Monday, October 22nd 2018

Alpenföhn Intros Matterhorn Threadripper CPU Cooler

Alpenföhn today introduced the Matterhorn Threadripper, a variant of its popular Matterhorn series tower-type CPU air coolers, which supports AMD socket TR4 (SP3r2). The cooler comes with an enlarged copper base, which offers full coverage of the Ryzen Threadripper IHS. Six 6 mm-thick copper heat pipes convey heat from this base through the aluminium fin-stack.

The cooler ships with a re-tuned Wingboost 2 120 mm fan, which spins between 500 - 1,500 RPM, pushing up to 106 m³h of air, with noise output as low as 18.2 dBA. Both the fins and the heat pipes feature a ceramic black coating that works to increase surface area for heat dissipation. A tube of Alpenföhn Permafrost TIM comes included. Measuring 138 mm x 100 mm x 158 mm (WxDxH), the cooler weighs about 1 kg. The company did not mention the thermal load limits of the cooler, and whether it's recommended for even the 250W TDP models such as the Threadripper 2990WX. The company didn't reveal pricing.
Add your own comment

30 Comments on Alpenföhn Intros Matterhorn Threadripper CPU Cooler

#1
Vayra86
That is one helluva sexy fan color scheme.
Posted on Reply
#3
bug
I know this isn't an apples-to-apples comparison, but the hottest Pentium 4/D didn't need a 250W capable heat sink.
Posted on Reply
#4
Ripper3
bug, post: 3927429, member: 157434"
I know this isn't an apples-to-apples comparison, but the hottest Pentium 4/D didn't need a 250W capable heat sink.
That is an odd comment to make. Hottest Pentium Ds had 130W TDP. Cooling has come a long way since, even air cooling.
Posted on Reply
#5
bug
Ripper3, post: 3927443, member: 39016"
That is an odd comment to make. Hottest Pentium Ds had 130W TDP. Cooling has come a long way since, even air cooling.
It just struck me what was considered excessive heat back then as opposed to now.
Back then, a 500g heat sink was considered massive. Today it's only considered good to run mild overclocks, at best.
Posted on Reply
#6
noel_fs
Too many heat pipes for that size imo, probably will be overpriced
Posted on Reply
#7
HTC
bug, post: 3927447, member: 157434"
It just struck me what was considered excessive heat back then as opposed to now.
Back then, a 500g heat sink was considered massive. Today it's only considered good to run mild overclocks, at best.
Back then, the socket it was designed for was ... a tad bit ... smaller than what this particular cooler is designed for ...

That said, big coolers are required because, with the smaller and smaller fabrication processes, the more the heat is concentrated within smaller areas and that needs more "grunt" to be able to dissipate.

If it continues this way, it's possible that air coolers will stop being enough (beyond stock), unless there's some sort of breakthrough with either cooler or CPU design that enables the heat to be more evenly distributed across a wider area in the socket.

AMD's multi-chip design is a good approach in this direction but i think it's still insufficient as manufacturing processes advance: would not surprise me if AMD are forced to lower clocks because of temp issues or ship / require beefier CPU air coolers with which to work @ stock. Unless ofc they pull a "9900K-like chip release" which, IMHO, would be absolutely terrible ...

Remember: being made in 7nm makes the heat much more concentrated. It may actually be one of the reasons Intel is having so much trouble with 10nm as well.
Posted on Reply
#8
bug
HTC, post: 3927484, member: 51238"
Back then, the socket it was designed for was ... a tad bit ... smaller than what this particular cooler is designed for ...

That said, big coolers are required because, with the smaller and smaller fabrication processes, the more the heat is concentrated within smaller areas and that needs more "grunt" to be able to dissipate.

If it continues this way, it's possible that air coolers will stop being enough (beyond stock), unless there's some sort of breakthrough with either cooler or CPU design that enables the heat to be more evenly distributed across a wider area in the socket.

AMD's multi-chip design is a good approach in this direction but i think it's still insufficient as manufacturing processes advance: would not surprise me if AMD are forced to lower clocks because of temp issues or ship / require beefier CPU air coolers with which to work @ stock. Unless ofc they pull a "9900K-like chip release" which, IMHO, would be absolutely terrible ...

Remember: being made in 7nm makes the heat much more concentrated. It may actually be one of the reasons is having so much trouble with 10nm as well.
Yeah, I was questioning Ryzen's design. But since you brought that up, that is not a silver bullet either since in higher core designs IF ends up drawing more power than the cores themselves. A lot more. Atm I'd say it's hard to predict where future designs will go.
Also, I wasn't talking about density either. I was just noticing how TDP has grown over years. I understand we have more cores now, cores that are way faster. Fwiw, the same cores perform way better using way more power in laptops. Yet for the desktop, the TDP is what it is.
And that's all I meant: an observation. I suggest we leave it at that and move on?
Posted on Reply
#9
bonehead123
words:

graphene
nanotubes
monoexcongravitumalkalydictpolymerlithographs :D
Posted on Reply
#10
Camm
DeathtoGnomes, post: 3927417, member: 151150"
looks a bit small for 250TDP.
You would be surprised. My TR 1950X was easy to keep under 60c under Prime with a U12S in Push\Pull. Its the high socket size that does it IMO, that increased surface area is just a godsend.
Posted on Reply
#11
HTC
Camm, post: 3927809, member: 110377"
You would be surprised. My TR 1950X was easy to keep under 60c under Prime with a U12S in Push\Pull. Its the high socket size that does it IMO, that increased surface area is just a godsend.
It's working with 14nm / 12nm: how will it work with 7nm? The heat will be more concentrated and can, in a worst case scenario, force AMD to lower clocks (boost mostly but possibly also base, depending on how much we're talking about.
Posted on Reply
#12
Camm
HTC, post: 3927867, member: 51238"
It's working with 14nm / 12nm: how will it work with 7nm? The heat will be more concentrated and can, in a worst case scenario, force AMD to lower clocks (boost mostly but possibly also base, depending on how much we're talking about.
IPC should go up with an increased transistor count, so personally I don't really care what the clockspeed is.
Posted on Reply
#13
HTC
Camm, post: 3927869, member: 110377"
IPC should go up with an increased transistor count, so personally I don't really care what the clockspeed is.
That depends on whether or not they're forced to lower the clocks and, if so, by how much: if the IPC increases by ... say ... 15% but the speed drops by ... say ...20%, you would actually be losing performance, no?

Hopefully, we'll see AMD's Zen 2 not only increasing IPC while @ least maintaining clocks: fingers crossed ...
Posted on Reply
#14
ghazi
HTC, post: 3927867, member: 51238"
It's working with 14nm / 12nm: how will it work with 7nm? The heat will be more concentrated and can, in a worst case scenario, force AMD to lower clocks (boost mostly but possibly also base, depending on how much we're talking about.
This is one of those same tropes like "quantum tunneling will make the next shrink the last" that we've been hearing since 90nm was leading edge. If it wasn't a problem during previous node transitions it won't somehow become an insurmountable one now. Shrinking from 32nm to 14nm made a much greater difference in how concentrated the heat is.
Posted on Reply
#15
HTC
ghazi, post: 3927903, member: 175630"
This is one of those same tropes like "quantum tunneling will make the next shrink the last" that we've been hearing since 90nm was leading edge. If it wasn't a problem during previous node transitions it won't somehow become an insurmountable one now. Shrinking from 32nm to 14nm made a much greater difference in how concentrated the heat is.
Actually, no: no shrinking from 32nm to 14nm. You're forgetting previous arch was one chip while Zen is @ least 2 (not counting APUs here), which is actually a good thing since it splits the heat sources, but it still requires good cooling to be able to have good OCing.

There's a reason Zen chips have temp issues when one OCs them past roughly 4.0 GHz and Zen+ when one OCs them past roughly 4.2 GHz: there's a temp wall that requires more exotic cooling to be surpassed.

By shrinking the chip's size for Zen 2, AMD will be concentrating the heat even further: i think they'll manage to have good base / boost clocks (perhaps similar with Zen+) but the OCing ability will probably suffer.
Posted on Reply
#16
DeathtoGnomes
Camm, post: 3927809, member: 110377"
You would be surprised. My TR 1950X was easy to keep under 60c under Prime with a U12S in Push\Pull. Its the high socket size that does it IMO, that increased surface area is just a godsend.
is that stock clocks?

bug, post: 3927429, member: 157434"
I know this isn't an apples-to-apples comparison, but the hottest Pentium 4/D didn't need a 250W capable heat sink.
comparing a Threadripper to a P4... its just mind boggling and funny as fk.
Posted on Reply
#17
Camm
DeathtoGnomes, post: 3927946, member: 151150"
is that stock clocks?
P-States were increased rather than all core.
Posted on Reply
#18
ghazi
HTC, post: 3927915, member: 51238"
Actually, no: no shrinking from 32nm to 14nm. You're forgetting previous arch was one chip while Zen is @ least 2 (not counting APUs here), which is actually a good thing since it splits the heat sources, but it still requires good cooling to be able to have good OCing.

There's a reason Zen chips have temp issues when one OCs them past roughly 4.0 GHz and Zen+ when one OCs them past roughly 4.2 GHz: there's a temp wall that requires more exotic cooling to be surpassed.

By shrinking the chip's size for Zen 2, AMD will be concentrating the heat even further: i think they'll manage to have good base / boost clocks (perhaps similar with Zen+) but the OCing ability will probably suffer.
Zen OC wall is the result of the process being optimized for low-power applications. It's not thermally limited like on the Intel side where you see people easily pushing well over 5GHz, or like with Bulldozer where the process was made to gulp power. People seem to like to forget 14LPP was originally designed with Samsung's Exynos SoCs for cell phones in mind. The idea that the Zen OC wall has anything to do with heat dissipation is astounding to me when the 4C/8T die doesn't even clock as well as the dual-8C/16T MCM configuration.

I'm not sure how they can get better stock clocks but worse OCs when their chips literally do not overclock. Unless you call making the single-core turbo apply on all cores "overclocking".

Consider that Zen has 4 times as many transistors as Bulldozer in roughly half the area, and runs way cooler. Why would heat density suddenly start lowering OC potential from where it already is if you increased density by another factor of 2?
Posted on Reply
#19
Athlonite
Why use 6mm heat-pipes for criss sake just stop, Use 8mm heat-pipes and do the job properly and supply two fans for dogs sake oh and while I'm havin a rant polish the damn base would it bloody kill you to put a proper finish on it maybe even nickel plate it so you know it looks like a mirror it's not that hard and I'd pay 20 bucks more for it too

Also with regards to the TDP of this HSF it states on their website
MATTERHORN THREADRIPPER
210W TDP


so cutting it fine in the heat dissipation sector for an 2990WX I would think

but the Olymp has an 340W TDP
Posted on Reply
#20
chfrcoghlan
Athlonite, post: 3928155, member: 80893"
Why use 6mm heat-pipes for criss sake just stop, Use 8mm heat-pipes and do the job properly and supply two fans for dogs sake oh and while I'm havin a rant polish the damn base would it bloody kill you to put a proper finish on it maybe even nickel plate it so you know it looks like a mirror it's not that hard and I'd pay 20 bucks more for it too

Also with regards to the TDP of this HSF it states on their website
MATTERHORN THREADRIPPER
210W TDP


so cutting it fine in the heat dissipation sector for an 2990WX I would think

but the Olymp has an 340W TDP
The base of my Matterhorn Pure is polished, are the other's not?
Posted on Reply
#21
Athlonite
@chfrcoghlan well if you take a look at the OP's pics you'll clearly see that this ones base is not polished it maybe flat but it certainly looks like it's covered in machining marks

Posted on Reply
#22
chfrcoghlan
Athlonite, post: 3929073, member: 80893"
@chfrcoghlan well if you take a look at the OP's pics you'll clearly see that this ones base is not polished it maybe flat but it certainly looks like it's covered in machining marks


Oops, I didn't bother to check the pictures. :oops:
Posted on Reply
#23
HTC
Athlonite, post: 3929073, member: 80893"
@chfrcoghlan well if you take a look at the OP's pics you'll clearly see that this ones base is not polished it maybe flat but it certainly looks like it's covered in machining marks


Someone please correct me if i'm wrong but isn't it like this actually better?

I seemed to recall reading about it because the TIM would adhere better to cooler this way. Don't remember where i read that: sorry.
Posted on Reply
#24
bug
HTC, post: 3929283, member: 51238"
Someone please correct me if i'm wrong but isn't it like this actually better?

I seemed to recall reading about it because the TIM would adhere better to cooler this way. Don't remember where i read that: sorry.
The best transfer is metal to metal. Whatever paste you put under there you put to work around metal imperfections. This is also why you use as thin a layer of paste possible.
Posted on Reply
#25
HTC
bug, post: 3929286, member: 157434"
The best transfer is metal to metal. Whatever paste you put under there you put to work around metal imperfections. This is also why you use as thin a layer of paste possible.
If the metal in the cooler is absolutely polished and flat, won't the paste tend to get squeezed out, when tightening the cooler? From what i recall, the imperfections actually helped with that.
Posted on Reply
Add your own comment