Inquirer: its CERTAIN Fermi will require a waterblock or phase change refridgeration

[josiahsuarez]

http://www.theinquirer.net/inquirer/news/1567017/nvidia-fermi-gpu-cut
What seems certain is that Nvidia's fab partner TSMC is finding it challenging, to say the least, to manufacture the chip, and that it will have to be cooled by a waterblock if not a more exotic cooling method like phase-change refrigeration.

there you go, confirmed by a reputable (?) source

[Eliminator]

yea real reputable source there

[shadow001]

yea real reputable source there

Charlie isn't working there anymore at least...:p

[fasedww]

I'm waiting, I'm going to watercool a pair in Sli on rig #2 below:D

[mailman2]

I'm waiting, I'm going to watercool a pair in Sli on rig #2 below:D

FX 5800 II. Windows ME of GPUs.

Its going to be delayed again btw, so you MIGHT have it up and running by summer of 2010. In the meantime I'm really enjoying my HD5870. :captnkill:

[Sazar]

We don't have to go all the way back to the NV30 days for a comparison honestly.

I am thinking back to the 2900 days from ATi/AMD :bleh:

[Muppet]

Charlie isn't working there anymore at least...:p

Like that makes a difference to the rubbish they post.

[Viral]

Yes, at worst it will be like the 2900. I'm still thinking it will definitely outperform 5870, but it could very well run very hot and not perform like expected.

There's no way it will be like the 5800 since it would have to underperform 5870, be noisy and hot, have crappy IQ and not properly support DX11. Nvidia wouldn't release such a GPU.... again.

[ViN86]

LOL the Inquirer.

[FearMeAll]

Well my 280 runs in my basement at 1280x720..good thing it's a low rez cus i will have this card for a while lol..

[Woodelf]

Sound's like they're just surfing the net, getting the same info we allready have (none).

I just hope that nvidia's claim of new architecture will make what we know about GPU's irrelivent. Meaning, the same way we look at modern cpu performance architecture... in the old days MHZ ruled so that's what we argued about (pre-release).

[Buio]

Charlie isn't working there anymore at least...:p

The article on the Inq linked to a post on Charlie's site, so Charlie was indeed involved. :P

[shadow001]

The article on the Inq linked to a post on Charlie's site, so Charlie was indeed involved. :P


In his defence though,and even though it's well known he hates Nvidia with a passion,and that many of his articles turn out wrong,if you look at the actual information with regards to fermi that he published on his site over the last 3~4 odd months or so,he's been pretty accurate so far.


He always stated that Fermi was only going to happen sometime in Q1 of this year,even when Nvidia's Own CEO was hinting about a late november 2009 release for Fermi,so where ever his getting his information from,it's been accurate,att least in this case.


The other thing to keep in mind is that we don't even know if the A3 revision will be the actual retail shipping product,and that they might need yet another revision....We'll know in due time.

[Redeemed]

I'm not holdin' my breath for Fermi. I'd rather expect little from it and have it surpass my expectations- than expect the world from it and have it fall short. ;)

[XDanger]

I run my 8800gt passive with an accelero, A similar cooler (with a fan) is all Fermi will need but that might increase the slot size unless they make a neat design.

Current stock graphics cooling solutions are crap.

The Fermi chips are big ya so they can put 6 DIRECT CONTACT PIPES on there and be cooled nice.

If the boards allow topside cooling that would be extra nice.

Currently graphics ram is heatsink cooled with an aftermarket cooler but if a heatpipe cooler is stock we may get pipe cooled ram which would be better, prahblee not though eh...


They always skimp on cooling.

1 big cooling part that goes over the GPU and RAM FTMFL!!:thumbdwn:

[Redeemed]

I run my 8800gt passive with an accelero, A similar cooler (with a fan) is all Fermi will need but that might increase the slot size unless they make a neat design.

Current stock graphics cooling solutions are crap.

The Fermi chips are big ya so they can put 6 DIRECT CONTACT PIPES on there and be cooled nice.

Really? Comparing your 8800GT to the top end Fermi? In terms of power consumption?

Fermi might be on a smaller process than your 8800GT (G92?) but that doesn't mean it'll run cooler.

When they went with the 55nm process for the GTX200 series, that barely made a dent in thermals and power consumption. In fact a GTX260 @ 55nm was near identical to the original GTX260 @ 65. Near identical power consumption and thermals.

Fermi might be on a smaller process, but it's also a very large chip. I mean, if it were as easy to do as you're implying, don't you think they'd have been able to put the chip to market by now? I'm sure part of the delay has to do with keeping the cards within a reasonable power envelope. I mean, we can't have a GTX360 that's near the 300watt limit now can we?

[XDanger]

Really? Comparing your 8800GT to the top end Fermi? In terms of power consumption?

Fermi might be on a smaller process than your 8800GT (G92?) but that doesn't mean it'll run cooler.

Oh no , I'm aware its a hot chip but like I said a heatpipe cooler will suffice and not phase change or water.

The fact its a large chip just means more space for pipes as far as I'm concerened,


They want to cut corners with cooling and its taken them time to realize they cant becuase they are stupid.

Listen to me Nvidia !! I am SMRT!!! ;)

[Redeemed]

Oh no , I'm aware its a hot chip but like I said a heatpipe cooler will suffice and not phase change or water.

Okay, I misread your post then. :o

I also agree phase change and water will *not* be necessary. A standard air cooler with heat pipes will do just fine.

Heck, I'm running dual 8800GTXs and when gaming neither one get's much above 50*C.

[Xion X2]

Heck, I'm running dual 8800GTXs and when gaming neither one get's much above 50*C.

The resolution that you run at plays a big part in temps, as well as the settings. I had a 7900GTX at one point that never loaded over 53C. At the time, I was gaming on a 1280x monitor. When I made the jump to 1680x, load temps went up to the low to mid 60s.

Just thought I would mention that since I remember you saying you ran at a low res (1024x/1280x or something.) For those users who run at 1680x, 1920x or, in the rare cases, 2560x, it heats a card up like crazy--especially if you apply AA (transparency AA, even moreso.)

On Fermi--Nvidia has to consider the worst "case" scenario (pun intended), meaning the user who has no idea how to ventilate his case properly. Fermi may run 60c in Joe's case and 90c in Bill's case because Bill doesn't know how to properly cool a case, or maybe Bill's case simply doesn't have any vents to intake air. Since there are so many different cases out there that a user could own, and various outside ambient temps that the card may operate in, it's an inexact science.

The problem w/ keeping Fermi cool is largely due to its transistor count. 3.0B transistors will obviously pack a lot of heat.

ATI improved their coolers for the 5xxx series; maybe Nvidia will too.

[Redeemed]

The resolution that you run at plays a big part in temps, as well as the settings. I had a 7900GTX at one point that never loaded over 53C. At the time, I was gaming on a 1280x monitor. When I made the jump to 1680x, load temps went up to the low to mid 60s.

Just thought I would mention that since I remember you saying you ran at a low res (1024x/1280x or something.) For those users who run at 1680x, 1920x or, in the rare cases, 2560x, it heats a card up like crazy--especially if you apply AA (transparency AA, even moreso.)

On Fermi--Nvidia has to consider the worst "case" scenario (pun intended), meaning the user who has no idea how to ventilate his case properly. Fermi may run 60c in Joe's case and 90c in Bill's case because Bill doesn't know how to properly cool a case, or maybe Bill's case simply doesn't have any vents to intake air. Since there are so many different cases out there that a user could own, and various outside ambient temps that the card may operate in, it's an inexact science.

The problem w/ keeping Fermi cool is largely due to its transistor count. 3.0B transistors will obviously pack a lot of heat.

ATI improved their coolers for the 5xxx series; maybe Nvidia will too.

You are right.

However, when I had my dual 640MB 8800GTSs they ran at about the same temp- and that was @ 1680x1050.

[shadow001]

Oh no , I'm aware its a hot chip but like I said a heatpipe cooler will suffice and not phase change or water.

The fact its a large chip just means more space for pipes as far as I'm concerened,


They want to cut corners with cooling and its taken them time to realize they cant becuase they are stupid.

Listen to me Nvidia !! I am SMRT!!! ;)


You have to realise why it's a hot chip in the first place....Your G92 GPU inside the 8800GT only has about 750 million transistors in it,while fermi packs 3 billion +,which is 4 times the amount and the die has a contact surface with the heatsink less than twice as large as the 8800GT does(G92 is 330mm^,while fermi is reported at 530mm^),although Nvidia does use a heat spreader,not to mention the higher clock speeds Nvidia are aiming for Fermi,again compared to your 8800GT.


So there's more power going in,spread thru 4X as many transistors,running at a higher clock speed,and with a contact surface from the Fermi die to the heatsink less than twice as large as the G92,so more heat is concentrated in a relatively small area,all things considered.

[Redeemed]

You have to realise why it's a hot chip in the first place....Your G92 GPU inside the 8800GT only has about 750 million transistors in it,while fermi packs 3 billion +,which is 4 times the amount and the die has a contact surface with the heatsink less than twice as large as the 8800GT does(G92 is 330mm^,while fermi is reported at 530mm^),although Nvidia does use a heat spreader,not to mention the higher clock speeds Nvidia are aiming for Fermi,again compared to your 8800GT.


So there's more power going in,spread thru 4X as many transistors,running at a higher clock speed,and with a contact surface from the Fermi die to the heatsink less than twice as large as the G92,so more heat is concentrated in a relatively small area,all things considered.

Yeah, this could be one hot chip.

[Xion X2]

You are right.

However, when I had my dual 640MB 8800GTSs they ran at about the same temp- and that was @ 1680x1050.

GTS runs a little cooler than GTX, so perhaps that's the reason. (If you're running the same case and airflow, that is.)

[Redeemed]

GTS runs a little cooler than GTX, so perhaps that's the reason. (If you're running the same case and airflow, that is.)

Nah. Different case now. My current one probably has slightly better air flow though. :D

[Xion X2]

Yeah, this could be one hot chip.

http://img228.imageshack.us/img228/5797/hotdamnfermigeeeeforce.jpg











http://img709.imageshack.us/img709/6337/fermisaucepan.jpg











http://img38.imageshack.us/img38/3231/firehosefermiiiiiii.jpg