Review and Measurements of Benchmark DAC3

[ripvw]

It would be interesting if you got hold of the Benchmark Dac1 and Dac2 to make comparisons

I have a Benchmark DAC-1 and recently sent it in to Benchmark to replace the volume pot. They are shipping it back to me and it should arrive next week. I've already informed Amir that I can send it to him whenever he is ready to test it. Understand that this is the original DAC-1 (over 11 years old) and does not have USB input. If someone has a DAC-1 USB that may prove to be a better unit to test...

 

[Palladium]

Well, I didn't expect to see another sacred cow dying by the hands of Amir.

Sure the results are still very respectable, but I also don't find it unreasonable to expect a $2000+ kit to perform stellar in every single test imaginable in a market where diminishing returns kicks in extremely hard starting at $0.

 

[John_Siau]

Next, let's look at Jitter and Noise:
View attachment 13385


As always, comments, correction and criticism are welcome.

Amirm,

The sidebands shown in the above test are due to the use of ASIO4All. This "ASIO" driver is not a true ASIO driver and is not a Benchmark product nor is it recommend by Benchmark. Unlike a true ASIO driver, ASIO4ALL does not bypass the operating system's audio system. Software-based sample rate conversion is the cause of the sidebands. They are not being produced by the DAC3.

Benchmark provides a downloadable driver, but this Benchmark driver is not needed with Apple operating systems nor is it needed with newer versions of Windows 10.

The driver is available here:

https://benchmarkmedia.com/pages/drivers

Unfortunately the ASIO4ALL driver has corrupted many of the measurements in this review. The tests should be rerun without using ASIO4ALL.

 

[gvl]

While I agree ASIO4ALL should be avoided in these experiments if at all possible I thought ASIO4ALL used kernel-streaming which does bypass OS mixer, it is a hack that allows to write directly to Windows audio driver through ASIO interface.

 

[John_Siau]

Here is the comparison of unbalanced output with USB vs Toslink. Sadly the same spikes remain:

View attachment 13390

These sidebands are caused by ASIO4ALL. They are not produced by the Benchmark DAC3. The Benchmark ASIO driver should have been used for this test. See Graph 3 on page 48 of the Benchmark DAC3 manual. This test is almost identical except that the test tone is 10 kHz instead of 12 kHz. There are absolutely no sidebands shown in Graph 3. Likewise, there are no sidebands produced in Graph 16 on page 61 when jitter was added to the digital interface. The DAC3 absolutely does not produce jitter-induced sidebands, even under the extreme test conditions outlined on page 61 of the manual.

 

[John_Siau]

While I agree ASIO4ALL should be avoided in these experiments if at all possible I thought ASIO4ALL used kernel-streaming which does bypass OS mixer, it is a hack that allows to write directly to Windows audio driver through ASIO interface.

ASIO4ALL causes problems when the OS sample rate and output sample rates are different. There is no reason to ever use ASIO4ALL with a Benchmark DAC2 or DAC3 converter. It will only cause problems such as those shown in this review.

 

[gvl]

ASIO4ALL causes problems when the OS sample rate and output sample rates are different. There is no reason to ever use ASIO4ALL with a Benchmark DAC2 or DAC3 converter. It will only cause problems such as those shown in this review.

It's been a while since I touched ASIO4ALL, but if memory serves there is no OS sample rate there. You're probably confusing it with Windows Sound device settings where you can set audio format and Windows will indeed resample to that format all audio when device works in shared mode. ASIO4ALL bypasses that stack and writes to the audio device driver directly, there should be no OS re-sampling taking place during this process. Like I said it's been a while since I looked at it so if you have any hard data to disprove it please shoot.

 

[amirm]

Unfortunately the ASIO4ALL driver has corrupted many of the measurements in this review. The tests should be rerun without using ASIO4ALL.

First warm, welcome to the forum John. We definitely could use your expertise and wisdom here. I upgraded your title to appropriate one.

As for ASIO4ALL, I have confirmed that it absolutely is pass-through mode Unless the checkbox is made in its control panel, it won't touch the bits.

Here is an example run just now with Topping D50 with ASIO4ALL and its native USB driver:

As you see, it creates nothing of note here.

 

[rebbiputzmaker]

First warm, welcome to the forum John. We definitely could use your expertise and wisdom here. I upgraded your title to appropriate one.

As for ASIO4ALL, I have confirmed that it absolutely is pass-through mode Unless the checkbox is made in its control panel, it won't touch the bits.

Here is an example run just now with Topping D50 with ASIO4ALL and its native USB driver:

View attachment 13983

As you see, it creates nothing of note here.

Why do you think Benchmarks own results differ from yours?

 

[Blumlein 88]

I too have had issues with ASIO4ALL vs the correct ASIO driver. I only use ASIO4ALL if I don't have a driver and can't get a device to work any other way.

 

[amirm]

I have had it crash or not work/recognize a device. I have not seen it generate different results than bit-exact. The only difference is that hardware volume control in the control panel needs to be set to 100.

 

[gvl]

I too have had issues with ASIO4ALL vs the correct ASIO driver. I only use ASIO4ALL if I don't have a driver and can't get a device to work any other way.

Afaik ASIO4ALL was the only (easy) way to get bit-perfect out for devices that didn't have ASIO drivers before Windows 7 came along.

 

[gvl]

First warm, welcome to the forum John. We definitely could use your expertise and wisdom here. I upgraded your title to appropriate one.

As for ASIO4ALL, I have confirmed that it absolutely is pass-through mode Unless the checkbox is made in its control panel, it won't touch the bits.

Here is an example run just now with Topping D50 with ASIO4ALL and its native USB driver:

View attachment 13983

As you see, it creates nothing of note here.

Curious if this SPDIF test went through ASIO4ALL?

https://audiosciencereview.com/foru...asurements-of-benchmark-dac3.3545/#post-85548

 

[Blumlein 88]

I have had it crash or not work/recognize a device. I have not seen it generate different results than bit-exact. The only difference is that hardware volume control in the control panel needs to be set to 100.

When I had most issues it may have been from using an older slower computer. I haven't seen as many issues (but also mostly don't use ASIO4ALL) with a more recent computer. I did have a case where a device was putting out 16 bit results being fed 24 bit data that I never figured out. In the end I decided figuring out ASIO4ALL issues were best handled by not using ASIO4ALL.

Could it cause sidebands being discussed, I don't know, but I'd try the test with the proper ASIO to see.

 

[gvl]

@amirm , can't you use WASAPI exclusive mode in your tests if you dislike installing drivers that much?

 

[Grave]

high-end DAC which start at $10K and go up.

What DAC costs $10,000 or more? Also, why would anyone spend $1,000+ or $10,000+ on a DAC when cheap DAC's are audibly transparent?

 

[mindbomb]

@amirm , can't you use WASAPI exclusive mode in your tests if you dislike installing drivers that much?

That's also something I was wondering, but because I feel that is the more commonly used interface

 

[amirm]

Why do you think Benchmarks own results differ from yours?

Say, when you robbed that bank last night, did you take all the money or just some of it?

Nothing has been shown to be different.

The measurements Benchmark has published are for the most part different than the ones I performed. Since I don't have the unit, I can't try to duplicate their measurements. Maybe John can try to do the reverse.

For now, there is one common measurement albeit, with different scales. I adjusted for that and here it is (THD+N vs Frequency):

Ignore the red line in theirs as that is distortion without noise (mine is with noise). As you see, the measurements are very similar. Theirs is just a dB or two better but I have that much variation between channels. And signal bandwidth is also different (mine goes to 90 kHz and I think John used 80 kHz above).

So until John comes back with different measurements, there is no conflict to resolve.

 

[amirm]

@amirm , can't you use WASAPI exclusive mode in your tests if you dislike installing drivers that much?

I wish I could. Audio Precision analyzer only supports ASIO. I almost did not buy the unit because of that. I complained about that and having to use ASIO4ALL and they said they use it all the time and it is fine. Ironically with a Benchmark DAC1 no less. So I got an evaluation unit, tested the same and accepted to go with it.

Longer term I hope to convince them to add support for WASAPI.

 

[gvl]

I wish I could. Audio Precision analyzer only supports ASIO. I almost did not buy the unit because of that. I complained about that and having to use ASIO4ALL and they said they use it all the time and it is fine. Ironically with a Benchmark DAC1 no less. So I got an evaluation unit, tested the same and accepted to go with it.

Longer term I hope to convince them to add support for WASAPI.

If enough people complain they will add it. The trouble is people who buy AP have ASIO drivers for their gear already