Pandora Vs. Eeepc 701 Processor Speed? Only For Curiousity

[(naw)mcx]

My main PC is a P3-s 1.4GHz(main one conked out) and my main laptop is a EEE701.

Any tests you want me to run?

(P3 preferable, no OS on my EEE atm)

 

[Exophase]

That's the point: to compare two different architectures with similar spec parameters.

Those "specifications" are all part of the architecture. Even then, your comparison was incomplete - when comparing caches you stopped at size and ignore latency, bus width, associativity, allocation policy, write back vs write through, virtual vs physical tagging, etc... just comparing size is naive. In addition to external memory bus width things may be a lot less similar than you think.

And I didn't understand your statement re ISA. CISC is more powerful AFAIK.

Oh god, where do I even start? This RISC vs CISC mentality is grossly outdated, it was probably outdated already when these things were coming out. x86 is optimized for code density in a lot of ways that may have made sense at the time but have proven themselves weak choices, especially as instruction set extensions have ended up chewing up a wider opcode space. Most of the more esoteric features of x86 go unused by the compiler, with the only real advantages over ARM being memory operands in ALU operations, larger immediates, and direct imm/reg + reg + imm addressing modes (there's also 8/16 bit arithmetic but this has proven dangerous to performance in a lot of scenarios on modern processors). Meanwhile, ARM has had since the beginning almost twice the general purpose registers, three address operations, more orthogonality (restriction to using cl for register shifts is the big remaining one on x86) barrel shifting on ALU operations, fully predicated execution, more general addressing modes (including pointer arithmetic writeback), block memory transfers, multiply accumulate, and so on... and that's ignoring the continued improvements made. Improvements in the x86 space have been almost exclusively in SSE over the last 15 years since Pentium Pro.

 

[Vitel]

That's the point: to compare two different architectures with similar spec parameters.

Those "specifications" are all part of the architecture. Even then, your comparison was incomplete - when comparing caches you stopped at size and ignore latency, bus width, associativity, allocation policy, write back vs write through, virtual vs physical tagging, etc... just comparing size is naive.

I'm not saying that my comparison is complete and comprehensive. But I think that size is one of the most important criteria for comparing. Besides we cannot choose most of the other architecture parameters even if we'd like to.

 

[Vitel]

In addition to external memory bus width things may be a lot less similar than you think.

It is 32-bit DDR (OMAP3) vs 64-bit SDR (P3), right?

 

[Exophase]

It is 32-bit DDR (OMAP3) vs 64-bit SDR (P3), right?

Not necessarily, but generally yes. There was an RDRAM chipset and later a DDR chipset for Tualatin P3s.

You should learn to multiquote btw. Replying to two separate people in consecutive posts can be overlooked but replying to the same post over two is really too much...

 

[SomeGuy99]

I think that size is one of the most important criteria for comparing.

Don't forget, good peformance is all about how you use it.

Wait... are we talking about CPU architecture?

 

[doublebeta]

I'm not actually sure what we're talking about anymore, i'm pretty sure we all got EXOwned. but no, really we were comparing two devices, the pandora and eePC 701.
It seems the EeePC is faster, but as has been said, even by myself, they're different systems that are designed to be used differently. Your octa-core 64bit system is going to still be crap if you're running 16-bit, single core only M$DOS (in fact i'm not even sure DOS would start). And your system would eat 2000x the power of the pandora. The pandora is a tiny device that gets an 8+ hour battery life, and the Eee is a relatively large device in comparison to the pandora, and got little more than 2 hours with no wifi and mid backlight when i was using mine.
Exophase, I actually DID think that Celeron was an archetecture! thanks for telling me it wasn't, i really was clueless i should research better :\. And what you said about the Atom really cleared things up too! I don't think Intel was going the right way, still, because they went back in core complexity, but seem to have forgotten that, while less complex, the older processor was still inefficient. IMO they should have started from scratch, even if that meant another f00f bug. But i'm going off topic.
(naw)mcx, thanks for the offer, but i think we should just leave it now . Nurbs might come along and say the pandora sucks if we probe any further. They won't get that the pandora is fanless and runs on miniscule amounts of power. Then again, these people don't deserve a pandora! .
Wish you all good luck, those waiting for their pandoras whom will have them arriving soon or are in the queue. And those who have them, i'm so damn jealous . Enjoy the games for me . Who else is waiting like hell for batch 2 preorders? I hope that doesn't start for a while yet, i gotta get some $ together before they're all taken in a nS.

oh yeah, and for anybody who's worried, that isn't a complete noob, it really DOES NOT MATTTER if the statistics say that the P3 is faster. Or even if it is in reality. The only reason we're all not using P3s now is because of programs becoming slower (for no good reason). If people can make decent programs for the Pandora, it will be a much better machine than a P3 box. I think we can all blame it on M$ actually...

 

[Exophase]

I don't think Intel was going the right way, still, because they went back in core complexity, but seem to have forgotten that, while less complex, the older processor was still inefficient. IMO they should have started from scratch, even if that meant another f00f bug.

Atom is a new design for Intel, it's not a modified Pentium 1 or anything like that. Both cores are dual-issue in-order, but that applies to a lot of other cores too. In every other way it's very different from P1 - much longer pipeline, very different sorts of stalls and latencies, SSE1-SSSE3, only one load/store unit instead of two, much more/wider cache, hyper threading, gimped x87 support, and so on.

Atom is really Intel's attempt to eventually enter the handheld space ARM dominates in. Taking netbooks instead was just a nice stop along the road; IMO netbooks never really had to be Atom (in fact, the original Eee you bring up proves this). Ultimately, the smaller/lower clocked Atom and its support chipsets get, the more the unavoidable price paid for supporting x86 over supporting other ISAs starts to hurt you. In the desktop world this stopped being an issue as die sizes got proportionately much larger. In the handheld world it's still an issue, and gives ARM an advantage over the long term, and is why they can afford to push Cortex-A9 which is a more sophisticated chip.