Intel's 1066MHz FSB - D925XECV2 and P4EE 3.46GHz Review

[Dave Baumann]

http://www.beyond3d.com/siteimages/b3dsmall.gif (http://www.beyond3d.com/reviews/intel/925xe/)Whilst one of the few truly universal constants appears to be the strength of Intel’s presence in the OEM market, they haven’t fared quite so well in the enthusiast space at the moment. Since the introduction of Prescott and the change to 90nm, they have been plagued by issues surrounding heat and power of the new processors, and the associated bad press that goes with it. These issues, and the apparent inability to scale Prescott to 4GHz, which was critical for further performance enhancements, has really given AMD a home run with enthusiasts and the AMD 64 platform.

Partially due to the issues surrounding Intel’s high end desktop performance, the presence of the newly introduced 915 and 925X desktop platforms, which bring numerous new technologies such as PCI Express and Azalia audio, hasn’t been quite as great as expected. Today, however, Intel are hoping to provide a further boost to the 925 platform with the Introduction of 925XE, and the 3.46GHz Pentium4 Extreme Edition, featuring a 1066MHz Front Side Bus. Click here to read more (http://www.beyond3d.com/reviews/intel/925xe/).

[3dilettante]

I'm a little confused about the CPU utilization graphs between the two CPU types.

I've seen utilization being compared between motherboards using different IO controllers and other integrated features used to offload from the CPU, but I don't know what to gather from comparisons between CPUs on utilization.

What mechanism would be involved in different rates of utilization between processors?
I can only think of a possible difference in how the decoder might use a different code path depending on the CPU, but this would muddy the waters a bit.

[Pete]

Perhaps PCMark04 should include a Spell Check, to complement its "Grammer" Check. ;^)

Good review, Dave. Thanks. I didn't realize how much faster the P4(EE) could be in many situations, but an A64 still seems somewhat more sensible for gamers. Particularly in a wicked Shuttle SN95G5. :D

[Dave Baumann]

You'll have to excuse the on/off nature of the launch of this - I think there were some confusions over the time conversions for when the NDA lifted! Anyway...

What mechanism would be involved in different rates of utilization between processors?
I can only think of a possible difference in how the decoder might use a different code path depending on the CPU, but this would muddy the waters a bit.

Assuming we are talking about the video decoding here?

If so, its might be the case that different code paths are being used, or it might actually be as simple as clock speeds (or a combination thereof). The utilisation is looking at the % of CPU time per second used on that particular task - video's are going to be capped at a maximum frame rate, so once the CPU exceeds the processing capabilities required to produce 24/30 frames (or whatever) per second then its not going to be tasked with anything else in that period.


Perhaps PCMark04 should include a Spell Check, to complement its "Grammer" Check. ;^)

What did I miss?

Good review, Dave. Thanks. I didn't realize how much faster the P4(EE) could be in many situations, but an A64 still seems somewhat more sensible for gamers. Particularly in a wicked Shuttle SN95G5. :D

Well, bear in mind that we are one release behind on the AMD front at the moment. However, yes, its not the case that AMD walks over in absolutely everything.

[3dilettante]

What mechanism would be involved in different rates of utilization between processors?
I can only think of a possible difference in how the decoder might use a different code path depending on the CPU, but this would muddy the waters a bit.

Assuming we are talking about the video decoding here?

If so, its might be the case that different code paths are being used, or it might actually be as simple as clock speeds (or a combination thereof). The utilisation is looking at the % of CPU time per second used on that particular task - video's are going to be capped at a maximum frame rate, so once the CPU exceeds the processing capabilities required to produce 24/30 frames (or whatever) per second then its not going to be tasked with anything else in that period.



Oh, that's what it was. I wasn't sure about what the utilization meant. I forgot about the frame cap on playback.

[Pete]

It's grammar, per my English teachers. (Maybe it's a Yank/Brit thing, but if grammar were spelled differently across the proverbial pond, I'd imagine it would be more like grammaur than grammer. ;))

Ironically, I think I misspelled "compliment."

Nothing like a little irony to colour a pedantic post.

[Anonymous]

Question for DaveBaumann?

Why nobody on net would tested new (925XE MB + P4-3.46GHz 1066FSB) using DDR2-667 memory, all I see DDR2-533, new intel board is capable DDR2-667, but for some reason nobody would test 667-memory.
I belive going from DDR2-533 to DDR2-667 would squeze another 5% boost; ----> before to give final judgement on new 1066FSB CPU.

[Anonymous]

Just on the CPU utilisation graphs for video decoding. How did you get the numbers for the intel systems? I've noticed that on intel HT systems a single threaded process seems to ba able to hit a maximum of 50% cpu utilisation as reported by task manager, with correspondingly lower numbers for lower utilisation. Maybe this is just an issue with the way winXP reports utilisation that has been fixed in an update.

However I was just thinking this could explain the lower utilisation on the intel systems when compared to the A64.

[Dave Baumann]

Why nobody on net would tested new (925XE MB + P4-3.46GHz 1066FSB) using DDR2-667 memory, all I see DDR2-533, new intel board is capable DDR2-667, but for some reason nobody would test 667-memory.

Well, I can't speak for everyone, but I don't have DDR2-667 memory to test on, so that would be one reason. Another issue is that we are dealing with reference Intel platforms here and the specification only goes up to 533 memory support righ now - I'd wager that when partner boards based on 925XE begin to be reviewed, that's when you'll see these types of test.

Just on the CPU utilisation graphs for video decoding. How did you get the numbers for the intel systems?

PERFMON.MSC - Select "Total" CPU utilisation and it will sum all the (virtual) processor utilisation; you can also opt to select the individual instances.

[mczak]

nice review.
There are IMHO though much more interesting processors around to review, even though they don't get that much publicity and don't even get a "launch". I mean, who the heck buys these "one-wafer-per-month" boutique parts like the FX and EE editions anyway? At least with the super expensive graphic cards, you get something like a 60 % performance improvement for buying a 500$ graphic card instead of a 300$ one. But with cpus, you get something like a 20% improvement for buying a cpu which costs 1000$ instead of a 200$ one.
So, the introduction of something like a A64 3500+ 90nm is imho way more interesting, people actually buy these!