Cooling Revolution.

Powderkeg

We're not "happily forgetting" anything like that. The problem is you keep ignoring one important fact:

GPU = 25 watts
CPU = 85 watts

The CPU requires the massive majority of the cooling power coming from those fans. Not only that but the rest of the system (motherboard ect) also requires some cooling from those fans, and no doubt you could do without such a complex motherboard without the 85 watt CPU.

Though that isn't to say that I believe Revolution will use the same GPU as 360, because I don't (for instance I expect Rev's GPU to have a single die).
 
Last edited by a moderator:
Thanks for the info Powderkeg I thought they are the ones how are more advanced in process terms and Intel, AMD (?) Nec should be ready for mass production in mid 2006, I thought they would be too (is IBM how will make their chips, right?), if so thatis bad news for Nintendo, or 90nm or low quantitity of Revs:cry: .

Teasy: I think that Xenus use 35W.
 
Even if Xenos consumes 35Watts, that's still only about 30% of the total power consumption (85+35= 120). A jump from 25% to 30% is very small and will hardly make any difference in cooling requirement for the GPU.

Urian said:
2 PPE with 2 VMX each and Shared L2 Cache. Total 100GFLOPS at 2.5Ghz.

Hmm..that's pretty interesting.

One source said that the processor was a dual-core one, whilst the other would not reveal any technical information.

If this is true then it's likely two identical cores vs a mother/daughter core as in (PPE+PPU) on same die.
 
Remember that when you double the clock speed of a CPU the power consumption becomes 4 times mores greater than the original consumption.
 
About PowerPC 970FX power consumption (straight from IBM docs):
Again, this application note will discuss the 2.0 GHz, 105°C maximum junction temperature (Tj), version of the PPC970FX with a maximum power dissipation of 39 watts. This is a large amount of power in a small die, (66mm2). It is critical to remove the heat properly or the processor will fail.

From what I could find the 970MP is a dual 970FX with twice the L2 cache (512KB->1MB) and is consuming a maximum of 100W with both cores running full speed.
I also found that the 970FX do 8FLOPS per cycle, which means 16GFLOPS @ 2.0GHz.

Unfortunately I couldn't find the relevant doc on IBM website :(

If Xenon is really consuming 85W, should I get that the most consuming part of the chip is the branch predication/Out of Order Execution part ?
Cause that's way lower than the 970MP which is dual core only...

[edit]
Found out that the new 970FX is supposedly consuming less power, but can't find any official figures, websites aren't very reliable since they don't say whether quoted Watts are "average"/"typical" or "maximum"...
 
Powderkeg said:
Sure the heatsink is small, but like PC-Engine you happily forget about the size of the fan and ducting that that heatsink requires to keep the GPU cool. Where in the Revolution are you going to fit 2X80mm case fans and a large ducting unit? You'll have to find somewhere to put them if you plan on using such a small heatsink.

Why do you think that two 80mm fans are needed, it is the CPU that needs the wast mayority of the cooling from the fans. Remember the XGPU uses less power than highend laptop GPUs. Anyway you could save the space needed by uisng radial fans.
 
Teasy: I think that Xenus use 35W.

Do you have a link/quote for that? Because I was pretty sure that it was 25w.

BTW I just realised that the XBox 360 GPU isn't even actively cooled at all. The two 80mm fans are ducted towards the CPU heatsink and the GPU heatsink is left completely outside the duct. Not only that but its totally covered by the DVD-ROM drive (the GPU heatsink is so close to the DVD-ROM its just about touching it). The way its all set up the GPU would get almost no air flow. This really just goes to prove that these fans are almost completely there for the CPU (the rest is just for clearing a bit of air from inside the case which any little fan could do easily).
 
Last edited by a moderator:
Ingenu said:
About PowerPC 970FX power consumption (straight from IBM docs):


From what I could find the 970MP is a dual 970FX with twice the L2 cache (512KB->1MB) and is consuming a maximum of 100W with both cores running full speed.
I also found that the 970FX do 8FLOPS per cycle, which means 16GFLOPS @ 2.0GHz.

Unfortunately I couldn't find the relevant doc on IBM website :(

If Xenon is really consuming 85W, should I get that the most consuming part of the chip is the branch predication/Out of Order Execution part ?
Cause that's way lower than the 970MP which is dual core only...

[edit]
Found out that the new 970FX is supposedly consuming less power, but can't find any official figures, websites aren't very reliable since they don't say whether quoted Watts are "average"/"typical" or "maximum"...

Xenon is a In Order CPU, so is something else that generate the heat.

The new 970Fx consumes about 16W at 1,6Ghz (probably are average or maximum as they are for notebooks mainly, I guess if they hit the 65nm they can make ~16W at 2Ghz), but the flops are low, yet if they get 2 of those (OoO and easy to programe= cheap to devolop games, which should be one of the main tatics from their strategy IMO) and a maths/vector coprocessor (IBM or ATI for good performance in heavy flop tasks, many ask for a PPU) should be a nice design price/heat/performance wise and if they really are using 1T-Sram (like is said)latency should be very low.

Are those the docs you want? If it is for the MP, I didnt find it too.

Teasy said:
Do you have a link/quote for that? Because I was pretty sure that it was 25w.
No I dont, I think :)wink: ) I read somehere from a reliable source (Dave B. or other article linked somehere in this forums) but I dont have the link or cant confirm it, but from what I remember is 35W:???: anyway it still very low.
 
Teasy said:
BTW I just realised that the XBox 360 GPU isn't even actively cooled at all. The two 80mm fans are ducted towards the CPU heatsink and the GPU heatsink is left completely outside the duct.

Looks to me like the GPU gets at least some of the airflow in this pic.
 
pc999 said:
Nice , I hope it will have somethingh to boost math intensive tasks too (beyond VMX).

Just wondering, when they talk about prototipe is final version (mass production ready) or is one that still need debugging/corrections and improvements :?:

BTW anyone knows when IBM can start mass production of 65nm chips (remeber that PS3 will be done in Sony plants, so it is not valid for comparition) :?:

Thanks

IBM is working together with Charter Semiconductor,AMD, Samsung, and Infineon.

IBM has created the SSDOI that also reduces the 970FX wattage.

http://www.cpu-planet.com/features/article.php/3321251

Broadway a Custom PowerPC 970MP @ 3.2GHz(15W) manufactured in 65nm with SSDOI and with PowerTune(IBM's powersaving optimized clock trick)

with a PPE Physics Processor (XGPU like heatsync).

"In addition, AMD purchased an $11 million license from IBM so it can produce 90nm and 65nm products at AMD's 300-millimeter Dresden fabrication facility, a third-party foundry or a joint manufacturing facility owned by AMD and one of its partners. Previously, all the work was being done at IBM's facility in East Fishkill, New York."

http://internetnews.com/ent-news/article.php/3412251

So IBM already has the technology and they already have the fabs plants(east fishkill in this case).

Sony paid $1.67 billion to manufacture its own 65nm fab line in 2003.

http://www.geek.com/news/geeknews/2003Apr/bga20030421019680.htm

They can easily pay IBM the license to use thier fabbing techniques like AMD did.

nintendo can have IBM build a custom 65nm part with the SSDOI process to cut back cost and wattage.(remember nintendo said quiet and affordable or nintendo's definition of affordable...$200-$250)

Broadway can be the most powerful CPU for next generation comsoles for doing general purpose code and ease of use(porting from PC will be easier)

Seeing how flops don't equal power or CPU complexity(more like the FPU's "power")I don't care about meaningless hyped PR statistics (remember Sega's blast processing and Sony's EE doing $120 million polygons same thing different era) or how GFLOPs will be the end all be all to computer archetecture.

PowerPC 970MP came out last month so by mid-late 2006 they will be cheaper.Manufactured with SSDOI the wattage is cut to 24.5W.

Built in 65nm, the 970MP would be even cheaper and cooler.Using the 970MP's built in Powertune technology it will be even more cooler.
 
At last I found it:
http://www-306.ibm.com/chips/techlib/techlib.nsf/techdocs/50059BE7FA1363DB872570910069037C/$file/PowerPC_ZCard_Nov2105_TGO01544.pdf

PPC970MP 90nm SOI Max power consumption : 70W @ 2.0GHz.


I bet "PowerTune" technology is a way to cut CPU power when IDLE, something you really don't care about on a console, since it's unlikely to be idle when you play...
So only the shrinkage (0.9=>0.65 ~30% reduction), and the SOI => SSDOI can be taken into account, as well as any architectural change.
 
Teasy said:
Do you have a link/quote for that? Because I was pretty sure that it was 25w.
Let's just say I find the idea of a 500MHz 200+ M transistor processor using only 25W power a little unlikely. Even 35W would be quite extraordinary, especially considering ATi's current products (which consume huge amounts of electricity).

BTW I just realised that the XBox 360 GPU isn't even actively cooled at all.
Better put on your glasses and take another look, because the sink is definitely cooled by the duct.

Really, there'd be no point at all in having a sink on the GPU if there was no active airflow around it. All it would do was delay the inevitable meltdown that would happen when the hunk of metal heated up enough.

Heatsinks are there to increase the air contact, but in a stagnant system they fill no useful purpose, they just turn into a heat store. It could just as well have been a solid bar of metal instead if there was no forced air across it.

This really just goes to prove that these fans are almost completely there for the CPU
No it doesn't.
 
PSXer said:
Looks to me like the GPU gets at least some of the airflow in this pic.
Yes, it at least is meant to get sufficient airflow from the duct, but I wouldn't be surprised if the airflow is much smaller than hoped.
The sharp 90 angle because of the DVD drive mounting could essetially stop much of the airflow into the very low duct going to the GPU, the corner could actually introduce a lot of air resistance by making air turbulences on the opening, and in the worst case just push the air back towards the fans.
However, I don't know if the air cannel to GPU is more rounded inside, but I doubt it.
 
Guden

I'd really prefer to hear what ATI have to say on its power consumption, rather then what you think is likely. I remember quotes from ATI in this forum about its power consumption, I just wish the search engine on this forum wasn't crap :D

Better put on your glasses and take another look, because the sink is definitely cooled by the duct.

I need glasses because none of the pictures posted in this thread showed the duct fully?, yeah that makes sense...

Really, there'd be no point at all in having a sink on the GPU if there was no active airflow around it. All it would do was delay the inevitable meltdown that would happen when the hunk of metal heated up enough.

I said no active cooling, not no airflow.. I even mentioned that the fans would remove the hot air from the system in general.
 
Last edited by a moderator:
Looks to me like the GPU gets at least some of the airflow in this pic

Ah right the DVD-ROM was hiding some of the duct in the other pics. So the GPU does get a little bit of direct air (little compared to what the CPU gets that is), thanks for the pic.
 
http://images.anandtech.com/reviews/system/microsoft/xbox360/no_cov.jpg

Looking at that pic it seems the air flow for the heatsink on the GPU mainly enters through the opposite side of the duct since the top of the heatsink is covered by the DVD drive, though there's small gap between the DVD drive and heatsink. This configuration allows higher air velocity, but not necessarily higher air volume. Higher air velocity is a very good thing for heatsinks if you can keep noise down. You can attain high air velocity with small fans too as long as you keep the air passage narrow like what's being done with Xenos's GPU.

BTW notice how close the CPU heatsink is to the DVD drive. I wonder if MS intended the DVD drive's metal housing to be some kind of heatsink too. I think it would've been really nice if they wrapped the DVD drive with a thin sheet of aluminum about the same thinkness as a cola can. That would've helped a lot.
 
I really hope the heat from the GPU is not rising up to the dvd player, if that it will probably kill the dvd player eventually. I have heard a lot of horrid stories about people putting their standalone dvdplayers on top of their receiver, and effectively killed the player in just a few days. Of course you cant do a direct comparison here, but I think it is to some degree reducing the xbox 360s dvdplayers lifetime unless all heat is directed in another direction with the help of that fan. Interesting anyway. And maybe the back part of the xbox 360 player is less sensitive to heat than where the lens is wich should be somewhere in the middle of it?
 
If the fans and ducting are doing their job, there should be very little heat rising from the sinks. The heat should be continually pulled toward the duct before it could even rise.

Anyway here are cross sectional diagrams I made of how I would design Revolution. :cool:

The real console should be closer to the top view diagram ie slightly longer than what you see on the side view diagram and the fan (in the side view diagram) should be slightly larger in diameter and height too. If required two fans side by side would also still fit..:p

Orange is slim DVD drive, blue is radial fan with rear exhaust, grey is unified GPU/CPU heatsink, red is heat resistant plastic air duct, black is GPU and CPU, and green is PCB.

For reference:

http://www.hardocp.com/images/articles/1130724723wQf6EywE1q_1_2_l.jpg

http://www.techspot.com/reviews/hardware/sapphire_radeonx1800xl/Image17.jpg
 

Attachments

  • Revolution diagram top.JPG
    Revolution diagram top.JPG
    45.4 KB · Views: 14
  • Revolution diagram.JPG
    Revolution diagram.JPG
    17 KB · Views: 13
Back
Top