Power consumption differences between different brands of the same GPU are a big problem - it might be that there is a much less consuming 5750 from another brand and that the 5850 reviewed was on the low side.

The cooler design makes baby seals cry. A formula, seriously, Asus?

The power usage, noise and temps of the 5750 cards can vary quite a bit depending on the maker. ht4u.net does very through test measurements (their power figures are very close to SPCR's). There are results for several 5750 cards in thier review of this Asus card:

http://ht4u.net/reviews/2009/asus_eah5750_formula/

The data charts are self explanatory enough to not require translations.

lol indeed. It uses 10w more at idle. Around 12w more under load. It looks stupid. What were they thinking?? It's got to be the worst graphics card I've seen this year.

http://ht4u.net/reviews/2009/asus_eah57 ... ndex11.php is the page with power numbers

10 watts lower at idle for 3 brands of cards plus the reference card:


I'd assume the rest were all similar to the reference card but the Furmark numbers don't show the MSI in such a good light


The furmark numbers are different for the SPCR review versus the ht4u.net review of the same Asus Formula 5750. I have to wonder if this is the old rename furmark.exe issue again (which includes differing driver versions under the same concept) or some other difference in testing?

Either way it's clear that I will want to avoid the Asus Formula 5750 just because the idle power draw is consistently higher across multiple reviews.

Additional sources in card wattage not total system power

http://www.xbitlabs.com/articles/video/ ... 750_5.html

http://www.techpowerup.com/reviews/Sapp ... 50/28.html

Basically of all the URLs I went to looking at 5750 power draw the only sites that did the math/work necessary to give card power not system power were:

xbitlabs.com
silentpcreview.com
ht4u.net
techpowerup.com

and ht4u.net was the most interesting source since they tested 5 different 5750 cards.

I've always been impressed with the thorough way that SPCR performs its tests compared to many other sites, but here I have a small problem with the test methodology:

I don't think it's valid to say that just because the CPU was fully stressed with CPUburn, it will consume the same amount of power when adding the FurMark. When I do a similar test, running LinX for CPU load (which is astonishing at heating up the CPU) and then adding RTHDRIBL (which is pretty astonishing for heating up the GPU, too) for graphics load, the CPU temperature drops noticeably. This means that the graphics benchmark puts less load on the CPU than just the CPU load by itself, and means the CPU will use less power.

Given how modern CPUs are good at saving power on unused subsystems, changing the nature of the load ie from floating-point heavy or just to a less cache-efficient code can have pretty large effects on the power draw even if the CPU is always nominally "fully loaded".

I wonder if it's possible to run with the integrated graphics and different programs to load the CPU and infer the relationship between CPU temp (disabling any automated fan control, of course) and power draw. Then you could by looking at the CPU temp during the benchmark infer how much power it's using.

Given the rapid increase in noise from the fan at speeds higher than you saw in your test, it also seems important to know that you really are stressing the GPU to the max. Otherwise someone might power up some particular game that does a very good job of maxing it out, and thus encounter noise levels much higher than you experienced.

cheers,

/Patrik

100% load isn't always a good indication of how much work a CPU is doing. If all the data and code is not in L1 cache, then there will be extra wait state clock cycles to retrieve data / code. Also if two programs are being run (depending on the OS and # cores in the CPU) such as a CPU and GPU load programs, then there may be task switching between the two programs which will also throw in wait states.

It seems that ATI has managed to get underclocking working on GDDR5 reliably. It's a shame they can't update the BIOS/drivers for the 48x0 cards which use it too.

It's also a shame you can't do crossfire with 4xx0 series cards as that would provide a nice upgrade path. It would also be handy for people like me who don't need massive graphics power most of the time because we could use the 5750 as the main card and just pop the older but still powerful 4xx0 card in when we need the performance boost. No need to even change the monitor cable over.

As it is I am considering just using Intel's on-board video with a Radeon for when I need performance. It's not an ideal solution as you need to have both Intel and ATI graphics drivers installed, and they don't play together well if you want to use the ATI card just for processing (e.g. Folding@Home).

Maybe I should get a new CPU as well, move to an ATI mobo and chipset with on-board Radeon. Seems like a waste of a good Core 2 Quad though.

Thanks for the review, excellent as always.

Marketing ppl, always working hard to bring us yet another innovative block of aluminum 0_o

So the card has the same layout and size of mounting holes, meaning the preferred AC S1 will fit fine, with just a little bit of trim. Just like on a 4770. Good.

MoJo: The high end of the Raden 5000 series have both low idle power and good performance. Maybe you could just sell your 4000 series Radeon and buy a 5850?

I wouldn't buy this card just for its design. Formula?

Passive or Vapor-X everyday.

I think that is what I will probably do.

The only issue is that if you only have one card and want to use it as a co-processor it makes the system unusable in the mean time. Screen updates are so slow you can't even browse the web. Thus having a second card is helpful.

Even if you can't use Crossfire with different generation cards I think it should be okay to have them both in the system at once.

What I'd most like to see in a 5750 review would be the stock ATI 5750 vs the power color fanless 5750 vs the Go Green 5750.

The Go Green version claims 20% less power usage than the reference version.

The newegg comparison for the price of the three 5750 powercolor cards is

$130 with fan
$150 fanless
$170 Go Green

the cheapest reference 5750 is XFX at $140. The $130 powercolor 5750 has an accelero heatsink with fan and has one less DVI connector than the reference card.

The "Go Green" version is probably just a hand-picked card which happens to work at a lower voltage than most.

I'd be wary of cards like that. How well have PowerColor tested the card? Can you imagine the fun you would have trying to a) diagnose a card which is failing at idle and b) trying to explain that and return it to a shop that will most likely install it and immediately run Furmark to stress test it?

You could always get a standard card and then see how much undervolting you can do yourself. At least that way if you run into problems you can be reasonably sure it will be fine at stock speeds, and still be a pretty efficient card.

It's passive but it's not an Accelero S1, the passive 5750 by Club3D uses an Accelero instead.

Edit: I've seen you mean the Accelero L2 with fan, not the passive S1

No, if you go to newegg you can see the cards are different in at least some caps, a heatsink, and the power connector.

I'm sure they do cherry pick the GPU so they can undervolt it some but the card has visible differences.

If you want to compare the pics it is the 3rd pic in each gallery for

scs3 and
green

I had problems making the second URL collapse. Had to remove "&Description=POWERCOLOR Go! Green AX5750 1GBD5-NS3DH Radeon HD 5750 1GB 128-bit GDDR5 PCI Express 2.1 x16 HDCP Ready CrossFireX Support Video Card" from the URL as SPCRs BB didn't like the length or one of the characters in the full URL.

The pics aren't from the same angle but if you compare them I think you will notice more than one change on the right hand side of the card between the 3rd and 4th heatpipe.

The 4th picture on each gallery shows the other side where if you zoom in it is clear a different PCB is used on each card.

One is version 1.0 and the other is version 1.1 so it is possible that you may be able to get the 1.1 revision in the non green card or it is possible that the 1.0 revision is for the non green and the 1.1 is the green. Just depends on if your interpretation or mine is closer to the truth.

I don't know for a fact that the green card is a different layout of the most current non green. I do know that there are two different layouts and that at least leaves the possibility that one is specifically for the green product.

Confirmation that the Go Green version is different in http://www.silentpcreview.com/forums/vi ... hp?t=56956

Thanks to Audiodude and the RBE guy(s) for the research into this.