Server Fault is a question and answer site for system and network administrators. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Anyone know the technical reason a MP Server 7000 series Quad core Tigerton Xeon 2.4 Ghz is approx. six times the price of same speed 5400 Series Quad Core DP Server processor.

What is so special about the MP


share|improve this question
up vote 7 down vote accepted

The MP is artificially expensive as it supports being used in configurations with more than two sockets. In practice, you can view the DP as simply a crippled version of the MP fiddled so it won't work in multi-socket configurations.

The MP is almost identical to the DP and almost certainly no more physically difficult to manufacture.

On the Opteron, the Hypertransport links for configurations with more than two sockets are structured slightly differently but not in any way that makes the chip more expensive to fabricate.

The pricing is purely artificial, structured to segment the market between people buying cheaper two-socket machines so they can charge a higher margin to the more well-heeled customers buying 4 and 8 socket boxes. That's all. You can see the behaviour in that they have fairly well defined price points for their 'latest and greatest', 'last generation' and so-on. As they bring out a new 'latest and greatest' the previous generation drops down a price point. This is not the marketing strategy of a company where the actual per-unit costs of the products differ significantly.

share|improve this answer
What has to be managed or managed better in a four socket system verses two sockets. Memory access, bus contention. I don't think there is L4 shared off chip cache or anyhting. If you happen to know ? – kingchris Jun 25 '09 at 17:07
On a two-socket Opteron system there is a single (IIRC) 16 lane Hypertransport channel between the two sockets. On 4+ sockets the system uses three 8 lane Hypertransport channels (or something like that - I can't remember the exact details). Anyway, the configuration of the HT bus is slightly different. In no way does this significantly affect the fabrication costs of the chip. – ConcernedOfTunbridgeWells Jun 25 '09 at 18:33

While I'm sure there are feature differences, I highly doubt that there is a major "technical" answer. Production line tool-up, poor yields on a new chip, and overall fab capacity probably account for the lion's share of the costs to Intel. Having said that, I can't imagine it really costs Intel that much more to make the higher-end chip.

This doesn't look surprising to me. Intel's highest-end processors are always priced well above the middle-of-the-market processors. Back when Intel published their "iComp" benchmarks, I would often compute the "dollars to iComp" ratio to figure out roughly where the sweet-spot in price to performance was. Now that Intel doesn't publish even proprietary benchmarks (they've had a no-public-benchmark publishing policy for a long time) it's a little harder to compute price-to-performance ratios today.

Intel's marketing people know that the segment of the market that needs the higher-end gear will often pay nearly any price. The MP series processors have additional "glue" to allow more than dual-processor configurations, whereas the DP series can only be run in a dual-processor configuration. If you need more than dual-processor operation you "graduate", in Intel's mind, to the higher-end processors.

share|improve this answer
Horrible, horrible. If only I could get my mad scientist crystal growing kit to clone their top-end silicon substrate. World domination would be assured. Cheers. I cant up-vote yet – kingchris Jun 25 '09 at 16:43

The cost of CPU's has very little to do with what it actually does and everything to do with how many of them can be crammed into a single silicon die, and how many of them are likely to be sold. Even if they are almost identical physically if Intel sell 500k 5400's vs 50k 7000's then the cost of the 7000's will have to be higher to eat up the fixed costs of managing that product line. The overall market volume for dual socket cpu's is definitely more than 10x the market for MP capable cpu's so that alone will explain a lot of the cost.

Even so the dominant reason they are more expensive is that the market for them will support a higher price. MP CPU's go into systems that cost 10x and more the price of DP Systems and the CPU vendors make a very healthy margin on those.

The same tiering applies even within CPU ranges - and in those cases there is absolutely no difference whatsoever between the CPU's - a XEON 5520 and a XEON 5570 can come off the exact same silicon die - they are separated out during testing and may even have passed precisely the same tests before being finally sorted. But since there is a bigger market for a $250 CPU's vs $2000 CPU's and the vendors make sure they have the correct distribution of each type even if that means labeling (and possibly electrically\physically limiting) a specific high spec CPU as a lower tier version.

This is sound business sense - everyone benefits by having cheaper average performing systems while those who really want the absolute max have to pay way over the odds for the privilege.

share|improve this answer
What a wonderful business world we live in. (Sarcasm, sarcasm) – kingchris Jun 25 '09 at 20:31

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.