Intel quietly ships 64-bit Prescott

  • Thread starter Thread starter Yousuf Khan
  • Start date Start date
After 33 years, you aren't deaf yet? :)

Though, your wife has claimed you seem to be deaf for the
past 32 years.
 
It seems like Intel wants to stick to its theme that 64 bits is not yet needed
for the desktop pc user, but is needed in the server space. That is why the
P4s with 64 bit compatibility are being called server chips. It will be
interesting
to see how soon those chips are available for consumers to purchase, and
how they will be priced.
 
JK said:
It seems like Intel wants to stick to its theme that 64 bits is not yet needed
for the desktop pc user, but is needed in the server space. That is why the
P4s with 64 bit compatibility are being called server chips. It will be
interesting
to see how soon those chips are available for consumers to purchase, and
how they will be priced.
Click to expand...

The prime use for the 64 bit Xeon is likely to be for better performance
for 32 bit applications. With more cache and faster FSB the processor
should be 10-15% faster for many CPU-bound applications. If prices are
reasonable it might be a good choice for some uses, even single CPU
applications, since the HT can provide some of the benefits of SMP.
 
Bill said:
The prime use for the 64 bit Xeon is likely to be for better
performance for 32 bit applications. With more cache and faster FSB
the processor should be 10-15% faster for many CPU-bound
applications. If prices are reasonable it might be a good choice for
some uses, even single CPU applications, since the HT can provide
some of the benefits of SMP.
Click to expand...

10-15% seems a little optimistic for just an FSB and cache increase.

Yousuf Khan
 
JK said:
Do you have any links to benchmarks for the 64 bit Pentium 4s?

Bill Davidsen wrote:
Click to expand...

Just the articles in InfoWorld. I assume they are online, but I no
longer have the mag. If you can't find them I'll go look in the library.
They were around 15%.
 
10-15% seems a little optimistic for just an FSB and cache increase.
Click to expand...

I don't know, the P4 saw a fairly big jump when going from 533MT/s to
800MT/s bus speeds (a 10% improvement on applications was not at all
uncommon), I suspect that the Xeon will show a similar jump as well.
It should especially help even more for dual-CPU systems where the
Xeon is even more restrained by the memory subsystem.
 
Tony said:
I don't know, the P4 saw a fairly big jump when going from 533MT/s to
800MT/s bus speeds (a 10% improvement on applications was not at all
uncommon), I suspect that the Xeon will show a similar jump as well.
It should especially help even more for dual-CPU systems where the
Xeon is even more restrained by the memory subsystem.
Click to expand...

Most of those P4's also accompanied the FSB increase with increases in
overall frequency. A 10% increase is believable if you combine the FSB
increase with the overall frequency increase together. Perhaps 5% due to
FSB, and 5% due to overall frequency?

Yousuf Khan
 
Most of those P4's also accompanied the FSB increase with increases in
overall frequency. A 10% increase is believable if you combine the FSB
increase with the overall frequency increase together. Perhaps 5% due to
FSB, and 5% due to overall frequency?
Click to expand...

Not really, for example there was a 2.8GHz P4 running off a 533MT/s
bus as well as one running off an 800MT/s bus. I would guess that the
average speed-up for applications was around 5%, but there definitely
were some applications that saw a 10%+ improvement just from the
increased bus speed (same motherboard, memory, etc).

Presumably this difference would tend to grow as clock speeds went up
(the Xeons in question were running at 3.6GHz) and if you add in any
improvements in cache it could make more of a difference still. Of
course, in the case of the Xeons there are a LOT of factors involved,
not the least of which being that it's a whole new core (the
Nocona/Prescott vs. Prestonia/Northwood comparison), and a new
motherboard + chipset.

Still, when you get right down to it, the Xeon, particularly in dual
and quad CPU setups, was getting rather limited by it's bus as
compared to some of it's competitors, so this increase is bound to
help.
 
JK said:
I found this article on the 64 bit Xeon 3.6 ghz.

http://www.anandtech.com/linux/showdoc.aspx?i=2163&p=1

The Xeon 3.6 ghz looks like a poor performer compared to the
much less expensive Opteron 150 in those 64 bit tests.
Click to expand...

The point I (and InfoWorld) am trying to make is that these processors
can and probably should be thought of as better 32 bit units, and the 64
bit capability as a bonus. The anandtech review is a good example of
finding something the CPU does poorly and using benchmarks to test that.

A better test would be running the 32 bit benchmarks, and since these
are most commonly used in servers testing as 32 bit servers running
server applications, preferably both for DOS and Linux, would be more
useful to the typical buyer.

I am not predicting the results, just saying the tests were not
representative of typical use for Xeon.
 
The purpose of having a chip with a 64 bit mode is for a seamless
gradual transition to 64 bit software while still running 32 bit code.
If the 64 bit mode performs much worse than a competing product,
then it doesn't make sense to buy the chip, but rather to buy one that
also has great performance with 64 bit software. There have been
many articles comparing 32 bit performance. There is also this
article about 64 bit performance.

http://www.computerworld.com.sg/pcwsg.nsf/0/EAAFAE24D76C416548256EED001242F2?OpenDocument
 
Yousuf Khan said:
Most of those P4's also accompanied the FSB increase with increases in
overall frequency. A 10% increase is believable if you combine the FSB
increase with the overall frequency increase together. Perhaps 5% due to
FSB, and 5% due to overall frequency?

Yousuf Khan
Click to expand...

Actually I remember those reviews. The clock speed actually went
*down* for the top of the line CPU at the time: 3.06 down to 3.00.

A bigger culprint was (IMO) that Canterwood/Springdale were a great
chipset right out of the gate.
 
Tony said:
Not really, for example there was a 2.8GHz P4 running off a 533MT/s
bus as well as one running off an 800MT/s bus. I would guess that the
average speed-up for applications was around 5%, but there definitely
were some applications that saw a 10%+ improvement just from the
increased bus speed (same motherboard, memory, etc).

Presumably this difference would tend to grow as clock speeds went up
(the Xeons in question were running at 3.6GHz) and if you add in any
improvements in cache it could make more of a difference still. Of
course, in the case of the Xeons there are a LOT of factors involved,
not the least of which being that it's a whole new core (the
Nocona/Prescott vs. Prestonia/Northwood comparison), and a new
motherboard + chipset.

Still, when you get right down to it, the Xeon, particularly in dual
and quad CPU setups, was getting rather limited by it's bus as
compared to some of it's competitors, so this increase is bound to
help.
Click to expand...

Running with HT on increases cache misses, larger cache and faster main
memory will give an improvement there.
 
Back
Top