Looking at my results and then comparing to the other host, I am certain that the differences between windows and linux is down to size and use of the L2 cache.
There are similar time differences under windows at similar clock speeds but differences in L2 cache.
This unit wuid=33662485 has been crunched under winXP on two C2D cpu's. Their speeds are only ~10% different, but the L2 cache on the E6400 is 2MByte but on the E6600 it is 4MByte.
But the competion times are very different,
E6400 - 107,386 sec
E6600 - 64,985 sec
I know it is only one comparison and there can be other factors involved, But unit 33623772, on same computers, show similar times as unit linked.
And comparing an AMD 4 3400 (clk 2100) L2 cache 512K to a PentM 1.86 L2 cache 2MByte, in wuid=33619879 the AMD takes nearly 50% longer, again WinXP.
I'm pretty sure that Linux manages small L2 cache better than Windows, but that if you have large L2 cache the differences between Linux and Windows is minimal.
Andy
I'm not so sure. The difference on my Core Duo is at least as big as on Gary's Coppermines, and I'm convinced they have less L2 cache (at least some of them, maybe not the Xeons, but our P3 has only 256 KB so I'm reasonably sure his won't all have 1 MB per core, as my CPU has, or more).
Maybe they're still figuring out what the problem really is... it seems to be quite difficult to even find out why certain CPUs are affected and others are not. Atm it seems to concern:
- all AMD CPUs
- all P3 architectures
- Core
The only CPUs known to be not affected are:
- P4 (only the Williamette or others as well?)
- Pentium M (maybe the enormous L2 cache really plays a role here?)
I'll be damned if I see a pattern. It must be more than just L2 cache, or Opterons would do okay and P4 probably wouldn't... And if it's Intel's on-chip-optimizer, why do Gary's and my Intel CPUs get hit about as badly as any AMD? How "modern" a CPU is can't play a role either, because the Pent M is newer than the P3 but older than the Core...
I'm not so sure. The difference on my Core Duo is at least as big as on Gary's Coppermines, and I'm convinced they have less L2 cache (at least some of them, maybe not the Xeons, but our P3 has only 256 KB so I'm reasonably sure his won't all have 1 MB per core, as my CPU has, or more).
The L2 cache for cpu's is
Coppermine 256KB @ 1/2 clk speed
Coppermine S 512KB @ 1/2 clk speed
My Pent M single core 2MB
My C2D 4MB
AMD's and this is being a bit general
Skt A 256
Skt 754 512
Skt 939 512 or 1024
Skt AM2 1024 or 2048
Your T2060 is an odd beast, it is a dual core mobile processor but it doesnt use the Duo name which have 2MB, but is not a Celery or a P3M
Two cores L2 1MB, it is not 2 * 512, so if one core is idle the busy core can use all.
If anyone is interested, the Linux distro I chose to use is PCLinuxOS 2007 TR-4. Check it out on Distrowatch. It's #3 on the Page Hit ranking and rising fast. Having installed it a few times now over the last couple of days I can really see why people like it :). It's really the first Linux distro I've ever used and it's very user friendly (for Linux/unix) :). I have the advantage of being familiar with BSD unixes and the unix CLI but the KDE gui of PCLOS is so easy for windows users to understand that this is a perfect way to introduce yourself to the joys of unix.
Actually, the nice thing about PCLinuxOS 2007 is its multimedia capabilities. Other Linux distro's can be made multimedia capable after you install them, but PCLinuxOS is multimedia ready with a default installation.
Having said that, for regular production-type work, I prefer something like Ubuntu, Debian 4.0, or one of the Red Hat derivatives. That's because the PCLinuxOS repositories still lack the productivity software that I need for my work.
Looking at my results and
)
Looking at my results and then comparing to the other host, I am certain that the differences between windows and linux is down to size and use of the L2 cache.
There are similar time differences under windows at similar clock speeds but differences in L2 cache.
This unit wuid=33662485 has been crunched under winXP on two C2D cpu's. Their speeds are only ~10% different, but the L2 cache on the E6400 is 2MByte but on the E6600 it is 4MByte.
But the competion times are very different,
E6400 - 107,386 sec
E6600 - 64,985 sec
I know it is only one comparison and there can be other factors involved, But unit 33623772, on same computers, show similar times as unit linked.
And comparing an AMD 4 3400 (clk 2100) L2 cache 512K to a PentM 1.86 L2 cache 2MByte, in wuid=33619879 the AMD takes nearly 50% longer, again WinXP.
I'm pretty sure that Linux manages small L2 cache better than Windows, but that if you have large L2 cache the differences between Linux and Windows is minimal.
Andy
I'm not so sure. The
)
I'm not so sure. The difference on my Core Duo is at least as big as on Gary's Coppermines, and I'm convinced they have less L2 cache (at least some of them, maybe not the Xeons, but our P3 has only 256 KB so I'm reasonably sure his won't all have 1 MB per core, as my CPU has, or more).
RE: (or do we already agree
)
Nope, they blow... ;-)
Brian...just being disagreeable
P.S. - Bernd? Akos? Any luck at getting us AMD/Win users back to at least not having a penalty?
Maybe they're still figuring
)
Maybe they're still figuring out what the problem really is... it seems to be quite difficult to even find out why certain CPUs are affected and others are not. Atm it seems to concern:
- all AMD CPUs
- all P3 architectures
- Core
The only CPUs known to be not affected are:
- P4 (only the Williamette or others as well?)
- Pentium M (maybe the enormous L2 cache really plays a role here?)
I'll be damned if I see a pattern. It must be more than just L2 cache, or Opterons would do okay and P4 probably wouldn't... And if it's Intel's on-chip-optimizer, why do Gary's and my Intel CPUs get hit about as badly as any AMD? How "modern" a CPU is can't play a role either, because the Pent M is newer than the P3 but older than the Core...
RE: I'm not so sure. The
)
The L2 cache for cpu's is
Coppermine 256KB @ 1/2 clk speed
Coppermine S 512KB @ 1/2 clk speed
My Pent M single core 2MB
My C2D 4MB
AMD's and this is being a bit general
Skt A 256
Skt 754 512
Skt 939 512 or 1024
Skt AM2 1024 or 2048
Your T2060 is an odd beast, it is a dual core mobile processor but it doesnt use the Duo name which have 2MB, but is not a Celery or a P3M
Two cores L2 1MB, it is not 2 * 512, so if one core is idle the busy core can use all.
Andy
Nope, it has a full 1 MB per
)
Nope, it has a full 1 MB per core afaik...
RE: Nope, it has a full 1
)
I can only go by the info on Tom's Hardware, Because it is not even listed on Intel's site. Or even in the Wikipedia.
I would even hazard a guess that it is a HP laptop.
Andy
edit] Just googled T2060 all indicate 1MB shared
It's a Samsung ;-) give me 10
)
It's a Samsung ;-) give me 10 minutes and I'll link a CPU-Z screenie.
EDIT: Okay, looks like you were right about the cache.
RE: If anyone is
)
Actually, the nice thing about PCLinuxOS 2007 is its multimedia capabilities. Other Linux distro's can be made multimedia capable after you install them, but PCLinuxOS is multimedia ready with a default installation.
Having said that, for regular production-type work, I prefer something like Ubuntu, Debian 4.0, or one of the Red Hat derivatives. That's because the PCLinuxOS repositories still lack the productivity software that I need for my work.
I think I'll stick to
)
I think I'll stick to Debian/Ubuntu... got used to the flavor, and like the software packages. It can be a bit more tricky to configure, though.