10 comments

[ 3.4 ms ] story [ 33.1 ms ] thread
There was ample motivation for AMD to be the first manufacturer with a heterogeneous chiplet CPU, it's hugely disappointing that they didn't make the dive. A stable 5.5+gHz clock is impressive, but I would have much preferred to see that engineering effort go towards making nice efficiency cores.
Efficiency cores are a marketing ploy for apple to charge you more for faulty chips.
I will save judgement when I see the benchmarks. AIDA64 isn't the target use care for content creation users. Performance is going to IO bound and/or have limited burst of higher CPU usage that might give the water cooler time to dissipate heat.
Am I understanding right that it’s a function of thermal density? They consume less power than an equivalent intel cpu, so being more efficient but hotter seems contradictory. It seems like it’s a limit of the ability to pull the heat out of a smaller area quickly enough. If that’s the case, one thing to consider is that, while hotter, it won’t heat a room as quickly. That can be a deal breaker for people.
Yes the total area of the 2 chips that compose a Ryzen 7000 with 12 or 16 cores is almost twice smaller than that of a Raptor Lake CPU, so even at a significantly lower power consumption it is harder to cool.

This is nothing new, it has always happened with the transition to denser technologies. For example when Intel Broadwell (14 nm) has replaced Haswell (22 nm) it was also much more difficult to cool than its predecessor.

But isn’t it normal for AMD? I have a 5950x and they normally operate at 100c? I have liquid cooling too
whoa dude thats uhhhh pushing the envelope a bit... I'd check your cooling system, 95C is considering TJ max before it throttles.
You very likely have a cooling system issue. Normal temps should be closer to half that (~50c). Of course it depends on ambient temps and if you are overclocking, but 100c is not anywhere near normal.
The 6900HS in the 2022 Asus Zephyrus G14 has a default boosting behavior that keeps it at 95°C most of the time (on the user's lap). It might have a liquid metal thermal interface. Given these two samples, the original story is not surprising.
> Do note that the information shared is based on ES/QS samples so final results may vary.

Pre-release CPU samples are usually worse as far as speed, power consumption, and heat output than the ones sold to customers.