Re: Pentium vs Celeron
From: cquirke (MVP Win9x) (cquirkenews_at_nospam.mvps.org)
Date: 05/13/04
- Next message: cquirke (MVP Win9x): "Re: Pentium vs Celeron"
- Previous message: Gordon: "Re: Opening attachments from email in Outlook Express"
- In reply to: Tim Slattery: "Re: Pentium vs Celeron"
- Next in thread: Alex Nichol: "Re: Pentium vs Celeron"
- Messages sorted by: [ date ] [ thread ]
Date: Fri, 14 May 2004 00:27:46 +0200
On Wed, 12 May 2004 09:11:52 -0400, Tim Slattery <Slattery_T@bls.gov>
>"cquirke (MVP Win9x)" <cquirkenews@nospam.mvps.org> wrote:
>>1) 512k half-speed L2 cache, 66MHz base
>>2) 512k half-speed L2 cache, 100MHz base
>>3) zero L2 cache, 66MHz base
>>4) 512k half-speed L2 cache, 100MHz base, SIMD
>>5) 128k full-speed L2 cache, 66MHz base
>>6) 256k full-speed L2 cache, 100MHz base, SIMD
>>7) 128k full-speed L2 cache, 100MHz base, SIMD
>>8) 256k full-speed L2 cache, 133MHz base, SIMD
>>9) 256k full-speed L2 cache, 400MHz base, SIMD
>>A) 128k full-speed L2 cache, 400MHz base, SIMD
>>B) 512k full-speed L2 cache, 400MHz base, SIMD
>>C) 512k full-speed L2 cache, 533MHz base, SIMD
>>D) 512k full-speed L2 cache, 800MHz base, SIMD, HT
>This is good stuff, but can you translate the acronyms? HT would be
>hyperthreading, I guess, but what's SIMD?
Sorry, my bad:
L2 = Level 2 cache; used to be on the motherboard in the 386 to
Pentium era, but since Pentium Pro has been in the processor.
Base speed is the speed at which the device is clocked. In the old
days, everything was locked to base speed, but the 486DXn started
running the core at a multiple of this speed. Since then, fractional
core multiplier have become common, other busses run at different
speeds, and multiple bits of data can be carried per clock pulse on
some busses such as AGP.
The RAM used to be run at base speed, but that's not always the case
these days - for example, Intel's 815 chipset used to run RAM at
100MHz even if the processor was based at 66MHz.
When clock-multiples started in the 486DXn era, I wondered whether a 2
x 25MHz would really approach a true 50MHz, etc. on the basis that the
processor core might have to wait for RAM, and thus waste the second
internal clock tic. In practice, performance is pretty close to the
internal speed, even when the multiplier from base speed becomes quite
extreme (e.g. double figures!).
The other kicker is that whereas 486DXn and Pentium had the L2 cache
on the motherboard and bound to the base speed, modern processors have
this inside the core and running at core speed. As L1 and L2 catch
most RAM accesses, the base speed serves to speed up only the small %
of RAM accesses that are missed and actually go to RAM. For the same
reason, faster RAM technologies often have less impact than expected.
SIMD = Single Instruction, Multiple Data; an extension of the basic
processor instruction set, much as MMX was in the original Pentium
era. Only Intel processors have SIMD, whereas all recent brands of
processors have MMX; AMD have their 3D-Now! alternative.
Unless programs are written to use these extensions to the instruction
set, they confer no benefit at all. Fortunately, several core
performance-sensitive parts of Windows (e.g. DirectX and drivers) can
use this code if it is supported, and thus benefit any software that
uses that OS subsystem - so it's not as useless as it sounds!
HT = HyperThreading, a new Intel technology that improves instruction
parallelism to the point that one processor begins to approach
dual-processor efficiency - as long as the program code is written
with an awareness of this in mind.
The overall effect of HT is expected to be less performance impact
from "underfootware" such as resident antivirus, firewall, and
whatever malware you have active at the time <g> Single powerhouse
applications such as video editing or graphics pckages are often
written to harness multiple processors and presumably HT.
>And is "base" the front side bus speed?
Umm... yes, he says cautiously ;-)
As I mentioned earlier, this isn't always the case. On the bright
side, 815 runs RAM at 100MHz when processor is 66MHz-based. On the
dark side, a new P4 capable of dual-DDR 800MHz to RAM on a POS
motherboard with 533MHz-limited chipset and 1 x DIMM isn't going to
dual-channel, and it isn't going to 800MHz. So much of the "Pentium
Tax" is down the drain; all you get is more L2 and HT.
It's easier to drop a P4 into a P4-capable motherboard than it is to
slide a P4-capable motherboard under a P4. Plus, it's cheaper to buy
a Celeron and good mobo than P4 and dud mobo. Plus, while good mobo
chipsets will fall around 20% in price, today's star P4 is likely to
fall to a third or quarter of today's price.
Add those up, and you can see why I prefer to build with good base
components (to which XP binds itself inextricably) and defer fancy
processors (which XP will generally accept quite easily).
>-------------------- ----- ---- --- -- - - - -
Tip Of The Day:
To disable the 'Tip of the Day' feature...
>-------------------- ----- ---- --- -- - - - -
- Next message: cquirke (MVP Win9x): "Re: Pentium vs Celeron"
- Previous message: Gordon: "Re: Opening attachments from email in Outlook Express"
- In reply to: Tim Slattery: "Re: Pentium vs Celeron"
- Next in thread: Alex Nichol: "Re: Pentium vs Celeron"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
