SPCR

Are you talking with the manufacturer or designer of the card about the high idle consumption? Honestly 50W idle is just plain ridiculous and I'm not going to touch it with a stick. It's almost an insult to nature.

Nobody really knows what's going on.

There's a theory that the 38xx and 48xx series were developed by two different teams at the same time, instead of the 48xx being a progression over the 38xx.

Time will tell if it's possible to lower idle power consumption. Modded BIOS files to downclock the card in 2D don't seem to help much, and it's doubtful that a voltage reduction would bring the 4850 to 38xx levels.

I suspect (and hope) that the idle power consumption will drop a fair bit with the next Catalyst drivers, or another hotfix.

Its pretty lame to have a modern GPU that uses half of its max power whilst doing nothing, although @ 100W max usage I am quite happy as it shouldn't need a cooling fan on the S1.


Andy

I'm thinking I'd quite like one of these cards but I don't "need" one nor do I play that many games that would benefit from the performance. (none actually!) But then need isn't what it's about!
I find the power consumption figures to be a little concerning; then again my X1950pro will increase the AC usage by 50w when running ATI Tool so it's not so far off. I’m definitely going to hold off for a while and see how the idle power consumption pans out. The Accelero won't fit my case, I'd be looking at Zalman VF-1000, good idea? I’ve fan-swapped VF-900 so I can do that to it if needed.
Final thought: 100w DC usage for the video card? I built an E8400 (Core 2 Duo "wolfdale" 3GHz) system with HD3450 graphics that ran SMP Folding@Home at 100w AC, ~80DC!
Regards, Seb

Matija wrote:Nobody really knows what's going on.

There's a theory that the 38xx and 48xx series were developed by two different teams at the same time, instead of the 48xx being a progression over the 38xx.

Time will tell if it's possible to lower idle power consumption. Modded BIOS files to downclock the card in 2D don't seem to help much, and it's doubtful that a voltage reduction would bring the 4850 to 38xx levels.

The team behind R7XX is the one behind XBOX's Xeno GPU. Hence the architecture is quite different.

As for the heat there is an issue with the GPU not down clocking during idle, if you search around many have modded the VGA BIOS to down clock and slightly bump the fan speed for a huge decrease in temps. Search around for Readon Bios Editor (RBE) on Xtreme Systems and the like.

Just wanted to say thanks for doing such a thorough review of the heat, power and noise of this card. I know it's common practice around here, but is uncommon considering what I've seen in other sites' 4850 reviews. It seems that everyone was in such a rush to get a review out, that they just about skipped the entire heat, power and noise portion of their articles. *coughanandcough*

Hypernova wrote:As for the heat there is an issue with the GPU not down clocking during idle, if you search around many have modded the VGA BIOS to down clock and slightly bump the fan speed for a huge decrease in temps. Search around for Readon Bios Editor (RBE) on Xtreme Systems and the like.

That's the thing - downclocking the GPU doesn't seem to help much, we're talking about a difference of 2-3 degrees in idle, which hints to a couple of watts.

Umm... am I reading those GPU-Z shots correctly, that the GPU clock at idle was 500MHz, memory at 750MHz? That's a fair way out of line with the modified BIOS settings (160/500) I've seen so there's certainly something screwy going on.

Lawrence Lee wrote:it should have been equipped with a dual slot cooler

Are you sure about that? Going by the load temperatures, the cooler is perfectly adequate (if horribly loud). It seems to me that the excessive 2D power consumption is the reason for the high idle temps. It looks like the fan speed profile has been designed for a much lower thermal output at idle than is currently the case, which gives further credence to the suggestion that the issue will be resolved by the next Catalyst update.

Even if the current cooler wasn't enough, I'm sure the problem could be fixed simply by using a bigger single slot cooler. Taking up an extra slot seems excessive.

Hi,

I gotta' say that the skull image is bad enough to make me not want to buy this card. I cringe when I look at it...

You're right Neil it is seriously cringeworthy. But in reality it's not as bad as the usual 3D-rendered-scantily-clad-faux-medieval-fembot-in-miniskirt-showing-sideboob-clutching-giant-sword-to-cleavage teenage wank fantasy we're normally subjected to.

I'm pretty sure that Powerplay simply isn't working.

But in reality it's not as bad as the usual 3D-rendered-scantily-clad-faux-medieval-fembot-in-miniskirt-showing-sideboob-clutching-giant-sword-to-cleavage teenage wank fantasy we're normally subjected to.

Andy

Test Platform.....
...* ATI Catalyst 8.5 beta graphics driver

Anybody felt a modification in power consumption at idle with the leaked Catalyst 8.7 beta? Or registering lower temperatures?

currently you can use HD 48x0 fan speed by creating a Catalyst Control Center profile after activating overdrive. After creating the profile you can change that target speed's and creating better temperatures.

Making the adjustment you need to open .XML file in Ati's ACE folder( courtesy of bretware from Guru 3D, thanks ). You need Excel or equal program to modify it:

if you make a profile in catalyst after turning on Overdrive( make sure clock and memory settings are correct). You can then go to the file "C:\Users\Bretware\AppData\Local\ATI\ACE\Bretware. XML. the xml file will have the same name as the catalyst profile you saved. right click and hit edit.

Code: Select all

<Feature name="FanSpeedAlgorithm_0">
<Property name="FanSpeedAlgorithm" value="Manual" />
</Feature>
<Feature name="FanSpeedRPMTarget_0">
<Property name="Want" value="0" />
</Feature>
<Feature name="FanSpeedPercentTarget_0">
<Property name="Want" value="65" />

Change the value to "manual" and Want value to "65"

Save the file then reload your profile in catalyst.

just remember you need to see hidden files and folders to see this. Profile can also be in: C:\Documents and Settinds\UserNameHere [or All users]\Application Data\ATI\ACE

rpsgc wrote:I'm pretty sure that Powerplay simply isn't working.

Maybe it's not even there.

If high idle power consumption was a software issue, it would have been fixed already.

If it's a BIOS issue, something would have leaked. Apart from a MSI BIOS that does lower the clocks in 2D, there's nothing.

There's something missing in ALL the reviews I read on the HD48xx series, and actually in ALL reviews of gfx cards.
How much power does the GPU itself use and how much the rest of the components (memory, VRMs, etc)?
From the 101W (SPCR figure) or 110W (Xbitlabs figure, that I'll use in the following lines) for the 4850, how much does the GPU draw?
My guess is 80W for the GPU and 30W for the rest. But some people in the forums believe that it's more for the rest, because undervolting the GPU didn't provide much benefit, while undervolting the memory lowered power consumption more.
I'm interested in the following: to get a 4870, undervolt and underclock it to 4850 levels, and keep the stock cooler of the 4870 at ~50% of stock rpms (not 50% of max rpm). Since for me the performance of the 4850 is good enough, I'm willing to pay more for the 4870, and the 4870's cooler exhausts heat, it seems a good deal. I might reseat the stock 4870 cooler and add a HR-11 to prevent it form ramping up too much. Your thoughts about this?

blackworx wrote:Going by the load temperatures, the cooler is perfectly adequate (if horribly loud).

Longest oxymoron I've seen so far.

Modo wrote:

blackworx wrote:Going by the load temperatures, the cooler is perfectly adequate (if horribly loud).

Longest oxymoron I've seen so far.

In which case it would be a contradiction in terms, not an oxymoron

I know what you mean though. Adequate in this sense doesn't necessarily mean "acceptable" - certainly not around these parts anyway

I think I'll skip this round for the 40nm refresh. Gaming performance seems great, but I am happy with my 3870 at the moment.

I believe the voltage of the card is not being altered when idle, which results in the offensively high idle power consumption. I think a driver update, or at least a driver and BIOS update should be able to fix the issue, as a driver update was able to remove the stuttering I experienced with my 3870 in early driver sets. This was a result of Power Play changing modes when the card was under load.

Great review. Mike, I wonder if you considered testing the card with an aftermarket cooler? S1? HR-03GT? Passive testing would be great. Did you also consider attempting the BIOS mod?

one last question. If they do release an updated driver that "fixes" powerplay, will you retest? I'd be very interested to see the power figures.

My 4850 arrived today, and rather than waste time testing out the stock cooler i threw my accelero straight on (with some new RAM heatsinks) and idle temps around the high 30's, with a 500rpm slipstream on. The stock cooler really isn't up to much, it does just enough to cool the card at stock speeds.
SPCRists can indeed still get fantastic performance without adding any noise. Horray!

ryboto wrote:Great review. Mike, I wonder if you considered testing the card with an aftermarket cooler? S1? HR-03GT? Passive testing would be great. Did you also consider attempting the BIOS mod?

one last question. If they do release an updated driver that "fixes" powerplay, will you retest? I'd be very interested to see the power figures.

I guess I'm Mike...

The card was actually purchased for my own personal use - the review was meant to be a one-off, since it's now sitting in my main rig at home. As such, I'm not going to risk completely killing it with a BIOS mod. If the power issue gets fixed with drivers I can share the before/after power figures as I do have an UPM power meter at home. We can extrapolate/guestimate what the difference would have been on the test rig in the lab.

I ended up putting a S1 on it (passive) and the temps are much improved. 50C idle and 65C load, IRC.

I think I'm going the same route as Lawrence and Bob - I just got the Accelero S1 a few minutes ago and the card earlier today. Wish me luck!

Lawrence Lee wrote: I guess I'm Mike...

The card was actually purchased for my own personal use - the review was meant to be a one-off, since it's now sitting in my main rig at home. As such, I'm not going to risk completely killing it with a BIOS mod. If the power issue gets fixed with drivers I can share the before/after power figures as I do have an UPM power meter at home. We can extrapolate/guestimate what the difference would have been on the test rig in the lab.

I ended up putting a S1 on it (passive) and the temps are much improved. 50C idle and 65C load, IRC.

sorry, I just saw that Mike made the thread, silly me. Still, great review. Good to know the S1 can work passively with the card. Can you give us an idea of the airflow in the case when it was used passively?

Great stuff!

blackworx wrote:You're right Neil it is seriously cringeworthy. But in reality it's not as bad as the usual 3D-rendered-scantily-clad-faux-medieval-fembot-in-miniskirt-showing-sideboob-clutching-giant-sword-to-cleavage teenage wank fantasy we're normally subjected to.

ryboto wrote: Good to know the S1 can work passively with the card. Can you give us an idea of the airflow in the case when it was used passively?

Antec Solo, Seasonic S12II 500W, Yate Loon D12SL-12 @ 5V, Scythe 100mm @ 5V, one hard drive.

I wondered why the 4850 seems to be better cooled by the S1 than G92 nVidia products, despite it's relatively high power usage.
My guess is that the larger die area of the 4850 allows for better transfer of the heat through all 4 heatpipes.
Or, to put it in other terms, the heat / surface density is smaller than that of the G92 cards, making the 4850 GPU easier to cool.

I don't have any stick-on heatsinks on hand for the voltage regulation chips - will a quick trip to Microcenter do the trick?