I’d guess 28w sustained because that’s what the heatsink is designed to handle, and maybe 45-55w max boost. It really depends on how framework configures it.
1 Like
Still 28W after PTM7958? I repaste my 7840U using PTM7950 and I got 41W sustained. I hope ryzenadj still works with Ryzen AI 300
thermal change! like the direction. This should be in more devices. All I need is the bios to unlock the TDP with warranty void so someone can create a custom thermal solution. These boards go into after life projects once they upgrade to the newer board anyways. why not allow TDP limits to be higher for that very reason.
This doesn’t affect warranty I’m pretty sure and there’s nothing to “unlock”
The radiator can dissipate much more heat than 28W specified, the bottleneck is the thermal interface. All these “high temp at low load”, “uneven CPU thermals”, etc etc are the result of (relatively) low conductivity from CPU to heat pipe. At a given CPU temperature, the hotter the radiator (heated by the CPU)the better the cooling efficiency.
I’m curious, was CAMM considered for the AI300 FW13 mainboard?
The CEO already answered that doing CAMM for the FD-128GB wasn’t possible due to how signals are routed. You do need two 128 bit camm modules to serve a 256 bit quad channel memory controller and it and it makes sense to me a dual CAMM configuration would have many additional issues,
The FW13 mainboard has a 128 bit dual channel SODIMM memory configuration.
The AI9-HX-370 chip lists DDR5-5600 (sodimm) and DDR5X-8000 (soldered) as speed
Crucial lists DDR5X-7500 CAMM2 modules
A CAMM modules has a 128 bit interface, so a single CAMM module replaces two 64 bit SODIMM modules, and the standard was specifically made to improve performance over SODIMM.
I think CAMM would make lots of sense for FW13. Assuming 7500 is achievable with the layout and routing, it would be a lot faster than 5600 while still keeping the modularity.
I’m curious if that’s a discussion Framework held with AMD and was unfeasible as well, or if it’s something being considered.
my 7640u is locked at 35w max boost 
I can’t make any guarantees but that is what the heatsink and chassis are designed to handle.
Potentially, however, because framework likes to ues the same components for as long as they can (one example being them using ddr4 for the 12th and 13th gen mainboards over ddr5), they chose a SODIMM. Again, they will continue to do this until there is a large change in the memory standard that makes all DDR5 SODIMMs incompatible with whatever platform they are using.
I’m sure they’d switch immediately, once LPCAMM2 modules become supported and widely available.
2 Likes
The specs page lists the power profiles for the new AI 300 series parts, and the highest profile is 30W sustained with a 35W boost.
1 Like
I hope Smokeless_UMAF works on these mainboards to get more power
1 Like
I’m skeptical that the cooling could keep up with it. My 12th gen framework’s cooling can’t keep it at the default balanced power profile for longer than ~20 minutes of Skyrim, and that’s after installing PTM7950. After that, framerates periodically drop to a slide show as it thermal throttles.
These new chips will surely get more frames from the same power draw, but I doubt the cooling setup will support a significantly higher power draw than the last gens. It’s the same chassis after all. Trying to buff the power draw in sustained workloads will probably just cause unstable performance when it repeatedly hits its thermal limit.
It could, but it would be a macbook situation. My FW can sustain 35w at 75-80c so there is a bit of headroom for perhaps up to 40w
Oh awesome, what generation do you have? Maybe the 12th gen got the worst of it then.
EDIT: Oh my bad I see you mentioned you have one of the AMD boards, sweet!
it has a lot to do with the coldplate/cooler design which has been improved quite a bit since 12th gen came out iirc, so it should be able to hold 40w sustained
My 7840u can do 42W indefinitely with liquid metal (ptm is basically just as good at least up to 35W so lm was probably not worth it but had to try it to learn that XD) in 25C ambient temperatures. That is with unobstructed air intake of course, if you put it on a blanket the cooling performance falls off a cliff.
I am kinda hoping the new heatsink will perform even a little better (and maybe even fit my 7940u XD).
from what ive seen at least the 7640u is capped at 35w sustained, so it can’t really draw more than that
You can probably bypass that using smokeless. Given how much power the 7840u can draw before saturating the thermal solution I doubt just removing 2 cores and a few gpu cus would physicall limit it to 35W
Does anyone cap the power at 15W in order to help with thermals/fan noise & battery life?
How much does it help, and how much of a tradeoff is it performance-wise?
Is it the same as running the laptop on power saver mode?
Thanks
whats smokeless? i tried to increase power limit with ryzenadj and it didnt do anything
I don’t do it a lot but I have played around with it.
The help bit is quite subjective but it certainly reduces the performance by quite a bit. However you do get the best performance per W around that power range and there is still a lot of it available, those zen4 cores are pretty beastly. Going much lower you start getting massively diminished returns.
I did a bunch of testing on that when I first got it.
this: GitHub - DavidS95/Smokeless_UMAF
ryzenad only lets you increase stuff to the bios limit, smokeless lets you adjust the bios limit. Keep in mind the ec keeps resetting the soft limits so your ryzenad changes will only stick till the next ec override (those usually happen when you uplug/plug in power).