DIY USB DAC v2:

The cover isn't complete yet though. Supposed to have a transparent top cover also.

On hindsight maybe I should've gotten a better-looking card case.

Oh yar, and the extra space at the right is for future expansion for 3.5mm. :D

And now I understand why GoVibe Petite is so overpriced. Coz it takes skill to design a case small enough to fit it just right.

But still, GoVibe Petite is still overpriced. :D

I've joined the dark side...

...the dark side in this case being USB DAC users.

Originally I wanted to post a sneak peek when I first got it mailed to me, but decided not to spend time on a useless post. Time was precious during the past week and it's against my policy to post needlessly.

So here he have the DAC complete. It is based on the Burr-Brown PCM2702 USB DAC (USB audio interface w/built-in DAC, the same thing powering the FubarII/III and GoVibe Petite DAC and various cheaper entry-level DACs. (Some other DACs also use this series of DACs for the USB interface, but uses I2S output to another better DAC, e.g. PCM1796 or CS4398 with higher sampling rates and SNR)

Specification-wise, The PCM2702 isn't much of an upgrade over normal sound cards, with a 105dB SNR and a sampling rate of only 16-bit/48kHz. At least it's still better than it's cheaper brother PCM2704 with 98dB SNR. Speaking of which, the PCM2707 also has 98dB SNR, but most people use it for the I2S/SPDIF to another DAC, so SNR isn't of concern here.

I used to ask this question: considering that sound cards have much more and better components than a USB DAC, why do everybody say that USB DAC sounds better?

Now, I have the answer right in my hand. Towards the left on the table to be exact.

I wasn't expecting such a huge difference for a 5dB increase in SNR. But, the quality of the music really blew me away. The most obvious being WAY, WAY, WAY BETTER highs response. The highest notes (those tlink tlink type instruments) sounded very clear, the piano is also greatly affected. The soundstage is also way bigger with seperation between instruments and much higher clarity as well as more "space". Those sound that I couldn't hear/groups of sound that were garbled could be heard better now (but still not flawless).

Bass also improved, and not just louder and more boomy - it's tighter, and the higher bass where all the bass guitar action lies is louder. I can also hear the hitting of keys/strumming of strings better. Still, it's louder, louder enough that I went to check my subwoofer level. Wierd, it's at the lower level, so why does it sound so close to my higher level?
Plus, combined with the bigger soundstage and "space", some of the more important bass I can hear it boom and reverb more nicely. Nice impact.

But there's one thing that's stopping from calling it a total all-rounded improvement - the sound signature, as opposed to other sound cards/on-board, MP3 players and my PSP. Ignoring the Zen Stone that's bass heavy, the PCM2702 sounds more natural yet more harsh. I'm not talking about the brightness of the sound since this can be altered by changing the capacitor values, though this is a bit on the bright side. Wierd... more natural and harsh at the same time. Is it due to better accuracy or my poor soldering skills?

Talking about soldering, I really sucked at it. Lots of blobs, a few re-solders (because I soldered wrongly). Also I never used any expensive parts, only generic wires and solder. Quite a waste when I'm using WIMA caps and audiophile-quality RCA jacks.

And the most lol part about this project was when I plugged it to my com and yet nothing was detected. I though I have failed and wasted fifty-five bucks and lots of time. Then I realized I never plug in the jumpers for the power. Once that was done, everything was nice, and I was awed by the sound.

Worth every buck, and with this I've joined the dark side, my first step into dangerous territory.

One more pic of it in action

Now all I need to finish this up is cutting the holes on the candy box for the connectors. Then the "Candy Box DAC" will be complete.

And one thing about this DAC, when sound passes through it the red led on it lights up. For what I wonder, but it's kinda cute. :)

X1950 Pro Overclocking - 680/1600

Today (12/04/08) I decided to give it a shot at overclocking my (new?) X1950 Pro

Yes, I know it's an old card and is less than half the speed of current high-end cards, but that doesn't mean that I cannot overclock it right? I'm just six months/a year behind, but I too need the experience.

So here I am, firing up ATT and doing some simply overclocking. I read on forums that this card maxes out at around 660+mhz, so I should hit around that speed too.

So, start with 630mhz. No problem. 640, no problemo. 645, np. 650, np.

Have to start to be careful here. 654? Ok. 658, nothing wrong. 662? Easy.

Seems like 660mhz is easy for my card. I think I'll leave it to the auto-find-max. But that was a mistake. Coz it immediately jumped to 668, then 675, then 682, where it rendered (to my amazement) for a little while before hanging and blacking out.

But, if it manages 682mhz for that little while, it might be able to do 680. That's a nice stopping point.

So I tried. Ran for quite a while now, max temperature reached. Scan for artifacts, no artifacts found.

Woot, looks like I got myself quite a good card.

And the temperatures of this card isn't bad too. While normally it does 45~50 idling depending on room temperature (usually 46) and ~61 degrees load, it rises sharply too 75 degrees at 680mhz. 75 is still considered ok and even low for high-end overclocked cards, but that's because the normal temps are too low to start with. And that's considering that my cooling sux.

Ok, maybe AC Accelero isn't anywhere near sux, but my case is. Those kind of minimalistic silent pc with only the PSU fan spinning.

Lets test my ram now.

Unfortunately, my ram wasn't exactly the best. I set it at 800mhz (1600mhz) since it's around what everybody's getting. The card works at 680/1600 for quite a while in UT3 before showing checkerboxing - sign of memory failing. It didn't hang.

The fact that it took some time (and my core was at 75 degrees, with the surrounding not too far behind) suggests that it failed due to heat. And that's totally understandable; due to the Accelero, most of the ram chips are covered by it with little exposure to airflow. Eventually I'll be modding that cooler with a bigger, quieter fan, and I'll make sure the ram get the extra cooling from the bigger fan.

So, current highest stable, lets say 650/1500 for now. It was totally stable since I gave it so much distance from the maximum of 680/1600, which is just going to go higher.

I sure got myself a damn good card. With 1.5yr warranty left to boot, so I'm totally not afraid to spoil it and get a newer card in exchange. :P

But of course, I'm not going to run it at such high speeds. Default or lower is enough for me. Since I love silence and my CPU/RAM is bottlenecking (but not by a lot) anyway.

After gfx card purchase - Review & Reflection

Why is it that usually after you hurriedly buy something because you thought it is a good deal you will find a better deal somewhere else only a short while later? That has been the case for my previous 3 hoots as far as I can remember correctly - x800GT (too hot and no SM3.0 and not yet powerful enough), X-mini capsule speaker (because other places sell it for $35, just 2-3 bucks more, there are probably other slightly bigger ones that sound nicer, a stereo version is available soon, and I didn't use it enough), and the most recent and going to blog about, x1950pro (right after I closed the deal I found another guy selling the 512MB version for 10 bucks more (or 10% more). This extra ram will come in handy in newer, more ram intensive situations where the performance difference can be more than 20%)

The third one is easy to swallow when I simply tell myself that if something is too good it's probably been taken. Or it's just too good (to be true), get what I mean?

But lets talk about whether getting the x1950pro is a right choice when the new 9600GT and 8800GT that's owning the entire industry as the highest end gaming cards with obviously superior performance are only selling for $240-$210 lowest, new and 2nd-hand respectively. Plus, their performance in certain newer games that is more than 3x as fast really don't justify their cheap price.

And yes, in a whole majority lot of situations, the 8800GT is almost twice as fast as the x1950pro.

Actually, that kind of makes the x1950pro worth the buy.

Why is it so?

Firstly, we all know that graphics card performance increases so greatly between generations such that the newer gen cards are always more worth the money. Even more so for the 9600GT and 8800GT which are probably the best in modern graphics card history.

However, to get this benifit, that means you will have to pay the extra 2x price to get this "economy of scale". If you can do that with a $100 card, why not? It's the same case with HDDs also.

Secondly, current gfx situation is that, even if you have just a 7900GT or 8600GT or certain x1000 series and above, you'll still be able to play many games at max settings without being unplayable, just not so smooth framerates and low res./AA/AF. Seriously, I see no point with 1600x1200 and above resolutions, simply because my monitor isn't that big and it never will be. Anything above 2x AA yields considerably less benefit and anything above 4x is almost indistinguishable.
And so why get a card that can play UT3 at 80fps when I'm already fragging people at 30?

Thirdly, and rather important, it has the AC cooler. I mean, that saves me from some assembling of the Zalman fanless. And this is definitely way stronger if I were to use my own 80mm fan on it.

And finally, lets look at the alternatives - a 7600GT and a new HD3850, now price lowered to $200 with a better cooler.

The HD3850 is still not worth the price even with the better cooler, its performance at $209 just sux so much to the 9600GT one shouldn't even be considering.

The 7600GT, while it would've been the card of choice not long ago, just couldn't cut it for the newer games that are more intensive on shaders. In certain situations you see it losing to the 8600GT and x1950pro by a factor of half (meaning half as fast if my maths/grammar isn't good). That really doesn't make it worth buying.

So after some thoughts, the winner is still the x1950pro. Plus, the extra heat doesn't really matter, since with that cooler and that it only runs 7W hotter idling makes little of a difference. But, the big problem is still the lack of ICM/ICC profiles in overlay mode for games/videos, and I don't think it will ever be implemented. Guess I'll have to live with it.

Enough talking, lets fire up my x1950pro. :P

Going to buy a gfx card, so doing some research, some thoughts...

Going to buy a gfx card for two reasons. 1. My CPU isn't keeping up with 1080P videos without hardware H.264 acceleration and 2. I'm so late in playing UT3.

And with the advent of G92-based cards that brought prices of current cards down a lot, it's time to source for a cheap 2nd-hand card that's good. And must be able to play UT3 with decent speed/image quality. And still, it must be low-power, as I'm a pc-silencer. :)

While the x1950pro has the best bang-for-buck since its launch, ATi cards' high heat and lack of overlay color profiles (needed for accurate color in games/videos if you have a calibrator, which I do) turn me off. Same goes for the x1900GT. It's sad, since these two cards have very good price/performance ratio and are just strong enough to keep up with current games.

The 7600GT is just the opposite - low heat, overlay color profiles, less price/performance (but not by much, since it's also one of the best during its time), and almost barely unable to play UT3.

And the problem? Shaders. 7600GT era was when people didn't care too much about shaders. x1000 series changed that, as with 8000 series. And as such, 8600GT's performance in UT3 is so spectacular it was unbelivable - it is about twice as fast as a 7600GT and faster than a 7900GT.

Well, that's at very high settings though. At lower settings, the 8600GT's weakness - lack in everything else, should show.

Now this brings a very interesting situation. Since we all know that the 8600GT is crap for price-to-performance compared to the rest of the cards, especially 8800GT and 9600GT, the latter priced at only $239.

If we were to have an Unreal-per-dollar chart, the 7600GT ($70) would be 0.27, 8600GT ($100) would be 0.39, and a NEW 9600GT ($239) would be around 0.42.

Seriously, the 9600GT is one heck of a good card.

Now I'm confused, should I buy the 7600GT or wait till people start selling 2nd-hand 9600GT at $100+? That'd depend on whether that 7600GT is able to hit 800mhz mem, since I told the seller I'll take his card if it so. Well, at least with 800mhz the performance won't be too bad, since we know that 7600GT is memory-bottlenecked.