Thanks for the review. I was wondering if the 5570's fan can be adjusted via Catalyst Control Panel to a low rpm and provide a better dB/max temp point...

Low profile card + S1 cooler will likely interfere with mobo components

nice review mike. The 5450 sounds like a perfect replacement for the aging x1250 intergreted graphics in my htpc.

Is it possible that the 5450's low power draw could be less then the internal graphics on my Asus M2A-VM? Or would there be no offset to the internal graphics being disabled?

My HTPC uses a 90 Watt PicoPSU with an athlon x2 2.0ghz at 1.05 volts. (65nm). And laptop hardrive with 2 gigs of ram. So i dont think that is too much for a Pico, but still dont care to go over budget.

Wow, SPCR got a 55x0 series card before Anand or Tom's. Looks like SPCR is really moving up in the world. Well deserved, Mike! Keep up the good work!

EDIT: Haha, looks like it was just the NDA keeping the others from publishing....

I highly doubt 7w max load is going to put you over capacity on your Pico-90, as long as your power brick is rated 90+ watts. You might invest in a Kill-a-watt or similar device to make sure.

The options were greyed out. Most of the reference cards we receive from ATI can't be tweaked via CCC.

I'm rather disappointed in the 5xxx series so far, specifically in the power consumption advancements I thought I'd see compared to the 4xxx cards.

While the higher end and midrange 5xxx cards do achieve some remarkable idle numbers, overall it seems like they still run hotter and more power-hungry than the previous generation.

The new 5450 card -- it should have an edge on the other lower-end cards from the previous generation in terms of power and heat, and maybe even have a bit better performance. Instead, my 4550 is its equal or better. Mine idles at 6.6W (a few tenths of a watt less than the 5450, actually), and tops out at 7W.

I may just have a particularly good sample, but I would expect that it should be bettered in power usage by the next generation lower-end part. Maybe if I played a real demanding 3D game I'd see higher wattage from the 4550, closer to 10 or 15W. That's something I like about the 5450 -- it barely budges under load.

Still, I don't plan on upgrading to a 5xxx card any time soon. Not unless I see equal power and heat with better performance. Maybe I shouldn't get my expectations up so much every time a new generation of cards is released.

Since SPCR linked to Anandtech's review of the 5450, I am assuming that they are aware of the 5450's shortcomings in deinterlacing video. As ATI (unlike nvidia) uses its stream processors to decode video, the 80-shader 5450 can't keep up with the full suite of options in the Catalyst Control Center.

This should be brought up in any discussion of the 5450 as a HTPC card (in fact, its only real niche).

Yet there is no mention of it here?

For decoding video on the graphics card (DXVA), I think it is either on or off.

For deinterlacing, I think the 5450 might have enough shaders, but that deinterlacing test video that anand used is just over the top and unrealisticly difficult. It may work fine on normal video with vector adaptive activated all the time. It just switches to a lesser deinterlacer when the going gets tough. That's my understanding of it anyways.

Anyone have any idea what my 7300GT's idle power consumption would be vs the HD5450? Wondering if I can get a slight graphics upgrade and less power draw and heat into the bargain

There should be a disclaimer about the 4670 idle power draw. Numerous other sites got higher idle numbers than SPCR on that card.

For convenience instead of linking to a bunch of reviews I'll just link to TTAGPR.

For those not wanting to decipher insanely large charts I'll distill a few numbers into text:

Idle power 3D
5770 18W 78W
5750 13W 54W
5670 13W 51W
5570 10W 43W
5450 7W 15W

4770 26W 58W
4750 29W 55W
4670 9W 48W
4650 9W 30W
4550 7W 18W
4350 4W 15W


SPCR shows the 4670 vs 5570 idle comparison as 3W vs 8W but TTAGPR shows that comparison as 9W vs 10W.

Blame that on measurement error, sample variance (manufacturing variance), non reference vs reference cards, or any other factor you like but someone just buying a random 4670 video card is unlikely to see the idle power draw you are quoting here.

7300GT 13W 29W
HD5450 7W 15W

So yes you'd save power/heat by upgrading.

Thanks for posting those power consumption numbers. They seem to be more accurate -- or at least they jibe better with what I've seen on other sites. The 1 W difference under idle versus load for the HD 5450 just didn't seem right at all. SPCR might need to look into their testing methods a little more closely.

Minor typo on page 7:


Informative review though.

Still Not the Perfect HTPC Card according to Anand Tech: http://www.anandtech.com/video/showdoc.aspx?i=3738&p=3

Little NDA date confusion at SPCR?

Date? Nah, just NDA timezone confusion.

On topic: bring on the stock passive 5570s! In fact, bring on the stock passive 5670s too.

Man, yesterday was like a Matrix deja-vu moment.

"So that's what they changed!!"

Actually I trust SPCRs testing methods more than other sites. If they make a mistake they always own up to it and learn from it improving their reviews year after year.

Something was different about the 4670 they tested vs the 4670 most of us could buy. That isn't an issue with how they test just an issue of what they tested.

As to the 5450 I wonder if they ran into the ATI furmark coverup again. See http://www.geeks3d.com/20080826/amdati- ... ark-proof/ for just one past example. (for the record Nvidia has done this in their drivers as well).

It would be interesting to know if ATI Catalyst 8.69 RC3 limits power draw on a 5450 with furmark.exe but I'm ok either way the power draw is so low even on other 3D loads it doesn't matter to me if it is 7W or 15W in 3D apps.

I generally trust SPCR's power measurements more than other sites too. I'm not sure some of the gaming-oriented ones even know the difference between AC and DC.

In this case though, something is definitely up. Most sites don't run CPUBurn at the same time as FurMark, so I wonder if that has something to do with it. Of course, how much these things get throttled under FurMark is an unknown too.

Anyway, the difference between the rated 19.1 W load and 7 W load that SPCR got is potentially big for those us using smaller power supplies.

ht4u.net does actual direct measurements of video card power usage (using a PCIe extender card). Here's typical charts of idle and load power usage, including the 5450 & 5670: http://ht4u.net/reviews/2010/amd_radeon ... ndex13.php

You'll be able to figure it out without knowing German.

Wow, that ht4u link was nice. According to their numbers, I could get a whopping 25W idle reduction in DC, if I replaced my Nvidia 8800GTS 512MB by an ATI 5870 and while my system uses around 110W AC at idle now and my Corsair VX450 being about 80% efficient at that load, would reduce my idle by 31W AC to under 80W.

This would already make my 1m^2 walk-in closet containing the computer much cooler. At the moment it gets a bit warmer in there than would be good for the system in the long term.

And of course I would not get anything just for reduction in power, but I can't play GTA 4 on my system without making it look like lego blocks.

I guess I'm still waiting for the new Nvidia releases just to get a 5870 cheaper.

at the wall my UPS stated its consuming ~55 watts idle and 75 watts when spinning up a DVD and playing a video. So actual DC consumption should be much lower. I just have not cared to run a Pico at its peak capability. Also i take those numbers with a grain of salt.

The idle power draw for 4670 is just wrong. Noone else in the world could achive that!
SPCR schould measure again - as this review draws conclusions from it. We thrust SPCR, but not with this measurement.

Comparison between 'real' 4670 and 5670 would be nice.

Anandtech has a follow-up blog that notes the 5570 with an updated driver has enough power for all of the 5xxx series video processing. The 5450 has enough power for deinterlacing at 1920x1080 (although not enough for other post-processing effects).
http://anandtech.com/weblog/showpost.aspx?i=669