Heatspreader on GPU
- Thread starter tEd
- Start date
Heatspreaders tend to be slightly inferior to direct contact with the heatsink when it comes to cooling. For CPUs, heatspreaders didn't come into vogue until around the time that the clamping pressure on the heatsink mounts became great enough to regularly crack or chip the processor.
Unless one is likely to change heatsinks or mount them outside of the factory, as is the case for a lot of CPUs, heatspreaders aren't usually a win.
Unless one is likely to change heatsinks or mount them outside of the factory, as is the case for a lot of CPUs, heatspreaders aren't usually a win.
Matasar said:
This got me thinking about the PCI expansion slots, and now that the bulk of everything is integrated right on the motherboard, how many people still need more than one available slot for a card? Personally, I've got a couple systems that have one slot with an Audigy or it's ilk installed, and the on-board sound disabled... but other than that I'm really not using a PCI slot except for a sound card.
I remember everyone decrying the two-slot cooling back in the GF5900 days... but how many people actually needed that slot, and how many people simply wanted something to complain about? I generally don't use the slots next to the AGP/PCIE slot for the simple reason that I want as much airspace around the graphics card as possible.
flf said:
Depends on the motherboard, but you could end up with a NIC, sound, SCSI controller, physics add-in card <g>, etc. Both of my boxes right now only have a sound card, but that's because the motherboards sent out with review kits are fairly feature-rich.
Actually, I use far more PCI header add-in slots (is that what they are?) than actual cards. My motherboard came with 7.1 Envy24 audio, so I use the add-in slot to give me the optical in / out and coaxial in / out jacks. I also put in the add-in slot for the firewire connectors, the add-in slot for more USB connectors, and finally an actual PCI SCSI card for my external tape backup unit.
That's three slots I've used without putting any cards in, and another one from a simple SCSI card. And while I'm probably using more of that crap than a lot of people otherwise would, that's still not leaving me much room for other stuff.
Still, for the majority of folks, I don't think a large cooler would hinder much. Hell, I'm using a waterblock on my X800 Vivo right now and not having any problems with clearance between it and my SCSI card and the other nonsense slot filler things.
That's three slots I've used without putting any cards in, and another one from a simple SCSI card. And while I'm probably using more of that crap than a lot of people otherwise would, that's still not leaving me much room for other stuff.
Still, for the majority of folks, I don't think a large cooler would hinder much. Hell, I'm using a waterblock on my X800 Vivo right now and not having any problems with clearance between it and my SCSI card and the other nonsense slot filler things.
tEd said:
All the heatspreader is doing is adding an extra layer of metal between the core and the heatsink. Without a heatspreader, the heatsink itself is acting as a "heatspreader." Wouldn't matter if the core were only the size of a pinhead - direct cooling would still be better.
aaronspink
Veteran
tEd said:
Heatspreaders as used in most desktop products are primarily designed to protect the actual silicon die and surrounding components. There are some applications where a high end heat spreader can enhance the cooling, but in the desktop space this is generally not true since the spreaders themselves are just slugs of copper.
Ideally you would have the heatspreader funcationality integrated into the heatsink itself which is often the case anyways.
I won't even get into the issues of TIM junction inperfections from mass production, or what seems to be substandard TIM or assembly issues that appear to affect some AMD device over repeated mountings. I would actually like to see AMD and Intel move to a system close to that used in the graphics market with a mounting ring that provides the protection while allowing DIE-TIM-Heatsink mounting.
Aaron Spink
speaking for myself inc.
GPU's are typically getting larger or staying about the same size these days. They're just getting much more dense.tEd said:
Edit (side comment):
As GPU's optimize for branching scenarios and multithreaded execution, they're going to require greater I/O caches. Caches are typically rather low-power for the area consumed compared to logic. So GPU's are going to both get larger, and consume power over less of that area. So heat spreaders really won't help at all.
I popped off a K6-III+'s heatspreader a year or so ago to try to get it stable at 600MHz. The core was so fragile that even though I knew to be careful and tried VERY hard to do so, I ended up breaking the chip. But I've never broken any other chip by cracking the core.
Obviously though it was just a smart move to put those heatspreaders on. An extra few degrees cooling is certainly worth losing over reducing returns.
Obviously though it was just a smart move to put those heatspreaders on. An extra few degrees cooling is certainly worth losing over reducing returns.
aaronspink
Veteran
swaaye said:
No. The heat spreaders that AMD and Intel use are nickel plated copper. The nickel plating is to prevent oxydation which would degrade the heatspreader-tim-heatsink interface.
Some higher end products utilize high conductive ceramics that have higher thermal conductivity than copper.