AMD Ryzen 9 7945HX (Dragon Range) processor review
Click here to post a comment for AMD Ryzen 9 7945HX (Dragon Range) processor review on our message forum
fantaskarsef
Totally missed opportunity for MSI to build the first Dragon Range laptop and use Hilbert's catch phrase.
Texter
Can you TOUCH a dragon would also be a nice catch phrase...
nizzen
91ns memory latency is a gaming killer 😱
Horus-Anhur
That memory is really slow. Latency is extremally high and bandwidth is very low.
Is Expo even enabled?
Freeman
when talking ''laptop performance'' it would be interesting to log/note FPS Performance in the first 10minutes and after 1 hour.
Because after 1 hour it might be... very different, because all the System has been heated up.
Hilbert Hagedoorn
Administrator
Much like graphics card reviews, we warm up the gear prior to testing.
schmidtbag
Despite the memory, the CPU performs extremely well. Even if it were a desktop CPU and consumed another 50W, it would be impressive.
This is the kind of stuff why I'd rather focus on undervolting, because for 4K gaming, all your CPU needs is to just not be a bottleneck, and it really doesn't take much nowadays to do that.
Noisiv
I don't understand how is this 55W CPU? Or even 75W?
From AMD:
AMD Configurable TDP (cTDP): 55-75W
https://www.amd.com/en/products/apu/amd-ryzen-9-7945hx
From Asus:
Ryzen 9 7945HX: 55W in 'Turbo' mode; 65W in 'Manual' mode
RTX 4090M: 175W
!!On HWINFO screenshot it shows 103W Maximum Power for CPU Package!!! (core + soc)
(85W for AVERAGE, but who cares about average power which includes time spent out of maximum MT load)
Guru3D Laptop power measurements:
!! Idle: 33W, Single Thread 61W, Multithread!!: 161W !!
Yeah I get that it blinks and whatnot, but there is 130W of difference between Idle and Multithreaded.
Display ON in idle?
If it is then these 130W of difference likely comes from 90W Cpu Core + 10W CPU graphics + 30W blink-blink-brigther-display-disk-misc
Which is in line with Hwinfo showing 88W max for CPU core.
tunejunky
another fantastic review HH.
a couple of thoughts that may be pertinent...
please include noise levels like gpus when testing laptops.
my Strix Scar 15 has been superlative in every way including the noise level...and by that i mean superlatively bad.
Asus has since put a lot of work on the cooling w/ smaller and hot spotted processes but physics are physics.
also, despite folks freaked by latency isssues - this is a mobile proc and i would love to see someone do it better - literally
what i would love to see are the SFF micro-pc's like the Brix incorporate both Dragon Range and mobile 4xxx/7xxx gpus
that would sell
TLD LARS
Normal OEM latency.
A 100% stock 13900k or 12900k is in the 70-80 ns, hardly a gaming disaster.
tunejunky
ok... temporary hijacking of thread to show off my boner for rtx4090m
just between us and f.y.i. the number of 4090m chips vs 4090 chips (totally different dies) is going to average greater than 20:1
which is about the way it always is for flagship gpus. and the die itself for the mobile is higher yield than the 4090 die
can you imagine the demand for a card that rivals 4080/7900xtx @ 175 watts? that can be a double slot solution?
maybe it's because i didn't go to Stanford Business School ...
Gomez Addams
One small correction HH :
The i9-13980HX has 8 "big" cores and 16 little ones for a total of 32 threads.
Freeman
Ok, thanks
Hilbert Hagedoorn
Administrator
That's a very good idea, I'll make note of that.
Texter
The 4090m is LITERALLY a lower clocked 4080, and costs 25% more than an actual desktop behemoth 4090...
The RTX6000 Ada Generation, has the 100% functional 4090 GPU, just with a 300W (dual slot) cooler. And much lower clocks. (At $6,800.)
So the 4090 wiped its ass with the 6000...o_O
Typhon Six Six Six
Laptop DDR5 is slow. A cl 15-14 kit of ddr4 OC will get you in the 39-45ms range. You might want to disable core isolation as well.
TLD LARS
Yes this kit is slow at 91ns compared to a desktop DDR5 7800 CL38 kit at 58ns (Guru3d review), but almost slicing the latency time in half from 91ns to 58ns is only really 15% more FPS in games.
Compared to the higher requirements for the faster memory to work and design costs and 2-3 times higher power need to run, the 15% more FPS maybe has too many consequences and to little profit.
nizzen
I tested stock 12900k 4800mhz xmp vs 12900k stock with 7000mhz tuned memory last year.
Got 35% higher min fps in Battlefield.
Running 8533mhz with 13900k now.
TLD LARS
35% more min FPS is nice, but the consequences are high.
I am guessing here but you are probably using 1.4-1.5V for the memory and this laptop is using 1.1V, the power difference is something like 1-4 W depending on load.
On this laptop needing 1-4W extra power to feed and cool the faster memory, means that 1-2 cores needs to be parked on the CPU, maybe even hurting the min FPS more then what is gained by the higher memory speed.
The extra power needed for Mem is always, so the faster memory can double the idle power requirements of the CPU and memory, because memory is not slowed down at idle.
Is your 7000mhz setup stable in a 60-65 degrees C chassis like a laptop?
Laptop memory is generally half the size of desktop mem, so the heat density is twice as high on a memory module, that is mostly naked and stacked on top of eachother doubling the heat density once again for 4 times as much heat per size compared to desktop.
No airflow.
No cooling.
Krizby
Lol who made up SKU name at AMD? I thought 7945HX is some midrange mobile CPU
