Filling the Gap: ATI HD 4830

Post by **MikeC** » Mon Oct 27, 2008 6:47 am

FartingBob · Post by **FartingBob** » Mon Oct 27, 2008 7:08 am

Very smart move by ATI, they now have nvidia beat at all price points with their 4000 series. While nvidia got complacent being at the top for so long with their 8000 series and releasing far to many cards that are very similar to eachother, relabelling cards and confusing naming schemes that change every few months. Meanwhile ATI built 1 exceptional core that is cheaper to make and very efficient and have used that in all their cards. Nvidia have been forced into slashing prices of older high end cards to compete with ATI mid ranged ones.

cdamian · Post by **cdamian** » Mon Oct 27, 2008 7:34 am

Are there any 4xxx cards without a fan ?

I will build a new computer in November and I rather have a graphics card without fan. I am currently looking at the Sapphire Radeon HD3850 Ultimate PCIe.

I am not a big gamer, so I probably won't need the power. And I prefer a ATI card, because I am exclusively running Linux and ATI has shown a better attitude to open source.

rpsgc · Post by **rpsgc** » Mon Oct 27, 2008 7:44 am

cdamian wrote:Are there any 4xxx cards without a fan ?

I will build a new computer in November and I rather have a graphics card without fan. I am currently looking at the Sapphire Radeon HD3850 Ultimate PCIe.

I am not a big gamer, so I probably won't need the power. And I prefer a ATI card, because I am exclusively running Linux and ATI has shown a better attitude to open source.

Probably not the best place to ask that, a bit off-topic

Powercolor (not released yet)
Gigabyte (available)

MikeC, from reading the article I get the idea that you didn't verify if this card has the correct BIOS or not. So I would assume it has the wrong one. I don't know whether you're aware of the problem or not.

I think having 80SP disabled would affect its power consumption, no? So a retail card with its full 640SP could consume more power than the "defective" 580SP review samples.
Techpowerup's review certainly shows that, as the full 640SP card from Powercolor consumes more power than the AMD card with only 580SP.

92W @ idle for the AMD card vs 122W @ idle (almost same as HD4850) for the Powercolor and 190W @ load for the AMD card vs 208W for the Powercolor card.

It's too bad though... the performance difference is rather small (<15%) but the power consumption difference is much larger (~30% @ idle & ~10% @ load). Maybe someone will host the "defective" BIOS for those who don't mind the performance difference and/or value more the power consumption reduction, eh?

autoboy · Post by **autoboy** » Mon Oct 27, 2008 8:00 am

If you are really concerned with idle power consumption and just game once in awhile I would stick to the 4670 with its 3W idle power consumption. Amazing.

Does this card use the exact same heatsink design as the 4850?

dhanson865 · Post by **dhanson865** » Mon Oct 27, 2008 9:17 am

I'd really like to know if you tested with 560 or 640 and if you tested at 560 I'd like to know the difference in idle power consumption after the BIOS update to enable the extra circuits.

To me the best options for a moderate gamer are a 4670 or a 4830.

the 4670 is the best of the low idle draw cards currently available. If you go by the TTAGPR2 you have

HD4850 ______________#####Â¦)#Â¦]#####}| 41W-61W-110W
HD4830 ______________??????????????????????
HD4670 ______________#Â¦)######}| 10W-70W %

I'll take the 4670 at 10W versus the 4850 at 41W any day. My system is at idle graphics wise something like 90% or more of the time it is on.

If the 4830 can come in between those two numbers, say 20W idle then I'd be willing to pay a little extra in idle wattage for the extra performance the 4830 gives me when I do game.

Give me a 4830 with sinks on the VRM/RAM and an accelero S1 premounted at the factory and I'm all over that. Add a BIOS optimized for low power draw at idle and I'll brag about it.

So Lawrence, tell me was the 18 watts idle number you measured with a bios that enabled 640? If the card you have is updated to run 640 does it up the power consumption?

rpsgc · Post by **rpsgc** » Mon Oct 27, 2008 9:28 am

dhanson865, the Powercolor HD4830 has sinks on the VRM and RAM modules.

http://www.techpowerup.com/reviews/Powe ... /card1.jpg
http://www.techpowerup.com/reviews/Powe ... /front.jpg

Lawrence Lee · Post by **Lawrence Lee** » Mon Oct 27, 2008 9:48 am

dhanson865 wrote: So Lawrence, tell me was the 18 watts idle number you measured with a bios that enabled 640? If the card you have is updated to run 640 does it up the power consumption?

I wasn't aware of the problem until just recently. The beta version of GPU-Z we used reported it as 800, so I can't be sure if it was 640 or 560. Judging by the power consumption levels reported by TechPowerUp, I would have to guess that it was 560.

QuietOC · Post by **QuietOC** » Mon Oct 27, 2008 9:49 am

dhanson865 wrote:So Lawrence, tell me was the 18 watts idle number you measured with a bios that enabled 640? If the card you have is updated to run 640 does it up the power consumption?

TechReport said the power difference when enabling the extra cores was 1W.

rpsgc · Post by **rpsgc** » Mon Oct 27, 2008 9:53 am

QuietOC wrote:TechReport said the power difference when enabling the extra cores was 1W.

Techpowerup disagrees. We will never know.

zoatebix · Post by **zoatebix** » Mon Oct 27, 2008 11:09 am

rpsgc wrote:Maybe someone will host the "defective" BIOS for those who don't mind the performance difference and/or value more the power consumption reduction, eh?

I almost posted the same question in response to the review on Anandtech. I expect there to be more interest here.

Tim Connor · Post by **Tim Connor** » Mon Oct 27, 2008 11:20 am

Funny enough that this review came out right after I became familiar with the fan controller and the new Catalyst Control Center.

At least on my 4870 with the absolute latest version of of the CCC you now have the option to set a manual fixed fan level. I too found the change in RPM to be quite annoying. While the easiest solution it still leaves the fan louder than it needs to be at idle.

I chose to change the cooler and modify the BIOS to adjust the fan curve. I'd assume that you can adjust the curve on the 4830 in the same way while keeping the stock cooler. On HTPC applications you wouldn't need nearly as aggressive fan curve.

(More info here BTW - viewtopic.php?t=50678)

Regards, Tim

dhanson865 · Post by **dhanson865** » Mon Oct 27, 2008 11:27 am

rpsgc wrote:dhanson865, the Powercolor HD4830 has sinks on the VRM and RAM modules.

http://www.techpowerup.com/reviews/Powe ... /card1.jpg
http://www.techpowerup.com/reviews/Powe ... /front.jpg

The Powercolor version of the card is also said to draw considerably more power at idle than the stock version of the card.

Is that because of the difference in design of PCB, heatsinks, BIOS, or all of the above?

http://www.techpowerup.com/reviews/Powe ... 30/24.html

30W more at idle does not impress me.

rpsgc · Post by **rpsgc** » Mon Oct 27, 2008 11:31 am

dhanson865 wrote:
rpsgc wrote:dhanson865, the Powercolor HD4830 has sinks on the VRM and RAM modules.

http://www.techpowerup.com/reviews/Powe ... /card1.jpg
http://www.techpowerup.com/reviews/Powe ... /front.jpg
The Powercolor version of the card is also said to draw considerably more power at idle than the stock version of the card.

Is that because of the difference in design of PCB, heatsinks, BIOS, or all of the above?

http://www.techpowerup.com/reviews/Powe ... 30/24.html

30W more at idle does not impress me.

I've already discussed that
viewtopic.php?p=436288#436288

AMD card had 580SP, Powercolor card had 640SP.
Until SPCR can test the HD4830 with the correct BIOS I'll assume the difference is because of the disabled SPs.

QuietOC · Post by **QuietOC** » Mon Oct 27, 2008 11:42 am

rpsgc wrote:AMD card had 580SP, Powercolor card had 640SP.
Until SPCR can test the HD4830 with the correct BIOS I'll assume the difference is because of the disabled SPs.

That is completely illogical. Power use is proportional to frequency cubed using minimum stable voltages. The idle power differences must be due mainly to reduced voltages and secondarily to reduced clockspeeds. Yes, video card BIOS can control both of those and the number of active SPs--actually, has no one else thought it might be possible to turn these into full 4850s with just a BIOS flash?

Deactivating parts of a chip generally isn't benefitial to power/performance. My 9600 GSO may have 85% the performance of a 9800GT, but it uses more than 85% of the power. Oh well, at least it was cheap!

Monkeh16 · Post by **Monkeh16** » Mon Oct 27, 2008 11:54 am

QuietOC wrote:
rpsgc wrote:AMD card had 580SP, Powercolor card had 640SP.
Until SPCR can test the HD4830 with the correct BIOS I'll assume the difference is because of the disabled SPs.
That is completely illogical. The idle power differences must be due mainly to reduced voltages and secondarily to reduced clockspeeds. Yes, video card BIOS controls both of those and the number of active SPs--actually, has no one else thought it might be possible to turn these into full 4850s with just a BIOS flash?

Except they're probably 4830s because not all of the SPs are functional.

QuietOC · Post by **QuietOC** » Mon Oct 27, 2008 12:00 pm

Monkeh16 wrote:Except they're probably 4830s because not all of the SPs are functional.

Well, I have owned more than a couple of video cards sold like this that were fully functional. My 9600GSO is physically lacking a couple memory chips however.

dhanson865 · Post by **dhanson865** » Mon Oct 27, 2008 12:21 pm

rpsgc wrote:
dhanson865 wrote:
rpsgc wrote:dhanson865, the Powercolor HD4830 has sinks on the VRM and RAM modules.

http://www.techpowerup.com/reviews/Powe ... /card1.jpg
http://www.techpowerup.com/reviews/Powe ... /front.jpg
The Powercolor version of the card is also said to draw considerably more power at idle than the stock version of the card.

Is that because of the difference in design of PCB, heatsinks, BIOS, or all of the above?

http://www.techpowerup.com/reviews/Powe ... 30/24.html

30W more at idle does not impress me.
I've already discussed that
viewtopic.php?p=436288#436288

AMD card had 580SP, Powercolor card had 640SP.
Until SPCR can test the HD4830 with the correct BIOS I'll assume the difference is because of the disabled SPs.

And I read your prior message.

1. I don't trust the methodology of every random website

2. Tech power up got my respect by bringing the issue to the forefront but the text of their discussion of the issue focuses on performance not power draw. I have no reason to believe they did a thorough reexaminiation of the cards in relation to power draw after they changed the BIOS. Call be a skeptic if you will.

3. I am asking for a additional sources to test the issue. If I can see power tests from 3 trusted sources instead of 1 I'd feel better about it.

4. Even if you are right and the 640 version draws 30W more at idle than the 560 version then I don't want the version that draws 30W more.

I find it hard to believe a change that only affects performance by 3 to 5% would affect idle power draw by 300%. I have to assume there are other factors at play.

Look at it this way
Card SP Watts
4670 320 10W
4830 560 18W
4830 640 38W?

Can't you see how out of whack that curve is? If the SP count was the primary difference in wattage the 320 to 10W would be equivalent to 640 to 20W. It's a simple progression. Why is the powercolor 4830 not following the progression?

dhanson865 · Post by **dhanson865** » Mon Oct 27, 2008 12:34 pm

Lawrence Lee wrote:
dhanson865 wrote: So Lawrence, tell me was the 18 watts idle number you measured with a bios that enabled 640? If the card you have is updated to run 640 does it up the power consumption?
I wasn't aware of the problem until just recently. The beta version of GPU-Z we used reported it as 800, so I can't be sure if it was 640 or 560. Judging by the power consumption levels reported by TechPowerUp, I would have to guess that it was 560.

Do you still have the card? Can you flash the bios and retest?

If you still have the card and plan to retest it'd be nice to see power draw retested before and after flashing the bios with a clear addendum to the review added to address the follow up testing.

croddie · Post by **croddie** » Mon Oct 27, 2008 1:34 pm

Hello. I think this website should focus on silent computers since it is called "silent pc review". It should leave gaming cards to other sites or change its name. On the front page of the site there are three reviews of power-hungry graphics cards with fans.

rpsgc · Post by **rpsgc** » Mon Oct 27, 2008 1:54 pm

croddie wrote:Hello. I think this website should focus on silent computers since it is called "silent pc review". It should leave gaming cards to other sites or change its name. On the front page of the site there are three reviews of power-hungry graphics cards with fans.

Orly? Plenty of people want POWER and SILENCE. Or we can't have both? And there are aftermarket coolers for a reason. So I don't really know why you're complaining.

FartingBob · Post by **FartingBob** » Mon Oct 27, 2008 2:29 pm

croddie wrote:Hello. I think this website should focus on silent computers since it is called "silent pc review". It should leave gaming cards to other sites or change its name. On the front page of the site there are three reviews of power-hungry graphics cards with fans.

We are in a good time when its possible to have QUIET gaming rigs, which can be made near silent with aftermarket coolers and fans.
I for one have a ATI 4850 with a passive accelero S1 heatsink and a Q9400 CPU with a scythe ninja heatsink and a 500rpm slipstream. Coupled with a WD 640GB HDD and an Enermax modu82 PSU.
I cant hear the thing when im watching video on it or gaming. This takes some work to setup (installing the heatsinks, slowing down the fans) but now i have a system that can run the latest games at max or near max settings on 1680x1050 res and I CANT HEAR IT.

So yea, im an example of someone who would be interested in looking at gaming parts from a silencer perspective. Very few of your typical review sites care about sound or power consumption, and even fewer have the ability to accurately test and report it.

croddie · Post by **croddie** » Mon Oct 27, 2008 3:08 pm

You have to have the right proportions. Gamers are a minority of computer uers. (I mean the people who buy computer systems to play modern 3d games.) If you are a gamer who wants silence you can look to other sites for performance and power consumption which are the two relevant factors (since any fan will be replaced with a passive heatsink). You don't need an SPCR review. For power consumption try this link: http://archive.atomicmpc.com.au/forums. ... c=7&t=9354

dhanson865 · Post by **dhanson865** » Mon Oct 27, 2008 7:23 pm

croddie wrote:http://archive.atomicmpc.com.au/forums. ... c=7&t=9354

Where do you think I got the numbers I posted above from? Did you not recognize the format of the "bar graph"?

Do you notice that TTAGPR2 doesn't show numbers for the 4830 yet? You are posting in a forum of a site that actually did a review asking people to look to a site that only culls numbers from sites that do reviews. It's silly to ask us to leave the source and go wait at the collector downstream.

Besides the numbers are useless when there is a measurement issue unless the reviewer addresses the issue. That is exactly why we are here. We are trying to get more information about the cards in question.

If you don't want to read about video cards go read about hard drives or heat sinks or anything else that you do want to read about.

rpsgc · Post by **rpsgc** » Tue Oct 28, 2008 1:55 am

Matbe's review shows the HD4830 consuming slightly less than the HD3870 at both idle and load. It's a 560SP card though.

maf718 · Post by **maf718** » Tue Oct 28, 2008 9:48 am

Re the wildly differing power consumption of the two 4830's reviewed by techpowerup: the author stated clearly that the lower consuming one was running at much lower idle clocks, 150mhz vs 500mhz iirc. It had different idle clock settings stored in the BIOS and presumably it was also running at lower voltage.

Some of the difference in load power consumption would have been down to the deactivated SP units, but also maybe because of the different PCB design of the two cards, the cheaper Powercolor design could be less efficient in its power handling.

Lawrence, there is an updated version of GPU-Z available that the author is recommending for reviewers to use to test for the 560/640 SP issue, http://www.techpowerup.com/articles/155 ... /GPU-Z.exe , if you would like it. (Hope he doesn't mind me posting the link). As you are already aware there is an updated BIOS available too.

If the spcr sample was 560 SP, I am sure many people including myself would be very interested if you could rerun the power consumption tests with 640 SP to see what the actual difference is. Hopefully you have the time and inclination as spcr's power consumption figures are the only ones many of us really trust.

croddie · Post by **croddie** » Tue Oct 28, 2008 1:36 pm

dhanson865 wrote:If you don't want to read about video cards go read about hard drives or heat sinks or anything else that you do want to read about.

Sure but I have the right to make the comment that gaming is hugely overrepresented in SPCR articles. On the front page alone: reviews of a 700W power supply and three gaming-oriented video cards.

rpsgc · Post by **rpsgc** » Tue Oct 28, 2008 1:53 pm

croddie wrote:Sure but I have the right to make the comment that gaming is hugely overrepresented in SPCR articles. On the front page alone: reviews of a 700W power supply and three gaming-oriented video cards.

I beg to differ. I for one welcome the review of these "gaming" cards as a breath of fresh air. As was said, many of us are gamers and would like to be able to game with minimum noise. You speak as if SPCR does nothing but review high end cards.

And about the PSUs, well that's not really under SPCR's control. They just review what the manufacturers send them. They don't have a choice in what they send them. I think that's pretty clear.

Post by **MikeC** » Tue Oct 28, 2008 2:25 pm

rpsgc wrote:
croddie wrote:Sure but I have the right to make the comment that gaming is hugely overrepresented in SPCR articles. On the front page alone: reviews of a 700W power supply and three gaming-oriented video cards.
I beg to differ. I for one welcome the review of these "gaming" cards as a breath of fresh air. As was said, many of us are gamers and would like to be able to game with minimum noise. You speak as if SPCR does nothing but review high end cards.

And about the PSUs, well that's not really under SPCR's control. They just review what the manufacturers send them. They don't have a choice in what they send them. I think that's pretty clear.

Just a couple of things...

1. SPCR's review roster is a bit like a mirror of the industry. We have some choice but it's hard not to accept the products companies WANT to promote hard. If we say no, then when we ask for something that's low on their priorities (but high on ours), there's less incentive for them to comply. Scratch my back...

2. We may be reviewing high power products, but rarely the highest, and we always look at them from a PoV quite different than most others. No one examines the acoustics of the products -- or even the power envelope, in many cases -- like we do. As rpsgc says, that's valuable for many gamers.

cmthomson · Post by **cmthomson** » Tue Oct 28, 2008 3:15 pm

Generally a fine review, as usual.

However, I have to contest some of the conclusions, especially that this card is somehow the best mid-priced card available.

By contrast, my recently installed Gigabyte GV-N98TZL-512H 9800GT card (see viewtopic.php?t=50572) has very similar performance, and consumes about 40W less at load according to Anandtech (at stock settings). The price difference is about $10.

There are three important differences (at least to me): the N98TZL card has an excellent quality VRM with absolutely no whine. Also, it is only 8" long vs over 9" for most recent cards including the 4830. And, I could both undervolt and overclock my card for even better power consumption and lower temperatures (my card maxes out at 65C under load with fans <20dB using an Accelero S1 rev2 heat sink).

Neither card has a good stock cooler. I know most folks consider 21dB to be "silent", but I sure don't. Also, running a GPU at 85C is okay for a while, (for instance as long as it takes for the typical review), but after a few months it will take its toll. Investing in an S1r2 and a slow fan will improve any graphics card immensely, both sonically and thermally.

Filling the Gap: ATI HD 4830

Filling the Gap: ATI HD 4830

Fan Control