The two counters (Limited and Apod) are from different point in the processing pipeline.
Apod counter is based on analysis on the input, before the data enters processing. So the processing wonāt affect itās value.
Limited counter is based on analysis of the output, last thing before data goes out to the DAC. So here processing affects the value and this one should thus always stay zero.
Well, been testing the new Windows 11 Pro,13900K build and was able to do this with no external gpu. Using Asus AI Overclocking. 990 Pro and Corsair Dominator 32gb 6000 DDR5
With convolution and peq
Sadly, still stutters using sinc-Mx PCM-DSD512 on most of the higher EC modulators, 1 stutter on DSD5EC.
Whatās strange was I got this last night before enabling AI Overclock when it was stock and played without stutters. I tried all combos of check and unchecks of Adaptive and Multi and not working tonight.
I would suggest modulator ASDM7ECv2 and DSD256 instead, not sure if that requires more or less CPU though but I am pretty sure it will sound slightly better.
You can also lessen the CPU load by using āAdaptive output rateā, but then you have to select another oversampling filter (I use ext3).
sinc-Mx and other sinc- filters are ones that benefit from GPU offload big time. Without offload, it will eat a lot of CPU time and you would need to have 12 - 16 performance cores to spare. So those may be better on the AMD CPUs that have more cores.
It is 64 million taps at DSD1024 to process, quite a lot of computations to do at almost 50 MHz sampling rate.
I have a May now and R26 is still here taking a break.
xla @1024x48 is ok with AMSDM7 512+fs
I did get AMSDM7EC 512+fs at some point and took a screenie when it was stock.
I may have been heat soaked last night too as temps were creeping up to high 60ās and had a couple blue screens the last couple sessions. I loaded defaults and started over on the AI again thinking maybe I didnāt do it right originally. I never loaded default setting and did the Cinebench test and reboot as instructed in the guide. (First time playing with AI Overclocking and only second build since 6700k many years ago.) I only enabled AI and went in and messed about and later had a blue screen, rebooted and did Cinebench test and so on and getting random blue screens before.
Weāll see how it goes in a bit with the new AI training today.
I think xla, ext3, hp@1024x48 AMSDM7EC 512+fs will be stable goal with igpu
Sorry, I forgot to answer the other question. I do convo and peq because I didnāt make it myself and the convo is from Mitch Barnett with a neutral target but I like a bit more bass. So, itās just a minor .75db bass shelf on an otherwise amazing sounding filter.
That sucker got beefy!Ā” Itās been like that since I crossed 500 hrs. to now at 800. I am intrigued and astonished. I am going to mess around more tonitght and tomorrow and see what happens.
After months of listening to immersive audio (24/48 Atmos 7.1.4, 24/96 Auro3D 11.1 and discrete 10ch or 12ch DXD) thru HQPe as DSP engine with very long taps of FIR filters for DRC+XO, so far I preferred:
First choice: everything upsampling to 12ch DSD256 using poly-sinc-ext3 (1x) / poly-sinc-ext2 (Nx) + ASDM5ECv2.
The SQ is the highest but the computerās power consumption isā¦ more than 700w (sustained). It hurts the electricity billā¦
Second choice: 48 or 96 immersive audio contents upsampling to 12ch PCM 24/352.8 (DXD) using poly-sinc-ext3 (1x) / poly-sinc-ext2 (Nx) + NS1.
Output DXD to Mergingās DAC, the data path in the equipment will be the same as DSD (other than DXD/DSD all audio data will enter the FPGAās DSP) but the power consumption will be very low. This combo has deeper bass and more relax feeling, good for long listening.
DXD will be in bit-perfect mode as same as DSD (but without some restrictions).
For example, volume adjustment for DXD will be handled by DAC chip (same as DSD), not in FPGA. Volume adjustment for 1fs to 4fs will be in FPGAās control.
This can be proved by max volume. 352.8 and DSD are 0dB but other sampling rate can up to +12dB.
So, DSD1024x48 AMSDM7EC 512+fs is not sustainable on my setup. It may have played without hiccup when the pc was cool again. Thermal throttling happens even when stable overclock in place. 139k is just too hot for my sff build in the Meshilicious and TGP with 280 Z63 Kraken cooler. I did the ai training a couple times and even just stock it stutters eventually. When AMSDM7EC 512+fs is selected at 1024, package temps are hitting 100c spikes when the stutter occurs.
I can live with AMSDM7 512+fs@1024x48 and XLA.
I also tweaked the low shelf to 90hz 2.50g and .71q and this got me glued to my seat feeling really good last night. The couple weeks at .75 sounded good. Now itās heavenly.
I will say the difference between the two isnāt much for when it does play before it stutters. Also, rates at 512 in comparison to 1024 isnāt night and day. The differences are even more minor than going from 256-512.
Iāve been listening to native source files on my May KTE in NOS today. Listening to PCM all day did make me realize something, I much prefer the sound of PCM vs DSD on the May. Native in NOS is very nice indeed, and I could most likely be very happy listening without HQP in my chain. But I went back to Sinc-Mx LNS15 (1.536M) and this is really the sweet spot for me with the May KTE.
On the May, NOS only sounds decent with 96KHz and higher files. Redbook (44.1KHz) sounds terrible. Of course, I am not a fan of the NOS sound.
I think the May KTE sounds great with HQPlayer upsampling to 32fs. But, I think that May KTE sounds best upsampling everything to DSD256 using the ADSM7ECv2 modulator.