Intel 925X/915: Chipset Performance & DDR2
by Wesley Fink on June 19, 2004 3:01 AM EST- Posted in
- CPUs
Final Words
It was hard to resist being glib and titling this article, "Much Ado about Very Little". That feeling of disappointment comes from seeing so much new technology introduced all at once, and then finding out the real performance benefit is extremely small - if it exists at all. If Intel wants us to turn our computer world upside down, there should be a real tangible benefit to the bucks we are expected to spend. Unfortunately, that performance advantage is pretty hard to find - at least for now. There are certainly a few gems in the total package, but if you're looking for a big performance advantage it just isn't there.The feature side of the equation is a lot easier to handle, as Intel has lavished all the features a techie could dream about on the new chipsets. High-Definition audio, Matrix RAID, a new bus with a bright future, and an 8GB per second bidirectional graphics slot are a few of those features that come to mind. Those new features do come with a price, however. You are also asked basically to give up IDE hard drives and also start the transition to SATA optical devices. This will be an easier task for some than for others.
There is also the question of future advantage. It appears that all these new technologies would set you up well for the future, but frankly, we're not absolutely sure that is the case. Time and the marketplace will answer this question better than any crystal-balling, and seeing where other manufacturers and the marketplace take these new technologies will soon provide a clearer picture of what will win - and what won't. There is no doubt that Intel wields enormous clout in the marketplace, but that clout did not make Rambus a winner for Intel. If prices for the new Intel vision of a PC are high and performance gains remain small, directions can change quickly again.
There were a few real surprises in the new boards and chipsets. The 925X is the Enthusiast part, and it is clearly the fastest of the new chipsets. We were surprised, however, to find the advantage over the mainstream 915 chipset was only 1% to 3%. 925X also must use DDR2 memory, which can make that 1% to 3% performance advantage very costly at current DDR2 prices. 915 will come from some vendors with DDR2, but there will also be a huge selection of DDR 915 boards, at least until DDR2 prices drop to more comparable levels with DDR.
We also clearly saw in our performance tests that a current Intel 875 motherboard, running fast DDR memory and a Northwood processor, was just as fast as a 925X running a Prescott at the same speed. This is not really a failing on the part of the new chipsets, though, as it indicates real progress in Prescott development. When Prescott was first introduced, it was a much poorer performer than Northwood. The fact that it is now competitive with Northwood using all these new technologies indicates a lot of progress has been made.
Last, we looked at DDR and DDR2 on the 915 chipset. We saw that DDR is just as fast as DDR2 on the 915 boards, given the current state of DDR2. If you are looking for a new Intel motherboard, don't hesitate to choose a 915 with DDR support because there really isn't a performance penalty. As prices drop and timings get faster for DDR2, then DDR2 will be a better choice.
There is a lot to like in all the new features and technologies that Intel has introduced with the new chipsets. We will be covering these new features in more detail in future reviews. We just wish that there were more real performance gains to get excited about. This is nothing new for Intel, and they are a company that always seems to have plans down the road that suddenly makes it all fall into place. Be sure to read the rest of the story, which compares the new Intel processors and takes a closer look at PCIe graphics and Intel's new integrated graphics solution.
57 Comments
View All Comments
Wesley Fink - Saturday, June 19, 2004 - link
#22 & #23 - This will all make more sense when Derek's companion piece is posted later today, which compares the 3.6, 3.4EE, and AMD chipsets.There is no 3.6EE, and we considered comparing the 3.4EE, but the early benches showed the 3.6 to be a fairer comparison. The 3.6 is the only new Intel chip and the top (3.4EE is just a new 775 package launch) and the FX53 is the top AMD chip. The FX53 runs at the same speed as the 3800+ (2.4GHz) but has twice the cache, and both are Dual-Channel Socket 939. The FX55 will not be released until late this year.
Comparing last year's 3200+, the first Socket 754 Athlon 64 and single-channel, to the just-released 560 (3.6GHz) compares nothing.
Bozo Galora - Saturday, June 19, 2004 - link
If the chipset supports only one IDE channel, why does the Gigabyte board have two green IDE sockets.Am I missing something?
I really wanna know, because only one IDE means I dont buy this stuff.
shabby - Saturday, June 19, 2004 - link
So when are the dual core cpu's coming out? Think ill wait for those.SDA - Saturday, June 19, 2004 - link
That's just it, though, the FX-53 is just a few tens of dollars more than the 3800+, so some people figure why not? AMD is overcharging because they know they can get away with it..stephenbrooks - Saturday, June 19, 2004 - link
Maybe replace '3500+' by '3800+' above.stephenbrooks - Saturday, June 19, 2004 - link
One interesting thing is the difference it makes which AMD chip people compare the 3.6E to in these reviews. Before coming here I read the one at http://www.bit-tech.net/review/326/ and came away thoroughly impressed. The clincher was that they compared to the A64 3200+ whereas this review compared to the FX-53.I'm not totally sure if the 3.6E _should_ be compared to the FX-53, being as it is marketed as a '560', in the 5xx 'mid-range' desktop segment for Intel. A better comparison for the high-price-end might be 3.4/3.46EE vs. FX-53 or '720' vs. FX-55 later on.
I'd think 3.6E vs. A64 3700+, or even better the 3500+ (as it's on AMD's new socket too) would be an informative comparison. It's just a shame Intel's high-end offerings are a bit thin on the ground right now.
Staples - Saturday, June 19, 2004 - link
Good review. I wish you had compared the 875 with the 925x however both using a Presscott CPU. With one using a Northwood and one using a Presscott fails to show which chipset optimizes the power of the CPU. The fact that you didn't further inphasises that Intel is in one hellava drought. The 3.2 Northwood/875 combo that came out like 8+ months ago is still faster than ANYTHING they have released.I am waiting for a reason to upgrade but this sure isn't something I was looking for. With Intel having these troubles, AMD is going to take their sweet time putting out anything faster.
Pumpkinierre - Saturday, June 19, 2004 - link
It would be great to see the 915 with Skt478- best of both worlds. I know that Skt775 is appearing on 865 mobos (Abit AS8). At that last Taiwan computer expo they had dual intel Sckt mobos but were they 915s?rjm55 - Saturday, June 19, 2004 - link
#16 - I disagree with you. I think Northwood on 875 is exactly the right comparison to Prescott 775 on 925X. Both at the same speed as done in the review. Prescott was not designed for 875, so comparing the best last generation 875/Northwood to the new 925X/Prescott is extremely fair. I'm sure Intel would have preferred a comparison of Prescott to Prescott, because they would have looked better. I think that was #4's point, which you apparently missed.Doormat - Saturday, June 19, 2004 - link
Now I dont feel so bad about buying that P4-3.0C/i875 combo last month...