To its credit, Intel never did hype the core, and never played up the launch, it was so downplayed that the press release touted the new chip as being made on the new 'High Volume 90-Nanometer' process. This was hardly a shot across the bow prepping the hungry world for the next world-beater.
Most of the enthusiast site world shrugged, the results showed the new core to be a little slower in most tests. Games didn't fare well, some streaming apps did quite well. The rare test was vastly better on Prescott than Northwood, and a few really did not like the new core. The overall reaction among a lot of the sites was: 'we waited for this?'
More importantly, the 'community' immediately launched into an all out attack on the chip, the AMD fanboys had a field day, and the Intel enthusiasts only made the odd muted comment. There were more suXX0rz in the message boards than usual, and the things the chip brings to the table were mostly ignored in the stunning light of Prescott's apparent averageness.
After about a week looking over the numbers, I have come to the conclusion that the chip far from sucks, in fact it has a lot of potential. Before we get into the argument, if you have not already read it, check out the review at Aces Hardware (Note: The site is moving servers Thursday, so it will be down for most of the day), it will give you a good overview of the chip.
Once you have read that, there are three things you should keep in mind about the chip. First is that is on average about equal to its predecessor, Northwood, in most areas. For general use, there will be just about no difference noticeable to the user in terms of processing power. If you are the one in a 100 people who actually need more CPU power, test your main application and see which core is better, one might perform much better than the other. The other 99% of you can buy either one without worrying about speed.
The more troubling issue is power use. Much as Intel would like you to believe otherwise, this thing sucks power like there is no tomorrow, it is really really bad. If you don't believe me, read this. The test took a system with a Northwood 3.2 CPU and measured the power used. It then swapped it out for a Prescott 3.2 and measured again. The chip took an astounding 37.5w more power at idle, and 48w more under heavy use. Take those blast furnace jokes seriously, but take these results with a grain of salt. While it doesn't explicitly say in the test, if the power was measured at the wall, power supply inefficiencies could account for a low-double digit percentage of the power used. While you still use that much more power when it comes to paying for it, the chip does not consume that much itself. Still, that is a lot of power used for not much, if any gain.
The last thing is that this chip in socket 478 guise is a dead end. If you are one of the lucky few whose motherboard supports the chip, there are currently no plans to put out a variant faster than 3.6GHz that will fit in S478. While this may change, you would still be rather foolish to view this chip as a sane upgrade path from current Northwood boards.
OK, so far, I have been no more than neutral, and mostly negative about the chip, but I did say said that it does not suck. Well it doesn't. In reverse order, the problems all have solutions, the socket part being the easiest. While Prescott does not offer a fabulous upgrade path, when the Socket 775 variant will be introduced in Q2, it will be the socket of choice for the next long time. If you buy a slow S775 Prescott, you will be able to upgrade to a much higher-clocked chip. The S775 chips are the first real Prescotts in my opinion, the S478 variants should never have been put out. S775 chips are a safe bet though, and they don't perform badly either, look here for your next upgrade.
The next issue is power. I have been harping on about the 90nm process for a long time, it has been going through a rough patch, and for a while, it looked to be blowing every power budget set for it. Between the bad times, about six months ago, and now, the process wizards at Intel have brought things mostly back into line. Power use, while not stellar, is reported to be much closer to where they wanted it than to where it was. Add in that Intel has strongly hinted at versions that will work with the older motherboards (See P3 of the Ace's review here). This means the 90nm process, while still not great, is improving fast, and should continue to improve. It may not reflect it in the specs, but I would be very surprised to see real world power usage not coming down. Again, waiting would probably be in your best interest -- look for good things to start happening around the S775 launch.
If you are still not convinced, and point to AMD's Cool and Quiet, rest assured that there is an Intel answer buried beneath the layers of Prescott. While I am not sure if it still has to be debugged, or not supported until S775, it is there. Remember, it took AMD six months to admit C&Q was there, and a few more months to turn it on.
That brings us back to performance. The current chips are not designed to be great performers at the same clock speed, they are designed for two things, not to fade as clock speeds ramp, and to be able to ramp.
The fading part is barely noticeable, but if you look hard enough, you can start to see it. Several benchmarks have pointed out that while Prescott doesn't outperform Northwood, at the slowest speeds, the gap is large, but at the higher speeds, the gap closes a bit. As Intel ramps the speed of Prescott this year, expect the chip to catch up to Northwood, and surpass it in almost every way. It was built to run at high speeds, and time will show what it is capable of.
The next part is the ramp itself. Intel is currently only set to ramp Prescott to 4.0GHz this year. But, as AMD ramps Athlon64 production, it would not surprise me to see Intel mysteriously pull a rabbit out of its hat, or a speed grade. So, with performance increasing with clock speed, and clock speeds increasing, the chip should look better and better as the year goes on. Availability may be poor on the high end right now, but with the process improvements that will change. Also, when a process matures, Intel will process some wafers differently, sacrificing a little yield for a little speed. These new processes don't exist yet for the 90nm chips, and no 'hot' wafers means a little less availability on the high end. Luckily, Northwood is on a mature process, and those boutique wafers do exist, and a quick pricewatch check shows that there are enough Northwood 3.4s to tide one over. For the third time, all signs point to S775.
All three of the big issues that confront Prescott should be solved in a few months. And, coupled with a new platform, this chip should blossom. Before you scoff, look at the P4, initially in Willamette guise, then in Northwood clothes. Let's be frank, the Willy 1.3 was a really sad chip. The Willy 1.4 compared to a P3/1.4 (Tualatin), there was not one thing this poor core did right. Stacked up against an Athlon 1400, it was woefully non-competitive.
Then something happened. By about 1.8GHz, Willy was much closer to competitive, and the ramp was just picking up. AMD tripped and fell at the same time, and if face planting with the early Palominos wasn't enough, they did it again with Tbred-A. Meanwhile, Intel pumped out 1.9 and 2.0GHz chips, and followed it up with the tweaked Northwood core. They then took off with the performance lead, and kept running. By the time the 3.2GHz chips were out, there were very few people would point to anything AMD did better. The wallflower had turned out pretty good after all.
AMD came back with the Athlon64, and retook the performance lead, but they have had chips out for almost a year, and Prescott is only days old. Next time you are thinking of posting something in a forum with odd capitalization and zeros instead of the letter O, think about this: the race is just beginning.
Intel promised next to nothing with this chip, and they delivered (sorry, couldn't resist). They also say that this chip will do much better with time. Last time they said that, they delivered, and I think they will this time, and Socket775 will be the launching point. µ