A paragraph about Cell

A paragraph about Cell

Summary: Cell crunchesthrough millions of lines of topographical and photographic data per second

SHARE:
TOPICS: Hardware
50
A Forbes article last week by Daniel Lyons under the title Holy Chip! included this paragraph about an IBM cell processor demo:

An IBM demo shows the contrast. A terrain rendering program lets you fly over Mount Rainier at 1,300mph. Cell crunches through millions of lines of topographical and photographic data per second to paint topographically accurate, photo-quality pictures at a movie-quality 30 frames per second. On a similar program a Pentium takes more than two minutes to sketch a single frame.

The article is worth reading, that paragraph above is worth reading again. Think about what this thing is going to do when applied in engineering or architectural design, bio-research, and chemical modelling.

Topic: Hardware

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.

Talkback

50 comments
Log in or register to join the discussion
  • That's gotta scare the pants off NVidia and ATI

    ;-)
    No_Ax_to_Grind
    • More likely to scare the pants off of Microsoft. They rushed their product

      to market which is a plain Jane PowerPC with multiple cores. If the games for the Sony PS3 are head and shoulders better than XBox because of the ability to do photo quality real-time rendering, sales of XBox will be falling like a rock.
      DonnieBoy
      • but what about cost ?

        Yeah, that might be true, but really, how much can they afford to lose with and at what price would Sony have to charge for PS3 ?

        And how will the processor actually work in a regular/business computing world. The question is really, would it and could it really challenge Intel/AMD (the x86's - don't forget VIA on the low power end) in regard to price/quantity ? As long as it can't in this two regards. Then this is going to be one of those high price niche item.

        Remember, Apple just switched over to the x86 because IBM couldn't get them the cost/quantity that they need. I'm sure that to save Apple from switching, IBM must have shown them the Cell and Appl decided that it just wasn't "enough" for them to stick it out with the PowerPC line.

        Remember, just raw horse power doesn't make a company (think Cray and Thinking Machine back in the 80's).
        JJ_z
      • I don't think MS is worried at all.

        Face it, once you reach a certain level of quality in the display it really no longer matters. I have a frined that is big time into gaming and he boasts his machine can crank out 90 frames a second. My reply is so what, you can't see them. It's all about how the human eye works...
        No_Ax_to_Grind
        • High frame rates are important

          because the game does not always stay at 90 FPS. When it dips below 30, that's when it's really noticeable.
          The higher FPS you can get, the more assurance you have of avoiding the low frame rates during the game.

          There was a graphics review last year that took a different measure in chipsets. It counted the average time spent below a minimum threshold and how noticeable it was. It was one of the big factors that pushed me to the AMD side for my new machine.
          Joeman57
      • M$ isn't scared

        M$ uses the Xenon chip - which is like Cell but only 3 cores instead of 8. If it is hard to program for multiple cores (parallel processing programming is HARD), then 3 or 8 doesn't really matter.

        Windoze for Xbox360 SHOULD run on the Cell anyway . . .
        Roger Ramjet
  • what was it, we were discussing yesterday?

    Reading this part of the article reminds me of the blog of yesterday, which was about using M$ or Linux software for dba functions and especially the way it was configured for benchmarks... by M$.

    Who was responsible for the configuration and who TESTED/BENCHMARKED these configurations against eachother.
    Arnout Groen
    • No where did anyone say this was a bench test.

      It was a specific application written to run on the Cell. But putting that to the side, an x86 CPU (pick your flavor) could not come anywhere close to this level of performance. It would be so far out of it's league there would be no comparison. It would be like bench testing a Yugo against a Porche. Why would anyone bother???
      No_Ax_to_Grind
      • More Amiga references

        THE reason that the Amiga was so much faster than a comparable x86 machine was its co-processors. The Motorola 68000 chip is INTERRUPTABLE whereas the x86 is not. This means that the Amiga could switch between its different co-processors whenever a task required. The x86 HAS to "stay in charge" and control everything - thus becomming the limiting factor on speed of operations. This is why SCSI is faster than IDE - IDE requires the CPU to control it, whereas SCSI does not.

        Amiga advocates called for an Amiga-on-a-chip for many years. Looks like IBM created one . . .
        Roger Ramjet
        • Not really, but that's ok.

          Yes I liked the Amiga, but that was then, this is now. The very nature of the design of the Cell and HOW YOU WRITE CODE to it is the deciding factor here. THe best comparison would be a MoBo with 9 x86 CPUs.
          No_Ax_to_Grind
          • Can't learn from history?

            You mean the Amiga, that had separate processors for sound and graphics, didn't have any parallels with the Cell in terms of programming? Creating a program on the Amiga that took advantage of the multi-processor, parallel processing was a skill that has direct application here.
            Roger Ramjet
          • In the losest of possible terms, sort of.

            The biggest difference of course is that in the Amiga the co-processors were dedicated to specific tasks. More like Intel's DPS than anything else.

            With the Cell each SPE can be assigned any task or have multiple SPEs assigned to the the task. In fact properly written code will test how busy an SPE is and balance the processing load.

            As an example, lets say you have an SPE driving video output. Now you recalculate a large spreadsheet. The video isn't changing during the calculation so there is no reason for that SPE to be sitting idle. In such a case you might throw all 8 SPEs at the immediate task.

            The other HUGE difference is in the communication bus. Really need raw power, put two Cells in one box (or connect boxes) and you now have 16 SPEs (all capable of talking to each other and sharing the load) as well as two 64-bit main processors that are anything but slow.

            The possibilities stagger the mind...
            No_Ax_to_Grind
          • Hardware-wise I would agree

            The Amiga used dedicated processors and the Cell uses general purpose processing cores.

            But what about the SOFTWARE? The similarity is striking between the two. How do you parallelize your application to utilize other processors? You still need to solve that same issues that Amiga developers faced. Some of that work was done by the AmigaOS, and some was done with compilers - but the biggest gain for speed lies with the programmer and his ability to parallel program.
            Roger Ramjet
          • Yes, coders will need to be retrained.

            It is a lot more than parallel processing. To be honest, I fear this need for retraining may be the thing that keeps the Cell from reaching it's potencial.

            Can you imagine a single box running Linux, Unix, Windows, and OS X all at the same time and working TOGETHER? All doable with the Cell...
            No_Ax_to_Grind
          • Can I imagine?

            I ran AmigaOS, MacOS and Windoze on my A2000 Amiga at the same time. The communication between them all was pretty rudimentary . . .
            Roger Ramjet
        • Which x86 would you compare to the processor used by your Amiga?

          an i386? And a slow one I'd bet. A i8088 is faster than a z80. We are talking about ancient history, correct? I'll bet you that the IDE drive in my oldest PC is faster than the SCSI that you had in your Amiga.

          The Amiga was great in it's day, but so was the C64 and they both belong in a museum. The Amiga was an adult toy, the expensive big brother to the Commodore 64. I'd think that the developers would find your Amiga on a chip comment insulting, There is no real comparison between the two architectures.
          balsover
          • Amiga education

            The Amiga was the best selling computer in Europe - it outsold PCs by a wide margin. I'm not sure if that qualifies as a toy (in the US, HAM radio operators were the early adoptors).

            Actually, when Amiga came out with SCSI, IDE didn't exist. You had MFM and RLL - and those technologies were REALLY slow - much worse than the SCSI 1 of the Amiga.

            If you don't see the connection between a computer that has separate chips for processing - CPU, graphics, sound - and the 8 core Cell that has separate cores for, well processing - then I don't know how to explain it to you.
            Roger Ramjet
          • apples and oranges

            The amiga wasn't designed like a PC with one central processor responsible for everything.

            Now if you want to compare the main MC68000 to an x86, the right one would be the 80386 since it was the first one with large memory (well, larger than 64k anyway) capacity and the ability to hold an interupt long enough to get a response from an 80387.
            murph_z
          • Z80 vs 8088

            Back in the day I remember doing speed tests on an original IBM PC (4.77MHz 8088) and a Radio Shack TRS-80 Model 4 (4Mhz Z80). The Z80 was actually faster. :-)
            mosborne
      • not my point..

        I used to run applications (all options enabled!)meant for 1.2 Ghz or higher on a 380 mhz system and it did well.

        So configurations DO MATTER when talking about performance... The fact that the software was written speciffically for the CELL does increase the performance on the processor, but there's a line I would like you to consider...

        (from FIA F1 racing) When the flag drops (end of the race), the bullshit stops.

        Meaning you can drive a Porsche in the race, but when the engine fails, the Yugo might/will (still) win the race
        Arnout Groen