|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
41.183.0.21
In Reply to: RE: : RE: *Power Consumption* Core i3-530 32nm + GA-H55M-UD2H posted by jackwong96 on May 19, 2010 at 08:26:24
I find setting DRAM Vterm to 0.500 gives better results after testing 0.600 & 0.350 - you have it at 0.350.
Have you installed PCI Latency Tool? It will list USB controllers - default is 0 (not programmable). I'm not sure if USB latency can be adjusted but its worth trying.
Use following formula:PCI Latency = ASIO latency (in samples) * 2 + 2
For example, if ASIO latency is 64 samples then PCI latency should be 130. For 32 its 66 and 48 its 98. I have mine at 66 and there's good improvement.
Follow Ups:
Have you tried PCI Latency Timer at 0? I have great results with it.
This flies in the face of cics recommendation to set it at max of 128. Do you have a fully optimized cmp^2 setup?
Yes I do.
Do you have a juli@ sound card? Where do you set your buffer in cplay settings? Do you run 192 with src? Just trying to see if your parameters are similar/different from mine.
No, I don't have juli@, I have Cantatis Overture 192. I've recently flashed it with JULI@'s firmware and tried native ESI drivers. I didn't like it at all - artificial, clinical whatever, it just didn't sound like music to me. Back to VIA 5.20 drivers and cantatis overture FW. If you have a minute, try those VIA drivers (5.20 version only). They're compatible with every Envy24HT-based sound card. I'm curious about your remarks...
Since I use I2S to my Buf32s I don't use drivers. But I do damp my I2S lines with 22 Ohm resistors. I do want to hear about your cplay parameters because if you are similar to mine (192/src 145/tiny buffer) I may try different pci latency levels.
Unfortunatelly I don't use cPlay, but XXHighEnd with Kernel Streaming adaptive mode and quad arc prediction upsamling to 176 or 192kHz
Then your system is not, by definition, a cmp^2 system. Cmp^2 implies a cmp shell fully optimized, and cplay playback. Nevertheless did you try 128 also for pci latency?
OK, my system is not cmp^2 by definition, you right, at least not on a hardware site. But I've done all system optms and underclocked/undervolted as much as possible. I've tried 128 pci latency, but it produces cracks in my setup. There is no rule for optimal PCI Latency value. Not for different PC setups.
Greets,
Marcin
Not trying to debate or argue with you just trying to understand your baseline.
Thanks, I set DRAM Vterm to 0.510 and it was better.
I installed PCI Latency Tool and adjusted the USB lantency. I could not make it work properly on my USB driver (Musiland 01USD) since its ASIO buffer was not defined by no. of sample and I have no way to determine the value. Instead, I installed ASIO4all. The ASIO buffer was set at 64 samples and I set the USB latency to 130, as per your advice. It worked fine with noticable good improvement.
I tried adjusting the DRAM timings which brought about small but noticable improvement too. my setting:
5
1
1
2
5
1
1
2
5
5
15
1
5
1
2
Please try and see if SQ improvement can be acheived.
I prefer Kingston ValueRAM to HyperX (both @1GB) and ValueRAM allows for more aggressive timings.
After some testing including suggested 5-1-1-2-... I still prefer Tcl = 5 and Tcmd = 1 (with all else remaining on Auto). Problem with other timings is CPU-Z does not confirm it and shows CL5 base timings (instead of 5-1-1-2). On 5-1-1-2 sound is more strident and fatiguing.
I have the same experience with the D2H mobo where bypass cap was not added, but not so with the modified UD2H. The difference in speed can be heard. This concerns the frequency response of PWM ps of the mobo, which is too slow responding to aggressive Dram timings.
Thanks for confirming PCI Latency and Vterm.
No need to use ASIO4ALL as cPlay reports buffer in samples (see diagnostics). For example:
ASIO Driver Initializing...
Name (ASIO4ALL v2)
Player ASIO version (2)
Driver ASIO version (2)
Message (No ASIO Driver Error)
Channels (inputs: 2, outputs: 8)
Buffer details (min: 64, max: 2048, preferred: 64 , granularity: 8)
Sample rate ( 96000.0)
Output Ready? Not supported
Preparing buffers... successful
Latencies (input: 64, output: 64)
ASIO Driver initialized.
Item is reported under "Buffer details" as "preferred: ?". This is the ASIO buffer size cPlay uses. Only issue for Musiland driver is how low ASIO buffer can be set (try its ASIO Control Panel via cPlay if this can be set lower). Maximum PCI Latency is 255.
Will test DRAM Timings.
Thanks. I tried the Musiland driver and Cplay reported min.: 256. With 256 samples and 255 PCI lantency, ASIO4all was preferred. I gave up and am using ASIO4all for now.
Hopefully Musiland can offer lower latencies in future. Be sure to setup ASIO4ALL to use hardware buffers and not have any latencies of its own.
Tried DRAM Timings on Kingston HyperX DDR3-1375 1GB and unfortunmately, RAM did not work. Will be testing Kingston ValueRam DDR3-1333 1GB soon which is cheaper and doesn't "validate" timings.
Try PCI Latency Tool 2.3 (you'll need to reverse Minlogon, and only then can 3.1 be uninstalled after restarting DCOM and Windows Installer services). With 2.3 there's no installation, simply extract software (by running LtcyCfg.exe) to folder and run it. It has less bloat and doesn't suffer from loosing PCI devices (unfortunately its a bit more difficult to identify PCI device as your soundcard will show something like "Multimedia device"). Also, we can safely disable unwanted stuff in BIOS. There's more improvement doing it this way.
Note for settings to take effect in cMP mode, you need to setup a .bat file and run it each time on reboot. I use the "RIP" to apply latencies on startup:
@echo off
"c:\...\LtcyCfg.exe" /a
exit
See example for doing something similar with Cryptographic Services.
Thanks! Will try. Can the same .bat file be used to start up LtcyCfg.exe and stop Crypt Srv at the same time?
Yes. It would look like this:
@echo off
"c:\...\LtcyCfg.exe" /a
sc stop cryptsvc
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: