Patu

Ah, nevermind. I misread your original post. Now it makes a lot more sense :) It’s still surprising though that DSD1024 runs with 13900T but then DSD512 + correction doesn’t. I’ve had quite similar loads with DSD256 + correction and DSD512 + no correction. Thanks for testing!

Wow, these results are very surprising. I assumed 13900T would easily do DSD512 since my i5-12500 (non-T model though) with 2x8gb of DDR5 4800mhz RAM can do DSD512x48 with poly-sinc-gauss-long/poly-sinc-gauss-hires-lp and ASDM7EC-super 512+fs. i9-13900T, according to various benchmarks all over internet, should be a clearly faster CPU than i5-12500. I'm also very surprised that 14900K struggles with DSD512, it should be a walk in the park for that CPU. I'm also running DSP correction (convolution filters for L+R channels) and still DSD512x48 works quite easily. With ASDM7EC-light 512+fs my machine isn't even struggling. With -super it's close to the maximum my server can do but still no dropouts if I don't do anything else with the server machine while listening to music. With DAC correction on, I can do DSD256x48 with default filters and ADM7EC-super. I guess my Lenovo Thinkcentre m90q Tiny Gen 3 server machine works very well then. Only optimisation I've done is that I run Atlas OS version of Win11 Pro. It's a highly optimised Windows-version, mostly for gamers. In HQP settings I've set Buffer to 10ms, DSP pipelines to 2 and Blocks per cycle to 4. If blocks per cycle is a high number, like 12, I get dropouts. Now I'm little worried that DSD512x48 with DAC correction on isn't even possible without CUDA offload.

Thanks Miska. If it’s not too much trouble, maybe try if either cpu can run DSD512x48 + dac correction with default filters and -light or -super but without the help of cuda offload ;)

Nice setup. I’ve been thinking of building something similar (mini ITX case, SFF) but I’d like it to run without a separate gpu to keep the power consumption and thermals low since I’d need to place the server in a tight closed space. I assume that the DSD1024 runs with the help of gpu offload and RTX2080 in your build?

In what kind of configuration do you use the 13900T? I'm just wondering if the CPU is throttling much or can you extract its full potential. I've seen those low power T-family CPU's mainly in SFF corporate machines like Lenovo Tiny, which I use myself. My current server machine is Lenovo m90q Gen 3 Tiny with i5-12500 (regular, not the T) and DDR5 4800mhz memory. It runs DSD512x48 with the default filters and ASDM7EC-light 512+fs + matrix profile. But if I want to use the DAC correction, I need to drop to DSD256x48. I'd like to be able to run the correction on DSD512x48.

I guess we need to wait until someone measures this thing.

I wonder how the new D90 III discrete behaves with the HQP. Interesting release from Topping. https://www.toppingaudio.com/product-item/d90-iii-discrete

Awesome, I always thought that the settings export/import doesn’t affect the matrix profiles since it has its own save/load mechanism.

@Miska Sorry if this has been discussed before, but is there any chance of profiles function in future updates? Now we can save matrix profiles and export/import settings but they’re two different functions. I use different settings and matrix profile for headphones than for speakers. It’d make the switch quicker and easier if there were profiles available, which include all the settings and matrix profile etc. Even better if these complete profiles could be switched from HQP mobile application for iPad (just like you already can switch between matrix profiles). Thanks!

So the most often used CPU benchmarks on comparison websites are useless for evaluating HQP performance? That’s unfortunate. I understand that HQP’s own benchmark is the most accurate but it’s pretty useless for comparison purposes at this point since there’s no database of results online. Even if there would be, the results should be run with identical HQP settings to be comparable.

@Miska Sorry if this has been answered already, but when I compare CPU benchmarks online, which benchmark will reflect HQP performance most closely? Is it maybe Geekbench, Passmark, Cinebench or some other? Thanks!

There's something wrong with your settings. I run HQP server on Lenovo m90q Tiny Gen 3 business pc with i5-12500 and no separate GPU and it runs DSD512x48 - ASDM7EC-light 512+fs - poly-sinc-gauss-long without problems. For DAC correction, I need to drop to DSD256x48 but then I can switch to ASDM7EC-super and it still works great.

@OE333 I recently moved my dedicated headphone amp to the equipment rack. I connected it to the RCA output of DAC200. XLR outputs are connected to a preamp, which feeds my active speakers. Obviously both outputs are live simultaneously. But are there any possible negative side effects for having the both DAC200 outputs connected at the same time?

I went the other way around, I auditioned Tambaqui among some other DACs way back before I ever had used HQPlayer. Back then I actually ended up with Holo Spring 3 with the integrated preamp module, which I found more organic sounding than the other options. Tambaqui had great space and clarity though, but I demoed it connected directly to my active ATC's. Tambaquis volume control/preamp section probably wasn't the best possible and I'm sure it would've benefited of a good preamp. Anyway, only after that I introduced HQP to the setup and have used it ever since. Nowadays I have T+A DAC200 and still with HQP. I also have Holo Serene in the setup (and Holo Red as streamer). I keep Holo Serene connected every now and then and also run DAC200 directly to my active ATC's. Both options have different flavor and strengths so I keep the Serene in the equipment rack for now. It's just a bit misleading to talk about HQPlayer as one option only. There's hundreds of possible combinations possible with HQPlayer and they all sound different from eachother. You can choose either DSD or PCM, then all the rates with both of these, then filter and modulator selections... insane amount of options. I've spent a lot of time to find the best settings for my preferences. DSD1024 isn't necessary for most use cases, actually lower rates can objectively have better measurements and can sound better. It depends on the DAC used with the HQP. For my T+A for example, DSD256 is the sweet spot. This information comes from the developer of HQP himself. IIRC with Tambaqui it's recommended to use the highest possible PCM rate for input. Tambaqui oversamples everything internally anyway. Tambaqui is a great DAC and it would be interesting to demo it again with a proper preamp. Makua with the Tambaqui DAC-module would probably be the best option. Styulus audio player seems interesting. I hadn't heard about it before.

Great news! Anyone updated yet?