Quote:
Originally Posted by
GoldDragon
β‘οΈ
I would love someone (scanproaudio?) to test the new 14th gen with both DDR4 and DDR5. I wonder if the memory controller improvements would help DDR4 performance?
The difference between 4 and 5 for virtual instrument counts was very large and unexpected, almost suggesting a bug.
I am 90% certain to upgrade my 12700K to the 14700K.
My i9-14900K seems to have a better IMC than the i9-13900K yes also for DDR4.
I initially ran my i9-13900K with the 3200CL16 memory I had when I built my system during 12th gen launch. I recently upgraded it to a 4000CL18 set, but didn't get it stable initially in Gear1 (or possibly it was my CPU OC), but it ran fine at 3800CL17.
Now with the i9-14900K, the 4000CL18 XMP runs fine in Gear1.
Regarding the DAWBench report, I'm seeing slightly larger gains, although I agree it's a very small nearly meh update. FWIW the i9-14 runs cooler than the i9-13 with the same power use reported - my NH-D15 managed 220-240W on the i9-13, but now I've seen 270W and even more on some tests.
I did a very quck initial OC to mimic my i9-13 setup and was dissapointed at first (only 100MHz up accross all cores), but now with a little use I think my upgrade was worth it, and anyway as a plus the i9-13 can replace the i9-12900K in my home/gaming/spare system.
But also interesting findings related to RAM speed and/or timings effect on DSP results on lower buffers, and definitely an increase in the "snappyness" of huge Pro Tools sessions (even with no plugins and very low CPU load). Keep in mind that my upgrade also results in a latency reduction for every memory operation, from 10ns to around 9ns (google for the infos and calculations)
Keep in mind these are cooled with the Noctua NH-D15 which surely limits the multicore especially on the i9-14. I'm currently starting a very large high pressure mix (Atmos film score in 96kHz) and needed the new chip to be "good enough", will tinker more later.
These are my own "Pro Tools DAWBench Beta" sessions originally built while I helped Vin on the PT sessions (I prefer testing changes with these, as PT is my DAW and it's actually faster to test with the test either passing or PT reporting a CPU error, no audible glitches ever) with TDR Kotelnikovs. HDX as the playback engine, PT Ultimate "Hybrid" (Native) mixer.
Results, the i9-14 ones are preliminary and quickly done so not "squeezed", will need to work now.
i9-13900K 3200MHz CL16 DDR4 (Gear1) RAM
64 buffer - 920 TDR (fixed)
1024 buffer - 1042 TDR (possibly on a older PT version so could be very slightly too small)
i9-13900K 3800MHz CL17 (17-21-21-21) DDR4 (Gear1)
64 buffer - 946 TDR
i9-14900K 4000MHz CL18 (18-22-22-22) DDR4 (Gear1)
64 buffer - 1010 TDR
512 buffer - 1060 TDR
1024 buffer - 1080 TDR
Would be interesting to swap my mobo and upgrade to very low CL DDR5 RAM and see how the results would compare, but I think I'll stay on plan and build a new system for 15th gen.
Edit:
The i9-13900K OC was x58 on 2 P-cores, x56 on the rest, and x45 on the E-cores.
The i9-14900K OC I now have is x59 on 2 P's, x57 on the rest and x46 on the E's (I didn't get the 2 P-cores to x60 with this all core OC yet, but will investigate after this mix...)
Edit2: Found an updated result for the original i9-13 result with the slower RAM, from 916 to 920 at 64 on PT 2023, seems that PT updates have changed these a bit as well. the 2 later result sets are the same PT version, 2023.9 (but I didn't find the i9-13 with faster RAM and larger buffer result, I just remember that it was nearly zero gains, so the RAM effect on DSP seems to be mostly on small buffers).