Intel 925X/915: Chipset Performance & DDR2
by Wesley Fink on June 19, 2004 3:01 AM EST- Posted in
- CPUs
Intel 925X Express
The Intel D925XCV is the Enthusiast version flagship for all the new Intel technology.
That is not to say Enthusiasts will stand in line to buy the new Intel board, since Intel has not traditionally supported the kinds of features and adjustments that will sway enthusiasts to buy a high-end Intel motherboard. The top Intel motherboard is always, however, a top-performing motherboard at default speeds, and it becomes the chipset performance standard. It is also the standard-bearer for Intel chipset features.
Click to enlarge.
If you look carefully, you will see that the power connectors on the new motherboards, both 925X and 915, are 24-pin, and not the common 20-pin ATX connectors on current Pentium 4 power supplies. We found that server grade power supplies providing 24-pin connectors work fine on the new boards. However, if you have a new top-line PCIe graphics card you will likely find a new 6-pin power connector on the graphics card. You will need either a new power supply with the 6-pin graphics connector or a converter cable that allows two 4-pin molex to be combined in a 6-pin connector to power the new high-end PCIe cards. This is true of both 925X and 915 boards.
925X only supports DDR2 and PCI Express 16X graphics. 925X is also the only 775 chipset version that is stated to support ECC memory. We were not able to verify ECC support, since we did not have access to DDR2 ECC memory. The other features are related to the ICH6 version used with 925X, though you will more likely see ICH6R and ICH6RW (pictured here) paired with 925X.
The other distinguishing feature of Intel 925X is accelerated performance modes compared to 915 chipsets. Intel first differentiated their "Enthusiast" board with enhanced performance on the 875 chipset. The 875 chipset featured PAT, or Performance Acceleration Technology, which was supposed to be a unique feature of the 875 chipset. Chipset manufacturers began revising 865 boards to support PAT-like performance almost as soon as Intel launched 865/875. This was not a development that pleased Intel.
This time around, Intel is also differentiating their top chipset with enhanced performance. Intel did not share any catchy names like PAT, and the diagram above is the only information that they provided on how they make the 925X faster. When we asked for more information than the almost meaningless diagram above, Intel said their acceleration methods were "proprietary". This is not really a surprise after the PAT confusion with 875/865. Industry sources tell us that the new acceleration scheme used on 925X is much more complicated than PAT; it will not likely appear on 915 chipset solutions from other manufacturers.
Facebook
Twitter
RSS
57 Comments
View All Comments
Mike89 - Saturday, June 19, 2004 - link
This is a real shake up to me. I mean this is no small change that Intel wants everyone to do. And for what? The only thing that comes to mind is "just so I can say I did it". Makes no sense at all for me. Intel shouldn't have put this out until they could show it was really worth it. I don't see it in any kind of way. And with all this, they STILL can't beat AMD (in gaming anyway). This looks more like a power grab than anything else, another way to make more money for Intel. I'm not one bit impressed or interested.Anemone - Saturday, June 19, 2004 - link
Thank you Anand for giving us the real deal on these new things. I'd love to say different, but AMD is too pricey and Intel performance is pathetic, and their prices, most especially when you consider the insane cost of DDR2, are no better."in 2005 we'll be better" Yeah right. AMD performs great till you give it too much to do at once, and they won't fix that till they bring in dual core. And if you think they'll manage that in 2005... yeah right on that too. AMD will be lucky to even SEE production chips on 90nm in 2005.
Don't think Intel has some great recovery coming either. The 925/915 release and the costs that they want to charge for chips that don't perform, is ludicrous. Yeah they multitask, but they can't do anything much better than systems could a year ago. Dual core might be coming, but it seems clear they have to change a lot of things to do it so why even create the 9XX chipset? This thing is going the way of the 820, from day 1.
If I'm not paying $800 for a chip that DOES perform, what on earth makes Intel thing I'm goign to spend $700 for the new 3.6 or worse $850-1000 for a chip that can barely keep up with that $800 chip? I know what makes you think that, arrogance. It's not desperation yet, but that's what it's going to become soon.
I'm disappointed in all sides of this. But I'm grateful Anand put it in a clear and honest light.
mostlyprudent - Saturday, June 19, 2004 - link
I agree with the comment toward the end of the article regarding Intel's ability to make things fall into place. Remember how disappointing the first itterations of the P4 were...Remember how dismal Prescott's performance was when first released? 2 years ago I bought a cutting edge system only to watch intel up the FSB from 400 to 533 to 800mhz, introduce 875/ICH5R and see SATA mature into the better hard drive option (see Anantech's Q2 hard drive round up), all within about a year. So I am glad to see Intel introducing a range of features all at once instead of changing the socket, then adding a new chipset a little later then a few months later revising the southbridge and so on. For those of us who do not have limitless upgrade budgets it is painful to see new features and chipsets added on what seems like a monthly basis.Of course I'm still planning on making the switch to A64 in the very near future.
SLIM - Saturday, June 19, 2004 - link
The funniest part of all of this benchmarking is that you STILL cannot buy a 3.4g prescott on pricewatch for any chipset (4.5 months after launch). People bickering about using the FX53 (launched earlier this month and widely available NOW) for comparison purposes should do a quick reality check and know that the Anandtech guys tried to show Intel in the best light they could... Raid for 925X vs. mid range seagate for AMD, wrong timing on AMD memory (wasn't it AT who said 2-2-2-10 was best performance), etc.Good review overall as usual for AT; looking forward to the upcoming, more indepth comparison.
AtaStrumf - Saturday, June 19, 2004 - link
The new chipset appears to be faster, but the CPU and DDR II appear to be a bit slower, so in the end it preety much evens out, with one BIG difference: PRICE.It is now very obvious why AMD is in no hurry to go PCIe or DDR II, but once this picks up AMD will have to play catch up again. Not great, but probably better than having to be the one to break the ice.
Today also seems to be the birth of the new naming scheme, and the 5 that prescott has gotten tells me we will be seing something more potent with the 7 in the name. That is in fact the point of these new names, since prescotts will get higher frequences, but not much more performance.
I'm much more excited about those CPUs than prescotts, which were DOA anyway. Franky I'm surprised it's doing as good as it is.
So if you dont't have performance to sell, what do you do? You sell features, which are the only good thing that this new chipsets brings for early adopters, with the notable exception of a single IDE, but as Gigabyte has shown, that is really not the problem. They just added an extra chip for that, just like it was with SATA in the early days, only now the roles are reversed.
Bozo Galora - Saturday, June 19, 2004 - link
thanx for quick reply and clearing that up wesley. you da man - I remember you from abxzone.Always thourough and focused - just like a Benihana chef.
Pandaren - Saturday, June 19, 2004 - link
29 - TrueWisdomPrescott might have been designed with 4-5 GHz in mind, but it's already dissipating 115 watts at 3.6 GHz. The Pentium 4's of yesteryear at least had potential for good scaling. Prescott has fallen flat on its ass as far as scaling is concerned.
When I think about it, this review isn't so much of a reflection of the awfulness of the 900 series chipsests as a reflection of the mediocrity of the Prescott Pentium 4.
tart666 - Saturday, June 19, 2004 - link
woo-hooo, NCQ for the masses! that alone is worth the upgrade...TrueWisdom - Saturday, June 19, 2004 - link
Well, while it is certainly disappointing to see that Intel's "hardware revolution" doesn't really do much at all, but I can't help but think about AMD two years back, when the AMD XP line was getting trounced by Intel's Northwood line. This sort of thing happens to companies, and I think that it's unfair to immediately say: "Intel is losing it."I think the biggest thing to remember about Intel is that they've always been convinced that clock speed is the way to go. That's why they're pushing this new standard so heavily: not because they think it will perform better at current speeds, but because they can push all of the speeds so much farther. (This was essentially the same thing that they did with the launch of the P4, and I'd say that ended up being all right.)
PCI Express allows for SO much more bandwidth that it would be silly not to adopt it, DDR2 can be pushed a hell of a lot farther than DDR1, and we already know that the Prescott was designed with 4 and 5 GHz models in mind. Is Intel's approach the right one? Maybe, and maybe not, and they certainly don't have the right to try and flip the industry on its head with so little immediate performance gains. But, they did it before with the P4, and after two years it became the clear dominator of the market, until AMD finally got the A64 out the door. So, while I totally agree that there's no need to upgrade now, that's no reason not to be excited for the future: look how far the P4 came, and look where it started.
Wesley Fink - Saturday, June 19, 2004 - link
#26 - The Gigabyte board adds 2 additional IDE channels with a GigaRAID controller, plus it also has the single IDE channel provided by the chipset. The new chipsets only support one IDE channel, but manufacturers can add in other chips if they choose.