2mb L2 prescott and ee3.73ghz reviewed!

[saaya]

http://translate.google.com/translate?u=http%3A%2F%2Fwww.x86-secret.com%2Fpopups%2Farticleswindow.php%3Fid%3D11 9&langpair=fr%7Cen&hl=de&ie=UTF-8&oe=UTF-8&prev=%2Flanguage_tools

wow... seems like intel worsened the L2 cache latency even more for the 6xx 2mb L2 cache prescotts, so the extra mb of l2 cache is barely faster than the current prescotts :o

in one test the 1mb prescott even beats the 2mb prescott! :eek:

and the ee 3.73ee is now based on the prescott 2mb core and performs about the same as the last ee based on the northwood/gallatin core!

64 bit performence sux compared to a64 cpus in almost all tests,and intel cut the 14 multi so it we wont be able to force a 14 multi anymore...


this looks horrable! :(

well x86 secret is known to not like intel, so im curious to see to what conclusions other reviews come... but it doesnt look good.

[agenda2005]

I'm also dissapointed with the 3.73EE, but something in me told me that the 2MB L2 will not offer much for prescott. In principle, the larger the cache, the slower it becomes. It will be apparent when you put the CPU under memtest and you will see that the L2 bandwith will be slower than that of 1M prescott which will also be slower than the Northwood.

[OneyedK]

Hmmm saaya, I think you read the article a bit too negative...

1) since 6xx will cost about the same as 5xx now, we always gain a bit...
If you check out the stock performance, 6xx scores as good as or better than 5xx.

2) the article doesn't state Intel cut the x14 out, it states that no processor will be able to switch lower than x14, so same as today...

3) comparing cpu's that support 64-bit instructions (but still have a 32-bit-core-logic) with real 64-bit cpu's is a bit stupid. It just gives us an impression of what the cpu still can do with 64-bit-OS and 64-bit apps... And we already knew that it would be bad. Bit of an unfair test.
I have a i560 without EM64T support, so I could start complaining that it will never run 64-bit OS... But it's a bit silly, since I already knew that when I bought the cpu.

4) overclocking... the x86-secret-crew seems to reach 4.3GHz with aircooling (CNPS7700cu), not exactly what I call bad news... 286Mhz fsb...

But hey, what the heck, same price, slightly better (or the same) performance, runs cooler and better OC...
Sounds nice...
Let's wait for dual-a nd true-64-bit-cores...

[ar7786]

Does it work on AS8?

[Tyrou]

well x86 secret is known to not like intel

I'm french, and I've read almost every article from x86-secret, they're not anti-intel at all, in most cases, we even blame them to be a lil' bit intel-addicted ;)

[wwwww]

In principle, the larger the cache, the slower it becomes.

Why would it? My cpu has a 2MB L2 and a 9ns l2 latency. Thats better than any 512k a64.

[Xassius]

Unfair
AMD with DDR at the TIGHTEST TIMINGS
and Intel with DDR2 at CRAP TIMINGS - the i925XE is rated stock for CAS3 and even most generic DDR2 can do DDR2-533 at CAS3 easily.

[OneyedK]

Before the cpu starts addressing the main memory, it looks into the L2 cache to see if the data it needs is not already present.

So yes, it will take the cpu more time to read 2M address-space then it would take to read 1M (at the same bus-speed).
But, chances that the data is present in L2 are also higher.

Worst case is that the data is not present in L2, and at that very moment, 1M L2 will give a better performance than 2M L2. (there is simply no use of caching something that does not exist yet, no L2 cache would be the fastest in this case)

In reality, data will be present most of the times, and you will get a better performance because the cpu doesn't have to call on the main RAM wich is at least 5 times slower...

[OneyedK]

I agree with Xassius, the ram they used does 3-3-3-8-4 (at least) and the x86-secret-guys have it running at 4-4-4-12... Nice move... Fair review ;-)

[Kunaak]

That a suprising bad review... I wasn't expecting much from a 3.73 cause lets face it, the 13 multiplier pretty much means it's useless for overclockers.

2 meg prescotts.

from what I see there in that review I am reminded of 1 thing.

cas latency versus FSB.

is cas 2-2-2-5 better at 250, then 290 at cas 3-4-4-8?
generally for overclockers, it is.
so a larger cache, thats alot slower, doesn't sound all that enticing to me.
I am sure to some people a large cache is better, regaurdless of how much slower it is... but for me, the P4 is already slow, the prescott is the slowest P4 around, so a even slower prescott... well, thats just not good :(

I wish intel would just realize what they have in the Dothan, and build on that, and stop trying to fool the general public with high mhz slow CPU's and just sell a damn good one that people can actually afford, like the 1.8 ghz dothans and such...
I mean, they are only about $200 but have a high IPC and large cache and don't cost a arm and leg, and are insanely low wattage...
just make a affordable, dual channel Dothan board, and watch how fast people snap those boards up...

if only intel would listen...

[LarsK]

I was rather discouraged after reading the review as well. There's absolutely nothingworth upgrading to from my system. Guess I'll wait a bit longer with upgrading and see what's on the horizon later this year.

[saaya]

Hmmm saaya, I think you read the article a bit too negative...you think? lol

http://www.x86-secret.com/articles/cpu/p46xx/p46xx-8.htm



1) since 6xx will cost about the same as 5xx now, we always gain a bit...
If you check out the stock performance, 6xx scores as good as or better than 5xx.if all new cpus or cpu updates would be just as minimally faster than the previous generation then wed still be at the performence level of a pentium3 today.
afte the prescott this is yet another dissapointment from intel...



2) the article doesn't state Intel cut the x14 out, it states that no processor will be able to switch lower than x14, so same as today...
hmmm ok, my bad, the translation from french to english is hard to read :D



3) comparing cpu's that support 64-bit instructions (but still have a 32-bit-core-logic) with real 64-bit cpu's is a bit stupid. It just gives us an impression of what the cpu still can do with 64-bit-OS and 64-bit apps... And we already knew that it would be bad. Bit of an unfair test.
I have a i560 without EM64T support, so I could start complaining that it will never run 64-bit OS... But it's a bit silly, since I already knew that when I bought the cpu. what do you mean? oth the a64 and the prescott are just as much 32bit cpus as the other, none of them is a real 64bit cpu...
one just performs well while the other one doesnt.
this sounds like they compared the prescott with 64bit extensions with an itanium or ultra spark cpu wich are true 64bit cpus.



4) overclocking... the x86-secret-crew seems to reach 4.3GHz with aircooling (CNPS7700cu), not exactly what I call bad news... 286Mhz fsb...dont know what the last 1mb prescotts oced to...



Unfair
AMD with DDR at the TIGHTEST TIMINGS
and Intel with DDR2 at CRAP TIMINGS - the i925XE is rated stock for CAS3 and even most generic DDR2 can do DDR2-533 at CAS3 easily.
true! thats what i meant with they like intel, thats why im looking forward to less biased tests.

in some tests they even run it at 5-5-5!, not very "intel addicted" i would say... ;)



That a suprising bad review... I wasn't expecting much from a 3.73 cause lets face it, the 13 multiplier pretty much means it's useless for overclockers.

2 meg prescotts.

from what I see there in that review I am reminded of 1 thing.

cas latency versus FSB.

is cas 2-2-2-5 better at 250, then 290 at cas 3-4-4-8?
generally for overclockers, it is.
so a larger cache, thats alot slower, doesn't sound all that enticing to me.
I am sure to some people a large cache is better, regaurdless of how much slower it is... but for me, the P4 is already slow, the prescott is the slowest P4 around, so a even slower prescott... well, thats just not good :(

I wish intel would just realize what they have in the Dothan, and build on that, and stop trying to fool the general public with high mhz slow CPU's and just sell a damn good one that people can actually afford, like the 1.8 ghz dothans and such...
I mean, they are only about $200 but have a high IPC and large cache and don't cost a arm and leg, and are insanely low wattage...
just make a affordable, dual channel Dothan board, and watch how fast people snap those boards up...

if only intel would listen...
yeah! wheres the desktop dothan???!!! :(


I was rather discouraged after reading the review as well. There's absolutely nothingworth upgrading to from my system. Guess I'll wait a bit longer with upgrading and see what's on the horizon later this year.
you probably said the same when prescott was launched, now its 6 months later and its the same situation again... im really starting to wonder whats up with intel... maybe they plan to leave the performence crwon to amd and supply the masses with more affordable cpus?

at least it doesnt look like they are planning to beat amd...
wich is bad since amd wont get off its lazy 4ss and lower prices or increase clockspeeds or cache sizes until its under pressure! :(

[OneyedK]

thx for the repies saaya...

You are right about point (3), I didn't look deep enough in A64... Sorry...

They both (A64 and i6xx) have 32bit memory-controllers, so (I think) that the very bad latency settings punished the Intel twice in 64bit tests...

Looking forward to see a better review of these cpu's...

[saratoga]

Why would it? My cpu has a 2MB L2 and a 9ns l2 latency. Thats better than any 512k a64.

Do you mean 9clks? 9ns is worse then Prescott if your chip happens to run at 2GHz.

Anyway, assuming your 2MB chip is actually a Dothan (which actually has 10clks latency on an L2 load, 3 from the L1, 7 from the cache), its L2 latency increased over Banias to accomidate the extra cache. Ands its latency is only slightly better then an a64 (1.66 ns faster if I'm remembering the numbers right). Possibly winchester or one of its 90nm decendents will surpass it, though thats not a good comparison since their cache is arranged differently.

[LarsK]

you probably said the same when prescott was launched, now its 6 months later and its the same situation again... im really starting to wonder whats up with intel... maybe they plan to leave the performence crwon to amd and supply the masses with more affordable cpus?

at least it doesnt look like they are planning to beat amd...
wich is bad since amd wont get off its lazy 4ss and lower prices or increase clockspeeds or cache sizes until its under pressure! :(

Indeed I did. Now I do have a Prescott system @ home as well, and while it's certainly ok for my 12 year old son, it's no upgrade path for me. I want high performance in games as well as proff. apps.

I had SO hoped that Intel would put a kick-ass product on the market... But no such luck yet :(

[Disposibleteen]

i kinda lke the 1mb prescotts, i plan on buying one @3.4 or 3.6GHz and hope to clock it up past 4GHz