Slashdot Mirror
4.7GHz IBM Power6 Spotted
Ilgaz notes that The Register has posted benchmark results from Oracle 11i running on four 4.7GHz Power6 chips. Quoting: "The speedy chips confirm IBM's boasting that Power6 would arrive near 5GHz. They also show that IBM's customers have a lot to look forward to in terms of raw performance." Rumor has it that the Power6 chips will be announced on Tuesday.
Despite the similar name, and somewhat related architecture, IBM's Power line are not PPC chips and aren't suited for desktop use. That's not to say that some technologies from them can't go in to other chips, but drooling over what is essentially a minicomputer/mainframe chip is silly.
The reason Apple switched is because, despite all the hype, Intel continues to make really fast chips for a good price. When Apple was on PPC I saw never ending arguments as to how much faster the chips were. All those never seemed to pan out in actual operation. Why that's the case isn't important from Apple's standpoint, they just want fast chips for low cost.
I suppose if you want to long for the days of Altivec and talking about tech stuff you don't fully understand, that's great, however Apple has to be a bit more pragmatic and realise that while Altivec might sound cooler than SSE3, SSE3 is an API for a damn fast vector unit and that's all that really matters. Most people don't care about contrived benchmarks, they care about the wall clock benchmark, meaning how fast does the system do what they want, and further how cheap can they get that system for.
POWERn != PowerPC
They (mostly) share a common ISA but the chips themselves have always been quite different.
The Power6 uses "under 100 watts in performance sensitive applications."
WAAAY too much for a notebook or a mini.
Not that it's the usual 27-page article, but still...
Who do you get to be an expert to tell you something's not obvious? The least insightful person you can find? -J Roberts
The stock price is more related to the iPhone and phenomenal iPod sales.
If your software has to take advantage of both chips it probably won't be optimized for either, and it's also a lot of extra unnecessary work.
// MD_Update(&m,buf,j);
Nope. This is the POWER6, not a PowerPC. IBM made the G5 (aka 970) as a derivative of the POWER4. IBM told Apple that they didn't want to make a derivative of the POWER5, so they were SOL on an upgrade path. This is not the kind of processor you would EVER see outside of a top-end workstation, server, or mainframe. It's not something Apple would have used.
why you should work twice more
Building universal isn't twice the work. Most apps don't have any intrinsic byte-order dependencies, and very few people ever wrote CPU-specific code that depended on Alitvec (for example).
-jcr
The only title of honor that a tyrant can grant is "Enemy of the State."
Building universal isn't twice the work. Most apps don't have any intrinsic byte-order dependencies, and very few people ever wrote CPU-specific code that depended on Alitvec (for example).
You can tell that to the Flash developers who worked their ass off to deliver the Intel version of Mac Flash quickly.
That little player has loads of ASM and SIMD instructions to be able to pull off what it does in this size and this speed.
Also you're not accurate about Altivec, multimedia apps like Photoshop make very good use of SIMD extensions, and interestingly enough, so do a big chunk of the audio instrument/processing apps.
The only Thinkpad in my requirements and price-range is $250 more than a Macbook, with a slower processor, smaller hard drive, and doesn't include bluetooth. (The X60 is $1,251.75 at sale price, and the Macbook is $1,000 after student discount).
Computers allow humans to make mistakes at the fastest speeds known, with the possible exception of tequila and handguns
Dude, all work's been done.
...or, you'll just pull in all the SSE work you did from the Windows Flash runtime since it's the same chip and these are all not OS dependent.
Either, you're doing a mac specific app and use the Accelerate.framework which handles conversion to SSE3 or Altivec depending on the platform...
Same thing for Photoshop. The plugin architecture makes it hella easy since they should have started with plugins for all the heavy stuff anyways. Recycling! It's not just for cans.
POWER isn't PowerPC. I don't think it's worth keeping two architectures in parallel like that for long term, not with Apple's current volume. Apple doesn't make that many Xserves last I heard, something like tens of thousands per year, when the other server companies exceed that by as much as 20x. IIRC, HP was selling 200k 1U servers when Apple sold 12k 1U servers. I don't know what IBM's numbers are, but POWER-based workstations and servers were a lot more expensive than Apple's stuff.
As to why Adobe can't be bothered to create a working flash player for (at least) 64-bit AMD64: I have no idea; we can't see the source so we can't see how difficult it would be to port it.
To be, or not to be: isn't that quite logical, Slashdot Beta?
For comparison, I think the Core Duo TDP in that machine is something like 30 W, maybe a bit more.
Very few individual applications need anything like this much. Video editing and 3D authoring are about the only two I can think of outside highly specialised systems. If you want to use more than 4GB of RAM, you still can on a 32-bit system. Modern x86 chips (i.e. anything newer than the Pentium Pro) support PAE, giving a 36-bit physical address space. This lets you have up to 64GB of RAM, although individual processes can only have 4GB of it each. I used a G5 for a while which had 8GB of RAM (sadly, it wasn't mine, and they wouldn't let me steal it), and it was really hard to get RAM usage above 50%. Even counting the disk cache, the amount of memory in use rarely went over 4GB, so I think it will be a few more machine upgrades before 64GB stops being enough.
By the way, the current generation of Apple laptops are limited to under 4GB of physical memory (3.5GB, as I recall) due to a problem with the Intel north bridge chip, not due to any limitations of the CPU.
On PowerPC, it makes sense to keep as much code 32-bit as possible, since that makes loads and stores of pointers cheaper, and results in faster code. On x86, going to 64-bit gives you a load of extra registers and a few other nice things, so it's not quite so clear-cut.
I am TheRaven on Soylent News
How on earth did the announcement of Power 6 turn into a debate about Apple and small consumer electronics? The Power 6 is designed to populate IBM's heavy-hitting AIX servers. They have large amounts of on-board cache and are designed to work in virtualized server environments - both hardware virtualization that IBM calls LPARs (logical partitions) and software virtualization (similar to Solaris zones/containers). A mid-sized server is capeable of running 50 or more AIX partitions and to copy one partition to another with a mere few seconds of interuption. The technology is very similar to the well known features of IBMs mainframes. IBM has strongly hinted that the P6 (and it's successors) will be the chip that will power future mainframes, AIX and I5 (as400) systems someday. The new chips use way too much power and are too large to fit in portable consumer electronics and I doubt any consideration was given to hand-helds during design. As for Apple - they have experts that can perform cost/benefit analysis on chip prices and this chip is going to cost a lot more that Intel (mobile).