What's new

DAWbench

Guru3D has the 7000 series and 13000 series idling within ~5W of one another, while total consumption was between 67W and 77W. No clear winner.

ComputerBase has them idling within a range of 14W, for a total of 56W-70W, with Intel 13000 being ahead.

Self reported package power is much lower for both, with Intel reporting lower Ws. I believe the difference between Guru3D's and ComputerBase's results is likely due to the choice of motherboard (I see a spread of 30W between Z690 mobos).

There's some eyebrow raising news about 'spikes of up to 100W during idle' on the 7000 series.

Most publications seem to agree that the 7000 series is more efficient under most workloads, including light load and full tilt.
 
Most publications seem to agree that the 7000 series is more efficient under most workloads, including light load and full tilt.
I consider this reliable enough.
 
Friend of mine has Ryzen 9 7900 (eco version) 12 cores. Cinebench something above 90W but in idle CPU never goes under 40W. Idle means completely nothing is running, just system itself and HWInfo. Same setup with 13600K (stock) 14 cores, 10-12W idle. Ryzen 7900 eats 400-500Wh per day for nothing but keeping PC on.
 
What buffer size are most folks (here specifically) using nowadays?
It is in the left column, in samples and for certain specific (low latency) card, I think it is their reference RME. You can see the results grouped by values 64, 128, 256 & 512 samples.
 
With some fairly simple tweaking, you can manage to have an Intel 13900K/F pulling under 200W for the same multithreading power (under Cinebench R23 load, scoring around 38500 pts; 170W for 32500 pts with HT off). That reduces the temperature by a huge chunk, and your machine becomes dead silent.

No charts showcase that sadly.
 
Friend of mine has Ryzen 9 7900 (eco version) 12 cores. Cinebench something above 90W but in idle CPU never goes under 40W. Idle means completely nothing is running, just system itself and HWInfo. Same setup with 13600K (stock) 14 cores, 10-12W idle. Ryzen 7900 eats 400-500Wh per day for nothing but keeping PC on.
Are you sure those 10-12W aren't CPU package power only? 6-10W at idle is what TweakTown's review samples self-reported in their 13900/13600 review (with AMD 7000 CPUs self-reporting 18-24W). The lowest I can find for system power consumption at idle with >=6 core desktop CPUs is ~55W measured at the wall. I've looked at a dozen reviews now and they all say the same. For the same class of CPU, the spread at idle between AMD and Intel at the wall is 14W or less - but maybe some publication out there had different results.
 
I purchased 13600KF yesterday and overclocked it by 200MHz.
My idle is less than 10W and in Cinebench R23 is around 158W, temp max. 64C. (room 24C)
I bought it against 13700KF as I was afraid of very high power consumption and temperatures. But it definitely does not look that bad. Cinebench score 24443 pts in multicore.

Here some pictures

Idle:
Idle.png

Cinebench R23:
R23.png

Geekbench result:
geekbench.png
 
I purchased 13600KF yesterday and overclocked it by 200MHz.
My idle is less than 10W and in Cinebench R23 is around 158W, temp max. 64C. (room 24C)
I bought it against 13700KF as I was afraid of very high power consumption and temperatures. But it definitely does not look that bad. Cinebench score 24443 pts in multicore.

Here some pictures

Idle:
Idle.png

Cinebench R23:
R23.png

Geekbench result:
geekbench.png
Looks like a solid machine! Congrats!
 
What buffer size are most folks (here specifically) using nowadays?
128 for playing in gives me about 3.5ms on the Babyface Pro FS. I don't feel any lag when playing piano so I don't bother trying to push 64. Mixing I go to 1024.
 
Just adobe, but there are many threads on the internet that confirm lower idle power consumption on intels in general
The more urgent thing to address is undervolting 13th Gen processors which several of us including the OP have posted links about. Some motherboards just simply push too much voltage to the CPU in the current Auto config that comes default in the BIOS. Testing temps at 100% CPU is pretty crucial... I can idle all day on Auto at 40C, but under load, your BIOS may need to be tweaked. Here's another example of someone doing that on an MSI board.

So here's a question thatI've not tested and don't know: when you undervolt your CPU so the BIOS doesn't cook it, does that lower power consumption?
 
Testing i7-13700KF with no overclocking. 25% faster but 50% more pawer draw in R23. So in my book i5-13600KF is the sweet spot by any margin. i7-3700KF idle is impressive imho, 8W...

13700KF Idle.png

13700KF R23.png
 

Attachments

  • R23 Score.png
    R23 Score.png
    31.1 KB · Views: 7
In general current technology is pretty boring. 13900k is only 2.3x faster in single core performance than 4770k, which is very weak result for 9 years of development.
 
In general current technology is pretty boring. 13900k is only 2.3x faster in single core performance than 4770k, which is very weak result for 9 years of development.
It’s more open ended than that. 2.3x single core in that time frame is not too shabby on its own but there are also tremendously more cores. The consumer (non “pro”) chips are essentially workstation chips. Think of the cost of a 2013 xeon workstation compared to a 13900k setup and it becomes much more impressive. 4 cores was the BEST you could get for a long time without going xeon, while right now you could get a 16 core 5950x for $500 that absolutely dominates everything that came before it. Even the $130 13100 is twice as fast as the 4770k.

I also think that software optimization is in need of a paradigm shift and that there is more performance to be gained in that realm.
 
It’s more open ended than that. 2.3x single core in that time frame is not too shabby on its own but there are also tremendously more cores. The consumer (non “pro”) chips are essentially workstation chips. Think of the cost of a 2013 xeon workstation compared to a 13900k setup and it becomes much more impressive. 4 cores was the BEST you could get for a long time without going xeon, while right now you could get a 16 core 5950x for $500 that absolutely dominates everything that came before it. Even the $130 13100 is twice as fast as the 4770k.

I also think that software optimization is in need of a paradigm shift and that there is more performance to be gained in that realm.
But multicore performance is not an objective indicator as in DAW world, you barely use more than 30% of multicore performance before audio engine overload
 
Top Bottom