SGI: Hardware

nongrato wrote:

Pontus wrote: I look at it as a generic PC and run Ubuntu on it.

Indeed, because 550 IS a generic PC, probably just a rebadged Intergraph, while 320 is based on unique architecture and I don't think you can run Ubuntu on it.

The zx10 was the rebadged Intergraph, the 550 is a PC based on an Acer motherboard . A Rambus monster with the 840 chipset. Should do Quake III fast: The RDRAM Avenger - Intel's i840 Chipset

Maybe I was biased but I always thought the 320 was the nicest implementation of Windows 2000 you could get. Honkin' big box, though*.

*Yeah, I know. If you had two more inches of dick you could get some new pussy right here ...

Not as big as the 540, now that's a monster

hamei wrote: Maybe I was biased but I always thought the 320 was the nicest implementation of Windows 2000 you could get. Honkin' big box, though*.

*Yeah, I know. If you had two more inches of dick you could get some new pussy right here ...

yes, that was a suprpise how big this thing actually is.

Could anyone show me the location of those custom chips - Lithium, Cobalt, Arsenic?

I haven't found a diagram from SGI that labels them, although it may exist.
It just seems obvious that Cobalt is the central one with the fan, Lithium is the lower one by the PCI slots, and Arsenic is the left one between the DRAM and the back panel.

nongrato wrote: yes, that was a suprpise how big this thing actually is.

You have to see a VW540. It's even bigger.

the memory modules look like cut-in-thirds Octane 1 GB modules - interesting.

robespierre wrote: It just seems obvious that

Thanks!

Geoman wrote: the memory modules look like cut-in-thirds Octane 1 GB modules - interesting.

A closer shot:


What is the real capacity of these? System detects 1024 megs, but from my understanding 1024 can hardly be divided by 12.


Big dissapointment: the system doesn't play any sound on startup. Welcome to PC world.

nongrato wrote: What is the real capacity of these? System detects 1024 megs, but from my understanding 1024 can hardly be divided by 12.

That sounds right. Each bank of 6 DIMMs is 512MB. Don't think of them as 6 distinct DIMMs; it's more like two groups of three pieces each. Geoman's remark about "cut-in-thirds Octane" DIMMs probably isn't too far off the mark. The Visual Workstations are known for their weird DIMMs!

See also the Visual Workstation 320 Owner's Guide on SGI techpubs . The memory chapter has some info on valid configurations. Yours is the maximum supported. Also, the Appendix on Technical Specifications suggests that the DIMMs are 48 bits wide each.

I beleive all those modules are idential. After receiving 320 I removed and cleaned all the modules and put them back in random order. Labels also seem to be ideitical. The wikipedia page states that the maximum memory per module was 96Mb which gives us 1152(96*12) in total.

nongrato wrote: The wikipedia page states that the maximum memory per module was 96Mb which gives us 1152(96*12) in total.

The Techpubs document says that the maximum memory modules are 16M x 48 parts, which does multiply out to 768 Mbits or 96 MBytes as Wikipedia suggests. But that doesn't account for the ECC (or whatever) overhead. Using six of those 48 bit modules gives an overall width of 288 bits, but the SGI documents show a 256 bit wide memory bus. So apparently they used a lot of bits (32 of them) for parity, ECC, or whatever other internal (non-user visible) purposes SGI has for the memory on these machines.

jpstewart wrote: 16M x 48 parts, 768 Mbits 96 MBytes ECC 288 bits, 256 bit wide 32 of them

nongrato wrote: Big dissapointment: the system doesn't play any sound on startup. Welcome to PC world.

A lot of the later stuff does not. The story is that the data in the prom has grown to where there's not enough room for music.

At least the VW320 has a prom, not a bios

nongrato wrote:

jpstewart wrote: 16M x 48 parts, 768 Mbits 96 MBytes ECC 288 bits, 256 bit wide 32 of them

Well, it's pretty standard - 1 bit of parity for each 8 bits of data. 256 bit /8 = 32 bit. But I think the chips are more like 64M x 12 than 16M x 48- the biggest module has 12 chips (6 memory chips stacked in two, on each side).

Alright, one month since I got VW320. This is how it looks now:



I'm still using cheap,I say "disposable", keyboard and mouse, but looks like i'm gonna get original VW keyboard soon. Windows 2000 appeared to be a better OS than I expected. After installing Microsoft Services for Unix I have a normal shell in Windows, a better POSIX compability and even GCC. But I miss mediarecorder, videoin and videoout from Irix. HackTv and Qcap are crap: slow and buggy. Also I missed red mouse pointers, but this was not a problem - I manged to create my own cursor set for Windows. It doesn't exactly match X11 cursor set, but If someone needs it, grab it here:

(link removed)

About the performance. While the Cobalt chipset is a little bit faster than MXE of my Octane(measured with Viewperf 7.0), dual R12k 360MHz still greatly outperform dual P-III 500MHz. On some tasks like rendering in Blender or raytracing with C-Ray Octane is nearly two times faster. Rendering in Maya is about the same, but I suspect this is just a result of a better optimisation for x86, because once my 3D scene needs fluid or rigid body simulations Octane is faster again. Although, web serfing seems to be more comfortable in Windows. At least Firefox doesn't crash.

Games played:

Quake 3 - runs fine. FPS is a little bit higher than on my Octane.
Half-Life - runs fine, with a decent FPS, but not in fullscreen.
Serious Sam - slow, after 20-30 secs of gameplay the systems reboots. Overheating?
Age of Empires II - runs great. No wonder - it's a 2D game.
Return to Castle Wolfenstein - satisfying, FPS varies from low at outdoors locations to good at indoors. But all sounds are delayed by a second. Probably DirectX's fault?

If I am not mistaken, I think the VW systems are the reason why the NT kernel uses ARC naming for raw device access. MS included some of the ARC (note, ARC is little-endian, ARCS is big-endian) standard in NT, especially for NTLDR and BOOT.INI (until they moved that into the registry in Vista).

I also believe that, hardware-wise, the VW's, at least the 320, is based on the same shared memory architecture as the O2. Both use the Graphics Backend chipset (in Linux, the gbefb driver). Not sure if the same applies to the 540, though.

Kumba wrote: If I am not mistaken, I think the VW systems are the reason why the NT kernel uses ARC naming for raw device access. MS included some of the ARC (note, ARC is little-endian, ARCS is big-endian) standard in NT, especially for NTLDR and BOOT.INI (until they moved that into the registry in Vista).

NT's ARC support predates VW by several years: ARC . NT supported ARC since it's conception. You could even launch Windows NT 3.1 betas on MIPS.

Kumba wrote: I also believe that, hardware-wise, the VW's, at least the 320, is based on the same shared memory architecture as the O2. Both use the Graphics Backend chipset (in Linux, the gbefb driver). Not sure if the same applies to the 540, though.

Yes, the 540 is a more expandable 320.

nongrato wrote: Does that Cobalt chip support any version of DirectX? Does anyone have any experience running games that require graphics acceleration?

Took me some time to check and the answer is no - dxdiag screenshot attached.

GL1zdA wrote:

nongrato wrote: Does that Cobalt chip support any version of DirectX? Does anyone have any experience running games that require graphics acceleration?

Took me some time to check and the answer is no - dxdiag screenshot attached.

A year and 2 months.. those things are so slow to boot!

uunix wrote:

GL1zdA wrote:

nongrato wrote: Does that Cobalt chip support any version of DirectX? Does anyone have any experience running games that require graphics acceleration?

Took me some time to check and the answer is no - dxdiag screenshot attached.

A year and 2 months.. those things are so slow to boot!

Mine hates me. It will randomly stop booting and requires juggling with the three CPUs inside. And there's something strange happening when I benchmark memory bandwidth (with SiSoft Sandra)- it should beat every Pentium III chipset except for the ServerSet III HE with MADP chips but the results are slower than the Intel 440LX with PC66 memory. And it's even slower when I'm using the multi-threaded version.