(!Warning! Fudzilla!) Eeeek 2900XT may need a 750W+ PSU

fornowagain · 3 May 2007 at 23:45

manoz said:
Aaah, didn't know that.

None of the PCI-e 2.0 specs I find mention anything about wattage so I take it it'll be the same, max 225w?

Apart from the increase from the 8 pin connector. The x16 slot will be 150W, not 75W as it is now. Better still the data rate will double as well.

http://www.pcisig.com/news_room/faqs/pcie2.0_faq/

Cyber-Mav · 3 May 2007 at 23:54

im still waiting, from power consumption alone id say the r600 is around 40% quicker then G80.

so lets see if r600 puts that power to good use or does it just end up sucking up more power, generating lots of heat, slowe than G80 and generally being inefficent all round.

the wait is on..............

manoz · 4 May 2007 at 00:19

fornowagain said:
Apart from the increase from the 8 pin connector. The x16 slot will be 150W, not 75W as it is now. Better still the data rate will double as well.

http://www.pcisig.com/news_room/faqs/pcie2.0_faq/

150w!!!, I knew the data rate would increase but I thought it would be due to efficiency or something else. Thanks for the link, DLed the PDF didn't mention wattage

drunkenmaster · 4 May 2007 at 02:08

if that shot is real, it could be from a crossfire support info, but i would say fake, because its ridiculous. if you're gonna make a fake use your brain. a 8800gtx which uses, what 175W or something can run fine and overclocked on a 500W psu, so a card that, even by his reckoning, "might" use 300w's overclocked, add 125w onto a 500W psu and me personally, i find the math doesn't add up to 750W. but maybe my maths is bad, or maths works differently when your an avid Nvidia fanboy.

drunkenmaster · 4 May 2007 at 02:17

Cyber-Mav said:
im still waiting, from power consumption alone id say the r600 is around 40% quicker then G80.

so lets see if r600 puts that power to good use or does it just end up sucking up more power, generating lots of heat, slowe than G80 and generally being inefficent all round.

the wait is on..............

going by Gibbo, 8800gtx is i'm sure its 175/185w(can't remember exactly?), its not much more than a 10% power use increase. things like more power to run a wider bus, which will help sometimes, but sometimes bandwidth isn't completely needed so the extra power isn't doing anything. but then its , i dunno, somewhere between the size of earth or down to a 65nm process ¬_¬ (think the last i heard was 80nm confirmed, but it was 65nm confirmed a month ago, and 80nm before that, and 90nm before that), it will probably save some power down to process size and they could have been working hard on dropping power usage in that extra time they've had, or they might not of.

basically getting at power requirements aren't going to give a great idea on efficiency or relative speed of the car. my guess will be faster, a not very noticeable amount faster i don't think. but its very hard to judge, has it got a lot of dx10 optimized stuff chucked in that really won't help at all under dx9? does the 8800 series have the same stuff in it? even if it doesn't bench well now, will the performance gap increase in the most advanced dx10 using games coming out later this year, or maybe the gap will decrease.

but as always, DO NOT pay too much attention to 3dmark results, quite frankly its cpu limited right now, also the high end cards generally are used for high quality settings and high aa/af, different cards take aa/af performance hits differently. what if they are same speed in 3dmark normal, but with 8xaa the 2900xt takes out a 30% performance advantage? take the best coded games as judged by top settings hurting performance and actually looking good, as opposed to say stalker, with high settings and HDR and looking pretty cruddy, then add aa/af, ultra texture settings and see how it performs, even then we only get a current dx9 overview, but that is fair, because for now thats all we've got.

eracer2006 · 4 May 2007 at 02:21

Gibbo said:
Hi there

Our system had an Enermax 1000W PSU.

I have the ATI Tech/Fact sheet here so upto 200W is requirement at stock clocks with 2x6pin connectors.

If overclocking then you need 1x8pin and 1x6pin connected and then it consumes upto 225W.

I hope ATI don't bash me for stating this but I don't want misleading information such as that going around.

Fact is a decent quality 400-500W PSU will run of these cards fine if your running a Core 2 DUO system or Athlon X2 based system.

Phew, i was starting to worry there cos i only bought the ocz 650 and not 750....

fornowagain · 4 May 2007 at 03:18

drunkenmaster said:
going by Gibbo, 8800gtx is i'm sure its 175/185w.

Its all about electrical rating. The TDP of a GTX is 177W. That's the maximum it's designed to use. It's not the maximum you can push it to. Overclock/overvolt outside the range it'll use more and the connectors/conductors/PSU will supply it until something gives/melts/trips. Now, in use, so far DX9 (no idea how DX10 will drive it) they've been measured around the 135W mark.

http://www.xbitlabs.com/articles/video/display/gf8800gtx-roundup_6.html

Overclock it, stock volts, maybe they get to 150W. The GTX has 2 PCIe 6 pin connectors because it's designed within the PCI standard and the TDP is more than 150W (slot+1x6pin) so it gets 2. Likewise, the 2900XT must use less than 225W at stock because it only needs 2x6 pins (note 20% tolerance as well). Once overclocked it must be close or even past this rating. Enough to force them to use the 8 pin connector, gaining a higher rating. As before, in use the power consumption is less than TDP. So 200W at stock makes sense. A TDP of 225W would be my guess. Overclocking past the normal range could take it a bit higher.

Edit: Saw this

The Halk · 4 May 2007 at 05:39

They really should rate them in Amps.

We all (or we all should at least) know the reason why they say 500W PSU is because there's a whole ton of "comes free with the case" PSUs out there that aren't fit to power a kettle.

These poor quality PSUs can't meet the demands on the 12V rails, and if you look at their specs you can see they don't cut it. There's 480W PSUs out there with 18 amps on the 12V rail. The Corsair 520W, for example, has 40 amps on the 12V rail.

There's only 40 watts of a difference there, supposedly. But once you've accounted for everything else in the case, one will run 2 cards in SLi/Crossfire, and the other won't even run one.

Edit : Incidentally... my new 8800 GTX was running on my old PSU. I had one molex. To that was connected a Y splitter. To each end of that Y splitter was connected another Y splitter. And to each of those 4 ends was connected one of the 4 connectors for the 2 PCI-E power connectors.

In other words I had it on a single molex. It worked fine, and you know why it worked fine. That's because the PSU I had was fine. The reason why they try and have it across as many "independant" rails as possible is to get as much steady current in as possible - which is only needed if people are using crap PSUs.

So I say stuff all that, and go back to a single floppy drive 4 pin connector on graphics cards like my old 9500 pro. It will work just as well. The only things that will suffer will be dodgy PSUs, and cheapskates that think they've got a "bargain" when they get one for less than 20 quid.

Just a note about OcUK. I've had a look at what appears to be a cheap and nasty PSU that they sell. It only looks cheap and nasty because it's up against all those nice Enermax, OCz etc PSUs. According to the manufacturer it does a combined 32 Amps on the 12V. Which is pretty impressive for a 450W PSU for 30 quid.

SteveOBHave · 4 May 2007 at 12:43

Fudz attempting to back up it's claim apparently -

http://www.fudzilla.com/index.php?option=com_content&task=view&id=829&Itemid=1

We created a lot of noise with the original 100 degrees Celsius article. You can read it here.

Luckily we have friends who wanted to help us out in our quest for truth. It turns out that testing the card without a case on in a 20 degrees C environment, the temperature of the Radeon HD 2900 XT will reach over 82 degrees.

There is a catch in 2D mode, as the temperature will be stable at about 58 degrees. In 3D mode the R600 chip heats up to an average of 68.4 degrees or in worst case scenario 71.3 degrees.

In a closed case enviroment the situation gets even worse, so you can expect even more than 82 degrees, not far off the 100 degrees that we originally reported.

Clearly 82C is close to 100C... Bad us. Still 82C doesn't ring entirely true, it's getting very close to an unsafe operating temperature for electronics and I struggle to believe that ATI could put out a card with such little leeway for temperatures. Of course I could be wrong...

Gibbo · 4 May 2007 at 12:48

SteveOBHave said:
Fudz attempting to back up it's claim apparently -

http://www.fudzilla.com/index.php?option=com_content&task=view&id=829&Itemid=1

Clearly 82C is close to 100C... Bad us. Still 82C doesn't ring entirely true, it's getting very close to an unsafe operating temperature for electronics and I struggle to believe that ATI could put out a card with such little leeway for temperatures. Of course I could be wrong...

HI there

A closed case has little effect on the cards temperature due to the heatsinks design on the card which effectively exhaust heat out of the case. So the temperatures do not rise by much, in my testing temps remained under 80c which is nowhere near reported 100c.

The difference between 80c and 100c is huge.

Gerard · 4 May 2007 at 12:54

Considering my x1900xt routinely hits 96c with no probs i wouldn't care if the new card was hitting 100c.

stickroad · 4 May 2007 at 12:57

This is the most ridiculous thread ever. Oh No the 2900XT may need a 750 Watt PSU, seriously, do you really think the people over at ATI would be so dumb?

I just cant wait till ATI release the card, all this speculation is getting wayyyyyyy out of hand.

Also does anyone have a list of ATIs full Direct X 10 card range? Or dosent anyone know yet?

londonbairn · 4 May 2007 at 13:16

Is it possible that the guy who is writing for Fudzilla is the (former) Iraqi Information minister. His claims are just as astonishing as his...

SteveOBHave · 4 May 2007 at 13:25

Gerard said:
Considering my x1900xt routinely hits 96c with no probs i wouldn't care if the new card was hitting 100c.

See that's just on the cusp of damaging the circuitry. 96C is waaaay too hot. My 7900GT tops at 51C when getting totally lashed and idles at 33-34C in 2D mode. (Room temp 22-23C)

Gibbo said:
HI there

A closed case has little effect on the cards temperature due to the heatsinks design on the card which effectively exhaust heat out of the case. So the temperatures do not rise by much, in my testing temps remained under 80c which is nowhere near reported 100c.

The difference between 80c and 100c is huge.

I know, it's a massive difference - they are clearly attempting to cover the earlier wild claims.

Dutch Guy · 4 May 2007 at 13:35

Tom|Nbk said:
they do give more power than previously needed, but the power can not exceed 225w as PCI-E cannot handle anymore than this. No matter how many fancy connectors you add

What is the limitation :confused:

There is only 75W coming from the PCI-E slot so if the rest of the power comes from extra connectors why can you (theoretically) put in a card that uses 500W with only 75 coming from the PCI-E slot?

Gerard · 4 May 2007 at 13:38

SteveOBHave said:
See that's just on the cusp of damaging the circuitry. 96C is waaaay too hot. My 7900GT tops at 51C when getting totally lashed and idles at 33-34C in 2D mode. (Room temp 22-23C)

According to ati theyre safe up to anything under 120c. I really don't undertand why they run so hot considering the cooler on it (leakage maybe?) but its served me well for over a year now.

dalin80 · 4 May 2007 at 13:43

all this of course presumes that your house itself has a good supply. i live in a mighty old house with 'dirty' power as the sparky put it i couldn't run a x1800gto and p4 3.2 with a 380w supply here would constantly crash under load. needed mid 400's for it to be stable.

my 2 8800's gts 320's wont be stable with anything less then a good 700w supply.

played havoc with our old cooker the clock was about 10% slow.

Dutch Guy · 4 May 2007 at 13:46

dalin80 said:
all this of course presumes that your house itself has a good supply. i live in a mighty old house with 'dirty' power as the sparky put it i couldn't run a x1800gto and p4 3.2 with a 380w supply here would constantly crash under load. needed mid 400's for it to be stable.

my 2 8800's gts 320's wont be stable with anything less then a good 700w supply.

played havoc with our old cooker the clock was about 10% slow.

And that is why AMD say we need a 750W PSU

drunkenmaster · 4 May 2007 at 13:47

why is 96C wayyy to hot exactly? based on the scale of temperatures we like to be in sure, 96C is crazy, 51C is crazy to us too. but we're not electrical components design to operate at the temps that they operate at. you can produce a card that is designed to run at 10C and only have components on it that are rated to 25C, as long as it runs under 25C its fine. you can also have a card that routinely hits 120C and has capacitors and other things that are all higher quality and rated at 150C. 96C is hot to us, its not hot to something that is rated to 125C.

it really really doesn't matter what temp it runs at, ok it will affect overclocking but if it runs stable at stock thats 99.9999999% of sales that will not be affected.

fornowagain · 4 May 2007 at 13:53

Someone should mention the 7950GX2, mine without a little cooling help lived in the 90's when loaded. Didn't hurt it one bit, nor did it throttle back as it was less than the overheat limit. I've a laser temperature probe for IC's and you can find the odd component on motherboards getting that hot. Its not ideal so best to add heatsinks, but there you go. Just as well solder melts at 180C.