VIA's New Nehemiah M10000 Processor Reviewed

Joseph Wharton writes "Mini-ITX.com has a review of VIA's new Nehemiah M10000 EPIA-M motherboard and processor. Some of the new features include a full-speed floating-point unit (finally!), SSE instructions, 64KB of full-speed L2 cache, and (get this) a hardware-based random number generator. Also, there's IO/APIC support in these new procs, potentially paving the way for dual EPIA boards."

Comments

a bit about the cpu since it's /.ed

[bpland]

It's almost dead but here is the page about the CPU.. interesting. hehe

"Nehemiah is the next generation C3 CPU, and features a number of improvements over the Ezra-T C3 used in all previous EPIAs. It has The 20.5 million transistors, and uses a 0.13 micron process. For comparison, a Barton Athlon or Northwood Pentium 4 have about 55 million transistors, and recent GPUs have over 100 million transistors.

The Nehemiah is designed to work at clock speeds of 1GHz and beyond - the Ezra-T is designed at up to about 1GHz.

Nehemiah has a die size of 52mm2 - the world's smallest x86 processor. It has been designed to minimize power consumption and optimise heat dissipation - VIA call this "Coolstream". Some active cooling is still required, but not very much. Let's hope for a Nehemiah Eden C3 version.

The Nehemiah features SSE instructions instead of the 3DNow! instructions featured on previous C3s. This should bring enhanced performance in 3D applications, which are optimised for more modern SIMD instruction sets. SSE optimised image processing applications should also benefit.

Full Speed FPU - the Nehemiah has a full speed floating point unit for the first time. The Ezra-T has a half-speed FPU. Floating point calculations are used heavily in 3D rendering, multimedia, and streaming applications.

Enhanced 64KB Full-Speed Exclusive L2 cache with 16-way associativity. An exclusive L2 cache gives a larger effective total cache size as it doesn't replicate the contents of the L1 cache. The more cache available, the more chance there is that program loops can run in cache and not comparatively slow main memory.

StepAhead Advanced Branch Prediction - Looks ahead and gathers the data needed to optimally run applications

A hardware based random number generator (RNG) has been added. This creates true random numbers from the random electrical noise on the chip. This is of much use in security applications, allowing a strong cryptographic key to be generated. VIA call this the "PadLock Data Encryption Engine".

Future Nehemiahs will feature IO/APIC support. An Advanced Programmable Interrupt Controller (APIC) provides multi-processor interrupt management - dual processor EPIA anyone?

The Nehemiah is available in EBGA or Socket 370 packages - both are low profile and require less board real estate."

Re:Makes for a great jukebox

[buffer-overflowed]

I forgot, don't expect one of these to run Doom 3 or UT2003. They will run office, and they will play DVDs. The earlier ones required no active cooling, it's still an extremely low power chip however.

I found the M10000 for $182 at directron, and here's what you get for your money:
VIA C3 1GHz processor
10/100 Ethernet
Firewire
TV-OUT (S-video, RCA(PAL and NTSC))
6 Channel Audio

Not a bad deal, methinks. Probably can be found cheaper, but I didn't want to look too hard.

Re:New Via

[tjrw]

A hardware random-number generator is useful for crypto. If you've ever tried porting something like OpenSSH to a platform that didn't have decent RNG support (i.e. no /dev/random or /dev/urandom like Linux has), you'll have run into the fun and games of trying to come up with a decent random source.

Hardware support for RNG is a "Good Thing(TM)", and doesn't necessarily have anything to do with the "Trusted Computing Platform" or whatever the DRM flavour of the day happens to be ! :-)

Re:64KB cache?

[mmol_6453]

Cache is only useful when you're working on a small data set. Multimedia applications tend to constantly move through a large data set, which makes the cache all but useless. Full details here.

Why Via names stuff after Christian Mythology

[Indy1]

The ceo of Via, Wen-chi Chen, is a Fundlementist Christian, so as a result this is the name source for many of their products (joshua, sameul, nehemiah).

Re:Why Via names stuff after Christian Mythology

[David Chappell]

Nehemiah was a 6th or 5th century BCE govenor of Judea during a time when Judea was under Medo-Persian rule. I would describe Joshua, Samuel, and Nehemiah as figures from Jewish history rather than Christian mythology.

Of course, in the 19th century it was popular to assert that public figures mentioned in the Bible are figments of the writers' imagination, but this view seems to be largely discredited. The names of too many of these 'fictituous characters' showed up on monuments and public records uncovered by archeologists.

before we start griping...

[iamkrinkle]

before everyone starts comparing this to p4 or athlon, it's not meant to compare to them. this chip is only 1 ghz, but the selling point is it's low power consumption and it doesn't run too hot (the slower cpu's use only passive cooling). So yeah, you're not going to be playing doom 3 on it, but you can do cool things like put it in your car or have a pc that is (almost) completely silent. So for around 200 you get a mobo/cpu/video card/sound card/etc... not too bad of a deal if you ask me...

Re:How is your experience?

[molarmass192]

Have an EPIA 800MHz, works great for MP3s but bought it for a media center. Not enough "nuts" for decoding MP2 video in real time. The fullspeed FPU on the 10000 would certainly help in that department. Bogos show up as 1200 but that's only for 1+1 stuff, not 1+1.1. The best part of these little boards is they're dead quiet and generate miniscule amounts of heat. For that reason alone, I'm looking into the 10000 as a replacement for my current EPIA.

Re:Makes for a great jukebox

[seafoodbuffet]

Anyone considering a M10000 should read this first. Basically, VIA released two separate boards under the M10000 name. The second of which is the only one to feature the Nehemiah core. The first series (the one from directron.com) still uses the older Ezra-T core. Right now, there's not many vendors who can reliably provide Nehemiah processors. The only places I've seen are idot.com and monarchcomputer.com.

More info from Via Press Release

[MikeD83]

VIA Press Release, April 15, 2003

Notable features:

• 10% drop in power consumption
• 50% drop in system noise
• Integrated MPEG-2 decoder
• ATA-133
• 10/100 Mbps Ethernet
• S-Video and RCA tv-Out
• S/PDif digital audio connection
• 1 avaliable PCI slot

I suspect

[Arbogast_II]

That this comment, although funny, is from someone who has never used the Epia 10000

Low power cpu - A great thing

[LocalHero]

I think it is a great thing that a company has started to make low power cpus. Imagine all those P4 and AMD cpus out there that waits for Word to tell them to do something. You dont need 3ghz for that. A modern P4 or AMD processor uses about 70W of energy for nothing.

Hey if you could reduce that to 35W you are not only geting 35W less for the cpu you are also lowering the power consumtion on the air condition. An office building that starts to take the power consumtion serius could save lots of cash on electrical bill and probably some on the environment to :)

VIA's reluctance to support Linux

[phliar]

On the Via Arena Forum for EPIA-M Linux support the outlook is bleak. Via seem reluctant to fully support Linux, in particular with the MPEG hardware and the EHCI USB. One person trying to develop a distribution for EPIA-M says:

The EPIA-M's aren't very well supported on Linux at all. The bare minimum will work, but there is a lot of work to be done, especially on VIA's part.
...
VIA should not claim that the EPIA-M is Linux compatible, unless it is actively supporting the Linux community by improving support to at least a level of being able to take advantage of most of the hardware's features. I hope VIA becomes more actively involved in the future.

Here's the complete summary of hardware support for EPIA-M by "jonthorpe" on March 10, 2003:

USB 2.0: I have been in contact with David Brownwell for the past two months in an attempt to resolve the issues with the VT8235's EHCI support in general. Most people will experience system hangs when they attempt to use the EHCI controller, although it is improving. VIA has apparently provided David with hardware, but with little support. As of late, the problem seems to extend beyond the VT8235's EHCI - on another VT8235 based board, it was revealed that there are IRQ/PCI issues causing any USB 2.0 card (e.g. an NEC which is usually stable) to hang.

Audio: ALSA should work for the EPIA-Ms with their ALC650, but the OSS drivers are still a better option in my opinion. The original EPIAs will experience problems with ALSA (crackly sound on new versions of ALSA, no sound at all on previous). The OSS driver in newer kernels (e.g. 2.4.21-pre5-acX) works well with OSS playback, but there are two major problems:

1. ESound does not work. This limits the user to only one stream to pass through the sound card. This can generate problems with applications such as web browsers which use the Flash plugin (especially if another sound application is already running - e.g. XMMS)
2. Recording does not work whatsoever. This is a major problem that I am yet to talk to Alan about, but there is absolutely no recording with OSS drivers. ALSA drivers have more success with this, but the audio quality is poor.

VGA: This will work nicely if you're running one of VIA's supported distribution versions. This is not good for people like myself who insist on using updated kernels to work with other features of their hardware. This limits us to using the Slim driver for the CLE266, removing MPEG2 hardware acceleration. I am developing a distribution that is to be bundled with EPIA-M systems, and even after mentioning that, sending two emails to VIA and signing up on the driver request page has proven fruitless and are completely ignored.

I haven't tested this myself, but I believe there are problems with VIA's CLE266 drivers when it comes to virtual screens. The driver simply cannot handle them.

NIC: This works flawlessly, with the via-rhine driver. There are no problems to report.

FireWire: I have no idea about this one either, but I can say that the host controller is at least recognised by the Linux driver. I'd love to hear from anyone who has tested the EPIA-M's firewire on Linux.