Why does the Celeron beat the living shit out of other CPU\'s

2002-01-21 15:30:52

Check this out:

http://www.tomshardware.com/cpu/02q1/02012...ron1300-06.html

Why does the Celeron own every other CPU in this test? It makes no sense to me. And it's even by a significant margin.

I know they are using Lame 3.89 for the tests, but still. I know quality and speed has changed since the 3.89 times, but has the relative speed difference between processors changed?

What's the explanation for this? :confused:

Regardless, there's no way I'd even consider buying a Celeron.

Why does the Celeron beat the living shit out of other CPU\'s

Reply #1 – 2002-01-21 16:22:45

Speculation:
These chips are produced in a new 0.13 micron fab, which in itself says little, however, it also means more on die cache (and/or perhaps better associativity).  As mp3 coding works in fairly small chunks, it's quite possible that the celeron essentially never accesses main memory in it's optimal solution search, but only to get new data.  The others have slightly smaller caches so for problems that are just too large for their caches but just fit inside the celeron's the celeron get's a BIG boost.  If both memory and cache have 2 cycles latency and the internal clock has a 10x multiplier, then that's a data latency reduction to 1/10 of it's original value (in the optimal scenario for the celeron which is absurdly unlikely to happen)  lame probably comes unlikelily close to this scenario in which the celeron hits and the other miss on cache.  Coincidence...  though amusing

(again, I think this is probably what's happening, but I don't know the new celerons exact cache size+type, so this may be complete humbug.)

Why does the Celeron beat the living shit out of other CPU\'s

Reply #2 – 2002-01-21 16:27:34

Sounds plausible. The Duron has a very small L2 cache. But the P4 and XP CPU's do have a L2 cache that is on par with the Celeron (AFAIR). The matter still puzzles me. The AMD line of CPU's even has the benefit of 3Dnow! optimized code. Don't know how much of an advantage that is though.

By the way, the specifications of the Duron 1300 and the Celeron 1300 are on page 2 of the review.

Why does the Celeron beat the living shit out of other CPU\'s

Reply #3 – 2002-01-21 16:43:25

I just checked out the specs on P4's, the old ones have the same amount (256KB) of cache.  Though P4's are weird anyway.  As to athlons... Those actually have more cache 256+64KB!! (Athlon cache is exclusive: data in L1 is not in L2, unlike intel chips).

Ah, here we go, the celeron has a 256 bit L2 data width, and the AMD's only 64 bit.  The code probably isn't P4 optimized (not many open source P4-freaks :-) )

Maybe.  Small specialized code fragments are really hard to analyze and compare...

Why does the Celeron beat the living shit out of other CPU\'s

Reply #4 – 2002-01-21 16:49:11

Dammit.

The Celeron's destiny is to suck! It must suck.  I hate seeing it beat every other CPU in the test. :mad:

"Could we have some optimizations over here, please?"

Why does the Celeron beat the living shit out of other CPU\'s

Reply #5 – 2002-01-21 17:52:49

I think the new Tualatin Celerons have hardware pre-fetch ability and perhaps this is where the performance gains are being made.

My 0.18u Celeron-II typically performs well in audio compression applications. Video is another matter.