IBM To Demo OpenPower 710 At SCALE 3x

An anonymous reader writes "IBM will demo their PPC based servers including their new OpenServer 710 at SCALE 3x this week. In addition they have their i5/520 running Power Linux, Intel Linux, AIX, i5/OS (OS/400), and Windows all simultaneously. SCALE will be held this weekend in Los Angeles at the LA Convention Center. Speakers include Kevin Foreman (Real Networks), Jon Hall (Linux International), Larry McVoy (CEO BitMover), Marc Hamilton (Sun) & 30 other sessions. In addtion to the talks there will be over 40 exhibitors including IBM & Novell. If you're in LA drop by on 2/12-2/13. There will also be a dinner and GPG Key Signing party. (For a free exhibit hall pass register with the promo code "FREE" or a discounted full access pass with "NEWSP".")

Comments

Benchmark recomendation

[Leroy_Brown242]

I think they should use the distributed.net client for benchamarking. maybe they can take a few decades off of the RC5-72 challenge, or finish up OGR. :)

Did they say windows????

[LWATCDR]

Are they running it under emulations or is there once again a PPC version of windows?

Re:Go PPC!

[Hal_Porter]

Why? x86-64 does pretty well in benchmarks, and it's cheap as hell.

Even if you pick your chips for assembler elegance, x86-64 isn't too bad. Most of the x86isms that hampered performance are either not used anymore (e.g. segmentation) have been fixed to some extent or no longer matter much in a modern chip running modern code.

In the "fixed to some extent" I'd include the lack of integer registers - x86-64 has twice as many, the nasty FPU architecture - x86-64 uses SSE2 instead. Also, variable sized instructions can be executed efficiently if you transform them into uops and give a higher code density than a typical Risc, so you get better cache hit rates for a given cache size.

And by "modern code on a modern chip" I mean that if you look at the code that comes out of a modern compiler, it's almost as good you would have got on a clean Risc design, but you run it on a chip with a clock rate that is much faster than Risc chips manage, because the x86-64 world now has two companies competing, and they can both afford high end fab plants.

In fact I'd say that x86-64 will eventually kill all the Risc chips, PPC included in the desktop/server world. The embedded world is different of course.

First-hand experience with OpenPower

[winmonster]

I just read this entire thing at -1, and there hasn't been one post that indicates anyone has even seen an OpenPower box. We purchased an OpenPower 720 box a few months ago. 2x 1.65 GHz procs, 4 GB RAM, 2 73 GB drives, 6 146 GB drives for about $22,000. We run SLES 9 on it, and use it for ITSM. I chose it opver and x86 box for one reason: I/O, which is pretty much where all big-iron architectures trounce x86. There have definitely been growing pains, though. I had to work with SUSE on a fix for installation onto an IBM RAID array. Also, I cannot utilize all of the disk space that was purchased. Both of those issues are results of IBM controllers requiring you to format a drive with 522 byte sectors. The results are workable and perform well, but not quite the "certified" solution I was expecting. Anyways, my 2 cents. I still think I made a good choice, considering management wouldn't go for a PSeries/AIX box.

Re:G5 = Opteron

[Hal_Porter]

Intel chips are x86-64 too, and they run up to 3.8Ghz.

Not that it helps them much on Spec benchmarks -

http://www.aceshardware.com/SPECmine/top.jsp

But for most of the last couple of years, the best performing x86 managed to run at a higher frequency than the best performing Risc, or turn in better SpecInt benchmarks - at least for single CPU, or sometimes both.

PC hardware tends to suck for multi CPU stuff though - even Opterons are a bit disappointing to be honest, considering that decent SMP performance was supposed to be one of the design goals. On the other hand, I don't think switching instruction sets would help here, it's something that will be fixed with multicore chips (because they make SMP more mainstream and thus more important) and better bus protocols - maybe AMD will license the Power5 one like they did for Alpha.

FP performance is a bit disappointing too - especially as the Athlon FX-55 result is with a decent 64 bit compiler - it's still a fair bit behind Itaniums and POWER5. On the other hand, how much mainstream computing is limited by FP performance?

Actually, Power5 does better than I'd expected - much better than PPC970, but it isn't enough to get people to switch from x86. Back in the Alpha days, Risc chips had a advantage in virtually all benchmarks, and conventional wisdom said that it would increase, but it still wasn't enough of an incentive to get significant numbers of people to switch over.

Which is my point really. It's not that x86 has an particularly good, it is that PPC and other Risc chips no longer better enough to make people switch.