn't happen no matter what the driver does with Powerplay.

One thing that would fix it is an explicit override of the BIOS MHz / voltage / fan table settings, ignoring those, and using driver based defaults instead.
The other thing that would fix it is to change the BIOS tables for Powerplay MHz / volts and the default fan settings in the BIOS. Then a proper Powerplay driver would switch frequencies down to 160 MHz or whatever is the lowest correct operating value.

I believe there are additional levels of Powerplay support that they may not even have enabled in past drivers for dynamic power consumption savings, and I look forward to those. However in the most gross case, a completely idle GPU in 2D mode, we should ALREADY be seeing Powerplay causing maximum savings via reducing clocks / volts to a very low level, and for people with working BIOS versions this already works, and for people with bad BIOS versions it does not and may well not ever work until they fix the BIOS or override it.

 

[Bateluer]

AMD's dual-GPU HD 4870 X2, code name: R700, is expected on 12 August dieses Jahres offiziell vorgestellt. August this year officially presented. Zu diesem Zeitpunkt soll offensichtlich auch ein neuer Catalyst bereitstehen, der nicht nur den ersehnten Stromsparmodus Powerplay 2.0 einführen soll, sondern auch die ersten Preview-Benchmarks noch einmal auf den Kopf stellen soll. At that time also seems to be a new Catalyst them, not only the desired power-saving mode Powerplay 2.0 introduce, but also the first preview benchmarks again on its head. Das "ATI-Forum" will dies jedenfalls aus angeblich geheimen Kreisen in Erfahrung gebracht haben. The ATI Forum "will any of this alleged secret circles have brought in experience.

Der so genannte Catalyst 8.8 soll signifikante Leistungssteigerungen im Dual-Betrieb, also unter Crossfire, mitbringen und insbesondere der HD 4870 X2 zu neuen Rekordwerten verhelfen. The so-called Catalyst 8.8 will be significant performance increases in dual-operation, also under Crossfire, and in particular to bring the HD 4870 X2 to help new record figures.
Mittels einer ersten Beta-Version präsentierte man gleich schon einmal einen Vorgeschmack. With a first beta version presented to the same ever been a foretaste. Im 3DMark Vantage legte ein HD 4850-Crossfire-Gespann um 12 Prozent gegenüber dem Catalyst 8.7 zu. In 3DMark Vantage, an HD 4850-Crossfire-team by 12 percent compared with the 8.7 Catalyst.
Auch der 2D-Stromsparmodus Powerplay 2.0 soll endlich funktionieren. The 2D power-saving mode Powerplay 2.0 will finally work. Auf selbem System sank der Stromverbrauch im Idle von 211 auf 174 Watt deutlich. At the same system decreased power consumption in Idle of 211 to 174 watts.

Dies könnte ein packendes Treiberfinale werden, schließlich soll Konkurrent Nvidia bereits morgen mit einem Treiber aufwarten, der auf allen Geforce 8, 9 und GTX-Karten PhysX-Support freischaltet. This could be a thrilling finale driver, after all, it should rival Nvidia tomorrow with a driver aufwarten, at all Geforce 8, 9 and GTX cards PhysX support unlock. Im September kommt mit dem Geforce 180 dann sogar das große "Big Bang II" . In September, with the Geforce 180 then even the great "Big Bang II".
Blättert man in der Grafikkarten-Historie, so trifft man äußerst selten auf "geschenkte Performance", die per Treiberupdate nachträglich bereitgestellt wird. Scrolls found in the graphics card history, so we are extremely rare "geschenkte Performance", by the driver update will be provided later. Nutzer beider Chipschmieden sollten also die Gelegenheit beim Schopfe packen und die Gaben dankend annehmen. Users of both chip forging should the opportunity Schopfe pack and gratefully accept the gifts.

 

[Creig]

Originally posted by: Leon
The problem with power draw is related to highly clocked DDR5 - drop the frequency to, say 200, and power and temperature decreases dramatically.

Are you sure about this? One of the selling features of GDDR5 is that it consumes less power than GDDR3 (1.5v supply voltage, low power modes, data and address bit inversion, etc).

 

[AzN]

Originally posted by: Leon
The problem with power draw is related to highly clocked DDR5 - drop the frequency to, say 200, and power and temperature decreases dramatically. Unfortunately, this cannot be automated at the moment - changing the memory frequency results in annoying screen flash. Hopefully AMD will fix this issue ... in the meantime, use AMD GPU clock tool and create separate profiles.

Leon

I doubt memory mhz speed eat away power. It's likely the core.

 

[Nafets]

For the HD4870 it's the MEMORY that is the power hog;

2D IDLE testing:

GPU/MEM
500/900 = 139W
160/900 = 138W
500/225 = 101W

Lowering the GPU clock from 500MHz to 160MHz saves only 1W.
Lowering the MEM clock from 900MHz to 225MHz saves 38W.

As Leon stated earlier, the only problem with using modified 2D IDLE MEM clocks in the BIOS is that you'll get constant on-screen flashing...

 

[Elcs]

Lets hope they do reduce any power draw and temperatures.

I might be able to cut down on another slither of PC noise if it does

 

[Leon]

I doubt memory mhz speed eat away power. It's likely the core.

No, memory is the issue. I created a custom BIOS with low memory value (200 mhz) for low power powerplay state, and while it works (and significantly reduces power and temperature), the screen corruption every time the cards switches between low/high power state is unbearable.

Leon

 

[Foxery]

Originally posted by: QuixoticOne
I don't quite understand about powerplay not working on the 48xx boards. I have a 4850, and I'm under the impression that in my case the factory BIOS simply lists 500MHz at some nominal voltage as being the only available clock speed for the thing at any of the powerplay settings, so even with a "working" powerplay *driver* it puts it into an lower clock rate / voltage in the idle/2D state that just happens (because of the BIOS table)

For what it's worth, the powersaving mode in the 3850/3870 sets the core clock to 300 MHz. The memory clock on mine doesn't drop significantly... but mine is overclocked, so I'm not positive what the default behavior is for the RAM.

It continues to baffle me how ATI could miss this before sending the design out the door. :roll:

 

[Bateluer]

More information on the 8.8s below.

Mad Shrimps claims PowerPlay is working under 8.8 betas

OC Forums Count Down thread - Claims Tuesday 19 Aug as the launch date

 

[ajaidevsingh]

Great news

 

[Alex C]

I'm a new, what exactly is Catalyst 8.8 and powerplay? Is it only for cards in CF, or will it pertain to a single 4850 as well?

 

[Nafets]

Lower power consumption from Catalyst 8.8 will only pertain to Crossfire setups. Single card setups most likely won't see any difference...

 

[Bateluer]

Originally posted by: Alex C
I'm a new, what exactly is Catalyst 8.8 and powerplay? Is it only for cards in CF, or will it pertain to a single 4850 as well?

Catalyst 8.8 is the next revision of ATI's Catalyst drivers for their products. ATI names their drivers after the year and month. Catalyst 8.8 is 2008, August release. 2008, September will be Catalyst 8.9.

Originally posted by: Nafets
Lower power consumption from Catalyst 8.8 will only pertain to Crossfire setups. Single card setups most likely won't see any difference...

The links and early tests suggest otherwise. If Powerplay is enabled, it should effect all 4000 series products, regardless of whether or not they are in Crossfire or single card.

 

[Foxery]

Originally posted by: Nafets
Lower power consumption from Catalyst 8.8 will only pertain to Crossfire setups. Single card setups most likely won't see any difference...

That's not true. Don't know where you got that idea. The Powerplay update will fix any and all 4000 series cards, in any configuration.

 

[Nafets]

I hope you're both right, and I'm wrong.

Regardless, Powerplay is already working properly for both cards, and has been since Catalyst 8.7.

The function of Powerplay is to detect the load on the GPU and apply the proper performance state, using values set in the video card's BIOS.

I don't see a way they can "fix" or lower power consumption (especially for the HD4870) as the 2D/3DLP/3DHP values that PowerPlay applies are set in the BIOS.

ATI isn't going to arbitrarely use GPU/MEM and voltage values that it's thinks are appropriate. It just isn't possible as every card operates differently with lower GPU/MEM clocks.

My HD4870 runs fine at 160/225 @ 1.083v lowered from the default 500/900 @ 1.263v, in 2D IDLE. But you can't say that every HD4870 would function properly at those speeds.

 

[Jessica69]

Originally posted by: Nafets

My HD4870 runs fine at 160/225 @ 1.083v lowered from the default 500/900 @ 1.263v, in 2D IDLE. But you can't say that every HD4870 would function properly at those speeds.

Why not? Each and every 4870 is built on a reference design....and they should all act exactly alike....

 

[Nafets]

For the same reason that every card doesn't overclock to the same limits. Running at extreme limits on either end (overclocking or underclocking), reduces stability at some point. And it just isn't the same with every card...

 

[PingSpike]

Originally posted by: Nafets
For the same reason that every card doesn't overclock to the same limits. Running at extreme limits on either end (overclocking or underclocking), reduces stability at some point. And it just isn't the same with every card...

They are all designed and built to a certain specification. If ATI doesn't know what those specifications and limits are on ATI cards, I don't know who does! They should be able to say, with relative certainty, what virtually all cards can safely run at.

You speak of the extreme limits...but remember, the cards aren't shipped with extreme limits. They're shipped clocked at a speed ATI determined would work the cards design with an acceptable defect rate. The same system would be applied to the lower limit as the upper. In short, they aren't going to clock the cards at the lowest speed your system can safely run at at the lowest voltage...they're going to clock at a speed they've determined say 99% of the cards can safely run at. At least, thats how I would do it. Obviously the extreme limit of power savings is going to have to yield to stability.