>No branch mispredictions at all. Mommy, Intel isn't plaing nice again! 8^) But, hopefully, this does mark somewhat a rather useful advance... we'll see how everything turns out. Benchmarks aren't fair to everyone - I'll live. The idea was to point to a benchmark that attempts to be less biased than most. You win some, you lose some.
>The UltraSPARC has always underperformed the Alpha on SPEC, and yet does much better in the marketplace.
And part of it is due to the systems that are built around it. Sun's I/O is dramatically better than that of most Alpha boxes. Always has been, and given the way that Compaq has treated the Alpha... always will be. A system with great I/O (like S/390 or AS/400) will continue to be more valuable in the mass marketplace than a system which prides itself on FP performance. Lots of database transactions... hmmm, mostly integer, and a lot that doesn't involve the processor at all.
Sun says 'we power the web', DEC/Compaq say 'I think we have some resellers, and I hope *they* try to sell Alphas'. The Alpha has endured a long history of poor marketing... great R&D (until recently, but maybe that's picking up again), but horrible product placement...
I'll certainly admit that the newer x86 chips can certainly outduel the PowerPC in terms of raw speed... Athlon 900 - 37.8 Int, 27.9(base) FP (oddly, the 1GHz lists at 42.9 Int (no FP listed)). They list the compiler and all of the optimization flags, and the change in flags between base and peak for all runs...
And yeah, there are ways around any benchmark... but that's why various tests get kicked out of the spec set, and why new ones are put in. Not a perfect system by any stretch, but a little better than one designed by the processer designers 8^)
My use of the PPC is all embedded, I've never owned a Mac. I personally don't care what happens to Apple (aside from the usual tripe about competition, but face it, the Intel/AMD war has pushed those two far faster than it can force Apple to beg for faster procs...), but the Power architecture is goodness, if not fully realized in Mac systems.
True, true, true... but apps like Photoshop that support MP while applying filters and the like can gain quite a bit from this. As usual, servers gain more than just about anybody.
I suppose it would be useful to encode some mp3s on one proc while image manipulating or whatever elsing on the other... and of course, Q-III (and whatever else) smp abilties get people fairly happy...
specbench.org lists a Dell 450MHz P-II: SPECfp95 13.1 SPECfp_base95 11.8 Yup, an estimated 21.4 and 20 aren't that much better...(~63.3% for Int and ~72.9% better for FP) Fairly dramatic, though... Note that a 466MHz Alpha 21164 rated a 19.2 specFP a 466MHz 21264 rates a 24.6...
So, much better than the IA-32 - close to the Alpha...
I think most of Apple's propaganda is pretty funny, but hey - somebody's gotta do it 8^)
Photoshop benchmarks between G4 and x86 would be useful, but the reference to Linux doesn't make too much sense. It's the hardware, not the OS that seems to be debated at the moment (so Windows vs. Mac vs. Linux strikes again, oh well)...
[Score:-25 ; Shameless Self-promotion] >Tower wins with a First Post plus a first B post.
>Is this a/. record?
Gee, I hope so - I've always wanted to win something. I was really hoping for the Pwerball or Big Game, but I'll take a few negative karma points instead.
I'd like to thank the people who made this possible...[random blathering, contributing to the 'N' in S/N...]...
Thank you. (for some reason, I feel guilty writing that last bit... (where is good old Don Knotts?))
>In the hands of good programmers, I find this vastly superior to pre-emptive (OS controled) multitasking
I'd agree (dot dot dot)... In the hands of anyone who is fallible, or doesn't fully test their code, this is a death sentence. Don't get me wrong, I wish it was that simple, but a good rule of programming (especially for lower-level OS design) is the same as when driving a car: Assume everyone else on the road (system) is an idiot. You usually end up being right, and you should never depend on anyone slowing down so that you can merge (or take control of your own OS back).
As Homer put is so eloquently "...in theory... Communism works - in THEORY."
Perfecly ideal worlds are hard to come by.
(note that I am in now way stating my opinions for or against communism, democracy, totalitarian oligarchy or any other form of government...)[/lame disclaimer]
The memory contention is dealt with via 'snoops' on PowerPCs (probably a similar mechanism on the Intel Arch). For example, if you have a PCI network card DMA'ng some data to main memory, the chipset can issue a snoop to the processor, asking if he has any data from that address. The processor can then cast out data from his internal cache, and the chipset can combine the data with the incoming PCI write, so coherency is maintained. The actual process and timing is interesting, but the system is fairly simple from a top level view. With extra processors, this can be an additional burder, and could get rather interesting...
Interrupts... well... there's a nasty rat's nest... I'll leave that one alone for now. Suffice it to say, the PPC has a few different kinds of interrupts (including an SMI)... Not sure about the full SMP actions there, but again, it all depends on the OS structure.
Asymetrical MP is so much *easie*, when you think about it... but it's not very useful with only 2 procs...
Hmmm... well, remember (like the Alpha), GCC isn't optimized as well for the PowerPC line as it is for the Intel Architecture (more people, more time). Of course, that isn't the whole story (nothing ever is).
That aside, I still can't find any specInt or specFP #s... which are a fairly good representation...
Yup, that's the problem with a fairly closed hardware platform... there's been clone PCs for a long time, and any Joe can slap one together, but very few can build a Mac (where do I get that ROM again...).
OTOH, the hardware is something to be proud of (note my self-indulging first post). Clock-for-clock, these babys really cook (and with less heat than a lot of other chips [cough]). A great chip - I just wish I could build a system with them myself...
Intel chips work on a shared bus architecture - real easy, like any other bus. Request and grant. Each processor requests bus mastering, and they share. Nice, but one (or more) can get held off.
The Athlon is based on the same SMP style as the Alphas (see earlier/. article today). Each proc gets it's own dedicated path to the chipset, where there's a switching fabric. This is essentially the same difference between a LAN hub and a switch. One Athlon could be bursting to memory at peak bandwidth while another talks to the AGP bus... good stuff. You only have to wait if the actual resource you are waiting for is being used. This is considerably more dificult (and more costly) to implement than a shared bus.
The G4 can use a shared bus (as the earlier PowerPCs could. I'm not privy to the specifics of this new chipset, but it stands to reason that anything priced in the consumer marketspace (ooh, buzzword!) would need to be a shared bus to stay cost-efficient.
The K6 was SMP capable, but not with Intel's chipsets and SMP structure. Therefore, it is essentially non-SMP capable, since nobody supported it. The Athlon has yet to be proven in this arena, but several boards are in the works (or so the rumor mill would have us believe)...
seriously - these are some really kicking processors, and the box would look nice next to the Alpha and K6 boxen. I'm still waiting for somebody to make some generic PPC motherboards (ATX, preferably). I'm a big fan of the hardware, but never been nuts about Macs in general - though they have turned my head a few times, I still think the system stability needs some work (as do they all). I might have to play with OSX sometime...
(oh, did I mention a Beowulf cluster of these would rock)
I'm presently using a Logitec Trackman Marble (basic - no scroll wheel). I used the scroll wheel for a while, and I love it for some things, but it really bothers me sometimes. Oh well. I had been using a Microsoft Mouse (2.0a) before that. Only two buttons, but, to this day, the best mouse I've ever laid palms on. Nice, solid feel, good shape, good motion. The trackball kept my wrist from getting sore, and I don't have problems with dirty rollers, etc.. Plus, I can use more of my workspace for my other clutter 8^)
The trackball is a little tough to pick up in a day, since you have to fiddle with the sensitivity for a while in normal 2-D land to get used to it, but it's well worth it (if you spend a lot of time at your computer, I'd say about two-four days isn't an unreasonable curve. There are some moves I've been able to do with a trackball that I never accomplished with a mouse - I like the speed and precision. Of course, there were many who were better at Q-II than I was, but I managed a fair number of frags on the stanford fvf server (back in the day)...
Windows-E - is that like Windows CE? Some of us don't have a keyboard that has those 'other' keys (too old - 1995, or the fact that they just get in the way... that's the handrest area of the KB).
>/. is, in effect, a town hall meeting. Anything you say is on the record. You have the right to speak up or to keep your mouth shut and just read to what's being written.
Yup. That's pretty much it. If you don't want it read (either via/. , Wired, dead tree, or whatever), then don't write it. You have to be aware of your enviornment.
>If you don't want people to associate you with what you've said, post as an AC. If you don't trust the journalists that run the site, don't post at all.
If I were prone to sudden outbursts, I might yell "Amen, brother!" for that last bit. But I'm not, so I'll just say: Nothing else needs be said.
When you post anonymously, you are giving up all rights to what you've said. You don't want to be associated with those words, or you would have logged in. (or you hate cookies.. or whatever). The point is, if you post via AC, it's not attached to you, so all reprinting rights are up to/. at that point. If you care, log in.
Slashdot Mirror
>No branch mispredictions at all.
Mommy, Intel isn't plaing nice again! 8^) But, hopefully, this does mark somewhat a rather useful advance... we'll see how everything turns out. Benchmarks aren't fair to everyone - I'll live. The idea was to point to a benchmark that attempts to be less biased than most. You win some, you lose some.
>The UltraSPARC has always underperformed the Alpha on SPEC, and yet does much better in the marketplace.
And part of it is due to the systems that are built around it. Sun's I/O is dramatically better than that of most Alpha boxes. Always has been, and given the way that Compaq has treated the Alpha... always will be. A system with great I/O (like S/390 or AS/400) will continue to be more valuable in the mass marketplace than a system which prides itself on FP performance. Lots of database transactions... hmmm, mostly integer, and a lot that doesn't involve the processor at all.
Sun says 'we power the web', DEC/Compaq say 'I think we have some resellers, and I hope *they* try to sell Alphas'. The Alpha has endured a long history of poor marketing... great R&D (until recently, but maybe that's picking up again), but horrible product placement...
Oh well...
The RS/6k numbers are fairly impressive 8^)
I'll certainly admit that the newer x86 chips can certainly outduel the PowerPC in terms of raw speed... Athlon 900 - 37.8 Int, 27.9(base) FP (oddly, the 1GHz lists at 42.9 Int (no FP listed)). They list the compiler and all of the optimization flags, and the change in flags between base and peak for all runs...
And yeah, there are ways around any benchmark... but that's why various tests get kicked out of the spec set, and why new ones are put in. Not a perfect system by any stretch, but a little better than one designed by the processer designers 8^)
My use of the PPC is all embedded, I've never owned a Mac. I personally don't care what happens to Apple (aside from the usual tripe about competition, but face it, the Intel/AMD war has pushed those two far faster than it can force Apple to beg for faster procs...), but the Power architecture is goodness, if not fully realized in Mac systems.
ah... missed that one 8^)
True, true, true... but apps like Photoshop that support MP while applying filters and the like can gain quite a bit from this. As usual, servers gain more than just about anybody.
I suppose it would be useful to encode some mp3s on one proc while image manipulating or whatever elsing on the other... and of course, Q-III (and whatever else) smp abilties get people fairly happy...
>http://199.104.132.208/brdata/PDFDB/MICROPROCESSO RS/32_BIT/POWERPC/MPC7XX/M951446479748co llateral.pdf
Thanks for the link.
specbench.org lists a Dell 450MHz P-II:
SPECfp95 13.1
SPECfp_base95 11.8
Yup, an estimated 21.4 and 20 aren't that much better...(~63.3% for Int and ~72.9% better for FP)
Fairly dramatic, though...
Note that a 466MHz Alpha 21164 rated a 19.2 specFP
a 466MHz 21264 rates a 24.6...
So, much better than the IA-32 - close to the Alpha...
I think most of Apple's propaganda is pretty funny, but hey - somebody's gotta do it 8^)
Photoshop benchmarks between G4 and x86 would be useful, but the reference to Linux doesn't make too much sense. It's the hardware, not the OS that seems to be debated at the moment (so Windows vs. Mac vs. Linux strikes again, oh well)...
[Score:-25 ; Shameless Self-promotion]
/. record?
>Tower wins with a First Post plus a first B post.
>Is this a
Gee, I hope so - I've always wanted to win something. I was really hoping for the Pwerball or Big Game, but I'll take a few negative karma points instead.
I'd like to thank the people who made this possible...[random blathering, contributing to the 'N' in S/N...]...
Thank you.
(for some reason, I feel guilty writing that last bit... (where is good old Don Knotts?))
>In the hands of good programmers, I find this vastly superior to pre-emptive (OS controled) multitasking
I'd agree (dot dot dot)... In the hands of anyone who is fallible, or doesn't fully test their code, this is a death sentence. Don't get me wrong, I wish it was that simple, but a good rule of programming (especially for lower-level OS design) is the same as when driving a car: Assume everyone else on the road (system) is an idiot. You usually end up being right, and you should never depend on anyone slowing down so that you can merge (or take control of your own OS back).
As Homer put is so eloquently "...in theory... Communism works - in THEORY."
Perfecly ideal worlds are hard to come by.
(note that I am in now way stating my opinions for or against communism, democracy, totalitarian oligarchy or any other form of government...)[/lame disclaimer]
All very good points - mod this one up a few...
The memory contention is dealt with via 'snoops' on PowerPCs (probably a similar mechanism on the Intel Arch). For example, if you have a PCI network card DMA'ng some data to main memory, the chipset can issue a snoop to the processor, asking if he has any data from that address. The processor can then cast out data from his internal cache, and the chipset can combine the data with the incoming PCI write, so coherency is maintained. The actual process and timing is interesting, but the system is fairly simple from a top level view. With extra processors, this can be an additional burder, and could get rather interesting...
Interrupts... well... there's a nasty rat's nest... I'll leave that one alone for now. Suffice it to say, the PPC has a few different kinds of interrupts (including an SMI)... Not sure about the full SMP actions there, but again, it all depends on the OS structure.
Asymetrical MP is so much *easie*, when you think about it... but it's not very useful with only 2 procs...
Hmmm... well, remember (like the Alpha), GCC isn't optimized as well for the PowerPC line as it is for the Intel Architecture (more people, more time). Of course, that isn't the whole story (nothing ever is).
That aside, I still can't find any specInt or specFP #s... which are a fairly good representation...
Anybody else know of them (and have a source)?
Almostthereware?
WhereWare?
Yup, that's the problem with a fairly closed hardware platform... there's been clone PCs for a long time, and any Joe can slap one together, but very few can build a Mac (where do I get that ROM again...).
OTOH, the hardware is something to be proud of (note my self-indulging first post). Clock-for-clock, these babys really cook (and with less heat than a lot of other chips [cough]). A great chip - I just wish I could build a system with them myself...
Intel chips work on a shared bus architecture - real easy, like any other bus. Request and grant. Each processor requests bus mastering, and they share. Nice, but one (or more) can get held off.
/. article today). Each proc gets it's own dedicated path to the chipset, where there's a switching fabric. This is essentially the same difference between a LAN hub and a switch. One Athlon could be bursting to memory at peak bandwidth while another talks to the AGP bus... good stuff. You only have to wait if the actual resource you are waiting for is being used. This is considerably more dificult (and more costly) to implement than a shared bus.
The Athlon is based on the same SMP style as the Alphas (see earlier
The G4 can use a shared bus (as the earlier PowerPCs could. I'm not privy to the specifics of this new chipset, but it stands to reason that anything priced in the consumer marketspace (ooh, buzzword!) would need to be a shared bus to stay cost-efficient.
The K6 was SMP capable, but not with Intel's chipsets and SMP structure. Therefore, it is essentially non-SMP capable, since nobody supported it. The Athlon has yet to be proven in this arena, but several boards are in the works (or so the rumor mill would have us believe)...
(after First Post)
seriously - these are some really kicking processors, and the box would look nice next to the Alpha and K6 boxen. I'm still waiting for somebody to make some generic PPC motherboards (ATX, preferably). I'm a big fan of the hardware, but never been nuts about Macs in general - though they have turned my head a few times, I still think the system stability needs some work (as do they all). I might have to play with OSX sometime...
(oh, did I mention a Beowulf cluster of these would rock)
Drool...
It's not a beer if you can read your watch through it!
Guinness.... mmmmm...
Yeah, it doesn't seem to bother my NT4 box (or, of course, my AIX workstation)...
I'm presently using a Logitec Trackman Marble (basic - no scroll wheel). I used the scroll wheel for a while, and I love it for some things, but it really bothers me sometimes. Oh well. I had been using a Microsoft Mouse (2.0a) before that. Only two buttons, but, to this day, the best mouse I've ever laid palms on. Nice, solid feel, good shape, good motion. The trackball kept my wrist from getting sore, and I don't have problems with dirty rollers, etc.. Plus, I can use more of my workspace for my other clutter 8^)
The trackball is a little tough to pick up in a day, since you have to fiddle with the sensitivity for a while in normal 2-D land to get used to it, but it's well worth it (if you spend a lot of time at your computer, I'd say about two-four days isn't an unreasonable curve. There are some moves I've been able to do with a trackball that I never accomplished with a mouse - I like the speed and precision. Of course, there were many who were better at Q-II than I was, but I managed a fair number of frags on the stanford fvf server (back in the day)...
Ah, the same arguments for my Trackball!
oh well. I find the trackball to be far more accurate after a little practice.
I used Enable Virtual Desktop. Not bad. Shareware. Along with a few other hacks, you can almost get your windows desktop as useful as fvwm!
Alt-ESC to lower a window?! No mouse click?!
Windows-E - is that like Windows CE?
Some of us don't have a keyboard that has those 'other' keys (too old - 1995, or the fact that they just get in the way... that's the handrest area of the KB).
Posting on an old thread... oh well...
Everybody over here uses EZ, except for those who use emacs. In fact, ATK runs on NT now too (not just on various Un*x varients).
>/. is, in effect, a town hall meeting. Anything you say is on the record. You have the right to speak up or to keep your mouth shut and just read to what's being written.
/. , Wired, dead tree, or whatever), then don't write it. You have to be
Yup. That's pretty much it. If you don't want it read (either via
aware of your enviornment.
>If you don't want people to associate you with what you've said, post as an AC. If you don't trust the journalists that run the site, don't post at all.
If I were prone to sudden outbursts, I might yell "Amen, brother!" for that last bit. But I'm not, so I'll just say:
Nothing else needs be said.
When you post anonymously, you are giving up all rights to what you've said. You don't want to be associated with those words, or you would have logged in. (or you hate cookies.. or whatever). The point is, if you post via AC, it's not attached to you, so all reprinting rights are up to /. at that point. If you care, log in.
>I would disable it--except that in Netscape it disables css too
style sheets - the cause of more problems...