Obtaining VIA Datasheets?

driv3l asks: "This is totally frustrating! I'm trying to write some drivers for my Epia M10000 with a CLE266 chipset (VT8235 southbridge and VT8623 northbridge). The problem is that I can't get datasheets or info on any of the Via chipsets. I have a driver that works on my Intel ICH4 (knock Intel as much as you want... at least they're forthcoming with their technology). The Via datasheets request page does not work, and from what I've heard it's damn near impossible to get datasheets from them. I dislike looking through other people's source code, so looking through Linux drivers is counterproductive for me, especially since the driver I'm developing is not for Linux...or Windows, for that matter. This project is on its last legs due to the frustration I have been experiencing. Does anyone have any suggestions on how I can solve this problem or how I can go about obtaining the necessary datasheets?"

Comments

+5: Obvious

[rylin]

Try calling them?
If you want it really bad (like it seems you do), try being social. It really doesn't hurt to talk to people every once in a while.

Re:+5: Obvious

[AllUsernamesAreGone]

You have looked at their contact numbers right? No technical contact numbers, the only numbers they show are sales and PR (and I'm sure they'd have the faintest clue what datasheets are...) viaarena, the "technical support" site is virtually useless and has no contact numbers.

Re:+5: Obvious

[jsfetzik]

Call the sales number. Almost all tech sales people know what a datahseet is and how to get them. They may not be able to decipher them, but they know how to get them.

Re:+5: Obvious

[Milican]

Well, the thing is damn near every chip manufacturer releases datasheets on their chips. That includes the obvious ones like Intel, Motorola, IBM, Texas Instruments, Fairchild Semiconductor, Microchip, Samsung, Micron, etc... and those are just off the top of my head. I used to have several 3" binders full of datasheeets printed 8 pages per sheet (4 pages front, 4 pages back) per project from other companies whose products we used in our designs. These datasheets were our bibles for the design. For VIA not to release datasheets keeps them from being put in embedded designs, new designs of motherboards, etc. Below is a copy of their policy on datasheets... quite crappy I might add.

"Datasheets are typically only supplied to development partners and large customers under a non disclosure agreement. If you are an end user, please be advised that datasheets will not be supplied. Requests from Open Source Developers and individuals creating system management tools and other utilities for public use are considered on a case by case basis. Driver developers can also find links to information from Microsoft and assistance for Linux developers here. Please complete the request form below and clearly explain what information you need and why.

Thank you for your cooperation." - VIA Datasheets Request Form

Now, you might be saying big deal. All you have to do is fill out a form. Now, notice they will only pay attention to you if you are a big company. Well, in my previous life I used to be an embedded hardware developer. I helped design new circuit boards for a variety of companies. Some big some small. I guarantee you that if I saw a statement like this I would not even contact VIA. We have deadlines to meet and we don't need any of this red tape crap. Its hard enough to get a design out on an impossible deadline without resorting to begging a company to give us info for their stupid products. Most other companies are very forthright with their info and will even make sure a Field Applications Engineer is there to answer your questions. I know alot of other designers will back me up and agree that we don't need one more roadblock. Then again, I guess VIA's focus is just on the few manufacturers that use their products. In addition, I think its pretty sad that they only release their datasheets on a "case by case basis" to OSS developers. Whatever. This makes me wanna hurl my Asus Terminator (with VIA chipset) out the damn window. Now, I know why the Linux kernels are so damn finicky on this system. I was seriously considering buying a little mini-ITX board and case before I saw this, but not anymore. I can use someone else, and I will buy from someone else for personal use as well.

JOhn

Re:+5: Obvious

[driv3l]

I did look for phone numbers... and the only numbers available are marketing and sales.

Re:+5: Obvious

[Anonymous Coward]

It helps if you have rice eyes and talk that ching chang chong chinaman gobbledygook.

So solly. Numbel one son have rost data sleet."

have you asked them yet ?

[sir_cello]

"from what I've heard it's damn near impossible to get datasheets from them"

Have you asked them yet ?

A quick search reveals mailing lists where VIA engineers freely handed out technical data sheets for earlier models.

"So"

[JohnFluxx]

"I dislike looking through other people's source code, so looking through Linux drivers is counterproductive for me"

I fail to see the link that the "so" implies.

Just because you dislike reading other's code, doesn't mean it is counterproductive.

Re:"So"

[Lars T.]

So if he wants to release his project under anything but the GPL, he better claim to not look at the Linux source. Or is having RMS on your back productive ;-)

Re:"So"

[cowbutt]

Remember that you still have the standard Fair Use rights allowed by your jurisdiction. These will probably allow you to study and write your own code based on GPLed code, as long as you don't include actual GPLed code.

--

Re:"So"

[JohnFluxx]

a) Your logic is screwed up. Saying that one doesn't like to read others code can not be concluded to that it would be counterproductive.

b) He can look at the code all he wants, and use it to work out how the device works, and then write his own driver.

Just looking at GPL code doesn't mean you can't then write your own proprietry code. What you can't do is just copy and paste the code.

Re:"So"

[RGRistroph]

You can take a GPL'd project, display the code in one window, and type in your own code, using it as the example, in another window, and as long as it is a different work, you have full copyright to the separate, non-derived work you just wrote. One way to try to make sure everything is different is to translate it into a completely different language. The important thing is to make sure you are not creating what would be a "derived work" under Title 17. An automated translation by a program would probably be considered a derived work, no matter what new language it was translated to.

Issues of trade secrets sometimes involve trying to find a "virgin programmer" and have them write a "clean room implementation" that can be proved to have resulted without examining someone else's code. Perhaps this is what you and the original poster were worried about. It is difficult to see how a trade secret could be in a GPL'd work because a trade secret looses it's status as a "secret" once it is published.

However, you should not confuse issues of copyright and tradesecrets. The belief that merely looking at GPL'd work can endanger other work you write is a product of the sloppy use of catch all phrases like "intellectual property."

Re:"So"

[driv3l]

Actually.... what I was trying to convey is that it is much easier and cleaner to work from a datasheet than trying to understand someone elses code.

Maybe it's just me, but I find the linux code fairly complex to read and follow.

It gets more hairy when trying to understand the interactions of the code with certains bits of hardware.

By the time you've finished plugging in the constant values (#defines) and looking through the 10 different source files, you could have written 4 drivers.

Also... working from a datasheet allows me not to be influenced by another person's approach and implementation.

Letters.

[Anonymous Coward]

Write a friendly e-mail to the webmaster, copied to the customer service department of Via, explaining the non-functioning link, and how much this would help you in supporting their product in another OS.

Then, looking at their who's who corporate page find the closest things to their technical department, and customer service department. If they take longer than a week to respond to your e-mail with something that isn't computer generated, write a letter.

The first paragraph should be high, how ya doing, I enjoy consuming your products. The seconds should be, but not so much this time around, and here's why. The third might use the slow e-response as a transition, into, "I'm a little worried. I'd appreciate attention. Thank you for your time." Naturally, a classic sign off and signiture to close things out.

This response might take quite a while longer. And should be addressed probably to both the technical and customer people, and the vice presidents of their divisions. After all, it's not just the plans you're having trouble with.

You might not get the response you wish, but I'd actually be surprised if you didn't, and truly shocked if you didn't get any sort of reply.

If they help you out, I'd probably write a "thank you" letter too.

Re:Letters.

[aminorex]

No, it's not worth it.

There are other vendors, other products,
and much much better ways of doing business
and treating your customers.

Instead, I suggest telling everyone in the
industry how VIA is hostile to its customers
and refuses to provide technical documentation
for its products, rendering them useless.

When you do a design, turn to a competitor.
VIA's intransigence can be resolved by
destroying VIA as an agent in the marketplace.

Only the good survive. Helping this lawyer-crippled company to survive
would just leave you with a deformed moron
sucking your life away.

Re:Letters.

[Anonymous Coward]

Have you actually tried writing letters when companies tick you off? You might be surprised at the effect. I certainly was the first few times I tried it.

In so far as your Introduction to Microeconomics viewpoint of 'All companies must respond to the marketplace or be destroyed,' goes, I can't say I disagree completely. But at some point someone has to let them know they're not serving the marketplace, and provide them with not only an impetus to change, but the opportunity as well.

It takes very little time, a similarly small effort, and can occasionally have huge results. Furthermore, reasonable correspondence from you, matched up against their unreasonable responses will go a long way in helping you trash their image. Assuming, one wants to pursue the public crusade against VIA, or which ever company had done them wrong.

Randomly enough, I was just as apathetic, and certain of corporate apathy, as you seem to be. What changed my mind, to some degree, were a couple of book written by Don Novello (Father Guido Sarduchi from Saturday Night Live). These books are collections of patently ridiculous letters by Mr. Novello posing as Lazo Toth, or Ted L. Nancy to corporations and public individuals. Only very rarely did the entities contacted not respond, and occasionally provide free stuff. Of these letters, not one can be mistaken as serious, or reasonable. (But they are damn funny.) Yet they have impact and get results. It would be akin to the trolls of slashdot, if they could abstain from certain pictures, and salty language, would not only almost always recieving a response, but sometimes presents.

It might not be very intuitive, but that's the way it appears to work.

VIA sucks for GCC too!

[MarcQuadra]

VIA sucks for the GCC folks too. You'd think that a CPU manufacturer with a dramatically new design would at least sponsor someone on the GCC team to write an optimized C3 target. Instead we had to use i486+3DNow!

Now with the nehemiahahaha CPU I think the -march=c3-2 target just points to i686+sse.

If they kicked a few spec sheets and a small donation to the GCC folks I'd bet GCC-generated code would run 30% faster on their chips. The C3 series should have it's own pipeline description and scheduling backend.

Re:VIA sucks for GCC too!

[Anonymous Coward]

Problem is this:

unless some set top box manufacturer tells via that they want to buy 10,000 EPIA's and MUST have full speed linux on it, you're gonna be writing that driver yourself.

VIA is a corporation, not a university so I don't see how making a donation to the GCC crew would increase VIA's bottom line. Remember, profit motive still drives corporations.

Re:VIA sucks for GCC too!

[RGRistroph]

Well, I can see myself buying 2 to 5 EPIA's over the next two years if they work well, are documented, and are actually fast enough to decode video on the fly. Why don't we start a list of petitioners who will sign a "I'll buy X units if . . ." statement ?

Re:VIA sucks for GCC too!

[cryptor3]

Yeah, I really wanted to take a look at what kind of stuff they had for the M10000's often-touted CPU crypto features. But I was very disappointed when I found absolutely no technical documentation on those things whatsoever. Nothing beyond marketing hype. What a gyp.

Re:VIA sucks for GCC too!

[Anonymous Coward]

Yeah, just find 5000 people like yourself that want to buy 2 units, and VIA will be happy to deal with you.

Re:VIA sucks for GCC too!

[Anonymous Coward]

What is dumb is that if VIA doesn't allow developers to make applications that use VIA chips (ie. by giving them specs that allow them to complete their work), then the "killer app" that uses VIA chips may never be written. And VIA will lose out on potential sales and profit.

This is obvious to most other manufacturers of general purpose processors.

Re:VIA sucks for GCC too!

[MarcQuadra]

The hardware video decode is also totally undocumented. I guess they COULD decode loosely encoded DiVX video, but it would totally hose the CPU, and I'll bet you'd skip frames. All the windows benchmarks you see use the accelerated video drivers that decode MPEG2 via an ASIC that Linux can't touch yet.

Right now XFree86 doesn't even have native drivers for the CLE266/CastleRock 2D video, you'll have to use the FBDev driver (slow as molasses) or wait for XFree86 4.4.

Overall these EPIA systems pretty much suck, the LAN chip, video chip, video-out, and overall board quality pretty much suck. It's all THERE but it's bottom-of-the-barrel quality. The onboard VIA RHINE II LAN chip was using about 30% of my CPU during transfers, I had to add an Intel NIC just to get it performing to my needs. The RCA video out quality was totally a bummer, as was the VGA video quality on my nice monitor, everything seemed a bit 'blurry' like when you use a shitty $20 KVM with noname cables. Sound was decent, but who the fsck cares when you can't see anything right?

Re:VIA sucks for GCC too!

[RGRistroph]

I have heard similarly from other people. It seems like the kind of system that needs all the help it can get from good drivers and an optimized compiler.

Re:VIA sucks for GCC too!

[Anonymous Coward]

VIA is a corporation, not a university so I don't see how making a donation to the GCC crew would increase VIA's bottom line. Remember, profit motive still drives corporations.

The profit motive was driving all the other corporations that at various times have funded work to make gcc work great on their hardware. They understood that if our hardware is more useful, more people will buy it, and we have more profit.

Re:VIA sucks for GCC too!

[MarcQuadra]

exactly!

If you intend to release a product that is inherently slower than even modern low-end systems you think you'd make a heavy investment in getting the most out of it.

Re:VIA sucks for GCC too!

[Anonymous Coward]

bullshit.
via's xstore crypto RNG is completely supported with a linux kernel driver IN THE STANDARD 2.4 KERNELS YOU DUMBFUCK. my m10k uses the via padlock feature all the time with debian woody and a 2.4 kernel (2.4.23pre6).

Ask in forums?

[ErpLand]

There are numerous "fan-sites" about mini-ITX boards, especially those from VIA. Maybe some of the folks on there have already managed to obtain some docs.

For example: forums.sudhian.com have sections about linux and mini-itx boards.

Or how about the kernel mailing lists? I think that Alan Cox himself mentioned in his diary (before he started writing it in Welsh, that is) that he's a fan of these machines and had done some driver development.

contact their rep or distributor

[Anonymous Coward]

someone [e.g. avnet?] will be a representative for them, and someone will be the distributor.
these are the normal channels to deal with a semiconductor mfg.
the rep will be able to provide the datasheets etc (perhaps under NDA, depends on company).

Re:contact their rep or distributor

[Almost-Retired]

I certainly wouldn't waste any time dealing with a distributor or field rep. They rarely have an adequate understanding of the language even if they speak english as a 1st language, won't relay the message till next tuesdays weekly confab, and it will get routed first to the legal dept. for a 30 day review, then get bounced back to sales, who A: won't know by then what was asked and B: will call you and ask you how many units you're talking about. You'll explain that its data sheets you want and then the cycle starts all over again, and if you don't become frustrated and say screw 'em and the camel that rode in on them after about 4 months (or less depending on your deadlines) of this, during which time your competitive edge in terms of time has long since expired.

I can get samples and data sheets here in WV to my home mailing address's mailbox in 7 days or less from National Semi, Harris, Motorola, TI, AtMel and at times from AMD if its not a cpu. And do it in one phone call after looking them up in 1-800 info. Lifes too damned short to screw with the likes of VIA and their proprietary information attitude. Its a situation which ought to be self healing by their going away, unforch the big mobo contracts will keep them going for quite a while I suspect. Which is a shame. IMO, whats for sale ought to be fully available to anyone with the sheckles to pay for it. Quantities matter in the pricing of course, but probably in nowhere near the proper ratio when the qty's get below the 10K piece level by the time the support costs are added in for that small an order. Anyone who expects to pay only a 500% premium for a onsie/twosie order over the price Biostar pays for qty 100K is living in a highly addled mental state and should get over it, they simply don't understand the economics that make giving you a sample or 2 far cheaper than billing you for 2 of them.

Other semi people are falling all over each other to sell you their latest doodad, and a bit of searching might even find a better device than the one from VIA you've been contemplating the use of.

Your call of course.

--
Cheers, Gene
A mostly retired old coot.

LinuxBIOS

[kinema]

You might want to check with the developers of LinuxBIOS. I know that they have had some luck working with VIA. There knowldge should be sufficiently low-level for you to be able to write drivers.

Good luck!

[Anonymous Coward]

In april I had a little eMail conversation with Fiona Gatt, but as she realized, after I replied to her second mail, that the Datasheet is for an obsolete chipset built in 1995, she must have lost my address...