Sun & Fujitsu Team On SPARC Chips & System

An anonymous reader writes "Sun and Fujitsu just announced a 20-year partnership to jointly develop SPARC based technology and systems. It looks like the long-predicted partnership that was hinted at earlier has finally come to pass in a much more comprehensive manner than I've heard anyone predict, i.e. not just chips, but a unified range of systems. My guess: Sun drops Ultrasparc III to provide the Throughput computing chips for the low end / web / network stuff, and takes up the Fujitsu provided SPARC64 chips for the high end and workstation market. Will this spark a new RISC renaissance for Sun and Fujitsu? Or is it a last gasp before Opteron / PowerPC / Itanium crush them? I for one will be interested to see what systems and processors come out of this. This could really revitalize the SPARC system market, especially if Sun's work on Throughput computing proves out."

Throughput computing

[shoppa]

I always thought Sun's only realistic market for "Throughput computing" (maximizing ops/watt) was dense server farms (e.g. blades). Now it is true that this market is playing out quite like everyone (especially Sun) wanted it to, but it is a real market.

For thin-client stuff, while low power consumption is a priority, it's not a big enough one to warrant the amount of money that Sun and others have spent on it. Maybe, just maybe, as a spinoff.

These "find a market for our new processor" discussions are getting a little depressing. I remember being excited about the DEC Alpha for embedded applications, but since then it all feels hollow.

What's actually going on here...

[Anonymous Coward]

1) Sun is having troubles convincing its partners that its multi-core "throughput computing" chip will be competitive. That Gartner report is causing people to ask questions about whether Sun can deliver on its promises. And who wants a 500Mhz 16 core chip anyway? Think of the memory bandwidth problems!

2) Fujitsu Sparc core spanks Suns own core.

My prediction? Sun will abandon its multi-core, asynchronous research pipedreams and farm out all CPU design to Fujitsu. CPU design is a very costly (comoditised) business for Sun to be in, and as Apple have shown its the system that matters, not the processor.

Throughput Computing and the Transputer

[dnnrly]

Sun's description of Throughput Computing and their approach of putting multiple processor cores reminds me of what Inmos [inmos.com] tried to do with the Transputer before they became STMicroelectronics [st.com]. The idea was to have many small processors positioned close to each other, communicating between each other closely. I seem to recall seeing transputers on eBay a while back for huge amounts of money. By all accounts, a transputer board was a very useful piece of kit for the right appplications!

provide the Throughput computing chips for the low

[Anonymous Coward]

provide the Throughput computing chips for the low end

It's more likely Sun will start using Opterons for the low-end. Why? Because (IIRC) Opterons scale much better than Intel chips in a multiple-CPU environment. And that multiple-CPU ability to scale damn near linearly is Sun's real strength in the computer market.

And they want to give that hardware away because they think people are clamoring to pay for the software they put out?!?!!??

Re:What's actually going on here...

[Anonymous Coward]

You want 16 cores because servers (especially Sun ones) tend to run many process, with many threads. You want to distribute those threads across multiple CPUs so one doesn't bog down the whole system.

Regarding memory bandwidth: look at Sun's I/O bus architecture.

Re:What's actually going on here...

[larien]

Well, there are other issues like lock contentions (making sure 2 CPUs aren't using the same chunk of RAM) but the point is correct. The problem with 16 500MHz cores is that a single threaded app will still run at the same speed as a single 500MHz core; you would be able to run more of them at the same time on the 16 cores, though.

Re:What's actually going on here...

[turgid]

Apache runs just fine on commodity hardware.

So it does. There's more to life than apache though.

I dare you to look at this [sun.com]. Then, think for a minute about what sort of things you'd use it for.

Re:What's actually going on here...

[FatherOfONe]

Oracle would absolutely take advantage of this. Any multi-threaded app would. Like say most of the mid-tier app servers. So in some cases it would be possible to have your database, app-server and web server on one box. Granted that box could be setup with different partitions, (similar to vmware in the low cost world, or perhaps user mode linux).

Re:20 years?

[Billly Gates]

And its not going good.

The reason Sun is losing is because the SparcV should be out that is comptetive agaisnt (theoritical) agaisnt the power, mips, and Opteron.

TI who actually fabricates teh chips is pulling a Motorolla in order to gain more profits by not upgrading their plants.

Either they innovate and skip the sparcIV and leep to the sparcV and develop the sparcVI or give in to Opteron now and save the company.

Open-source platform for games

[Anonymous Coward]

I wonder if funding a kind of open-sourced game development platform wouldn't help the hardware manufacturers (e.g. Sun in this case for the processors and ATI etc. for the graphics) sell their wares. All 3D-engine stuff wide open and free, so that a prospective game maker would not have to buy an engine license.

Problems of course:
- need an installed base to sell enough games
- state of the art engine does not grow on trees
- willingness of hardware types to work together

Possible Pros:
- open standard encourages performance improvements for the next generation that don't break older applications
- massively distributed bug fixing

Re:20 years?

[christophersaul]

Why 'give in' to Opteron? It fits in well with the Sparc kit and Sun are already aggressively marketing Opteron.

Sun are capable of having a strategy that can move with the market, as well as dictating to the market, as appropriate. I get tired of people on Slashdot claiming one company or product is 'dead' just because it has a competitor.

Sun also aren't particularly 'losing', as you put it. Unit shipments were up 26% for the first quarter of 2004, with the UIIIi systems selling extremely well - and they're positioned directly against the Xeon based stuff that Slashdot readers tell us is going to take over the world.

Re:What's actually going on here...

[turgid]

The new highly-multithreaded [sun.com] chips have much shorter pipelines unlike current headless-chicken designs like Pentium 4, which operate at obscene clock frequencies but take ages to refill after a pipeline stall hence spening a lot of their clock cycles doing nothing. The idea is that the chip has several simple low-latency cores with many "thread contexts" which can be switched in with a 0-cylce delay when the current thread blocks. So, going back to your Amdahl's Law, the "unimproved fraction" (1-P) on one of these processors is proportionally smaller in many cases where your clock frequency increase on your Pentium would buy you little benefit. Or something.

Witness intel's recent change of direction regarding the future of Pentium. They've all but EOL's the Pentium 4 Netburst architecture and are now going multi-core. It took an anouncement from their competitors though, and lengthy explanations and analysis in the industry press, before they did, once it was absolutely clear that the clock frequency wars were over. Intel is well behind, but they have an absolutely astronomical R&D budget.

The best use for Solaris Zones: N1 Grid

[_damnit_]

The real upside of of zones is when combined with N1 Grid. I know it sounds really amorphous, but with zones it really starts to make sense. Imagine rolling out a new zone or dozens when needed across a datacenter full of stock Solaris 10 machines (x86 or SPARC). The storage is on EMC or Hitachi, so you just hand out LUNS like candy for the zone's "/" with predefined pkgs and patches and you have a really dynamic environment for Oracle 10g, webservices, etc.
This is the story that needs to be told, but I sometimes think Sun is using Novell's old marketing team.

DISCLAIMER: I work for Sun but I try not to drink the company Kool Aid.