8

I want to know why the price of a Xeon CPU and a similar Core i7 CPU is so different. The CPU clock is more or less the same. I want to know this because I want to have an understanding of what kind of CPU is more suitable for me. I do heavy mathematical computations and simulations which may take up to two weeks to finish.

Example: Intel Xeon Processor E7-8893 v2 vs Intel Core i7-4960X Processor Extreme Edition

  • 1
    Reviewers: I've edited the question so that this is a more general hardware question, not a specific shopping recommendation. – bwDraco Feb 20 '15 at 18:04
7

The i7 series is focused at end-user computation in desktop environments where the Xeon processors are for non-consumer high performance computing, often used in servers, and are optimized for such. For example, see that the Xeon processor can handle around 1.5TB of RAM.

Clock speed is not the only factor which determines what makes a processor expensive. For example, the cache is nearly double the size, which reduces calls to RAM, which will significantly speed up the processor, as well as a significantly higher memory bandwidth. The underlying technology is much more advanced than a regular CPU.

For regular home computer use ranging from running games to running small-scale CAD projects, and i7 would be sufficient.

If you are doing large scale math computation and simulations and hitting certain deadlines are important, that it might be worth going with a Xeon, or getting a Xeon server and offloading your work to that.

Typically workstations only have Xeon processors and/or workstation GPUs if you are doing high quality simulations or renderings, necessitated by your industry.

If this is for home, don't get a Xeon, chances are you won't make the most efficient use for it, and it isn't worth the money. If this is for work, and the sky is the limit for price, then go with a Xeon.

13

Different Uses and Environments

  • Consumer-grade Core processors are designed for everyday desktop or gaming applications and are therefore optimized to operate at higher clock speeds. Most consumer applications cannot take advantage of more than a few processor cores and would benefit significantly more from a processor that operates at 4+ GHz than having 8 or more cores.

  • Xeon processors, on the other hand, are designed for business applications such as servers and workstations. These applications benefit significantly from having more cores. Because scaling a processor to extremely high clock speeds is difficult and inefficient, increasing the number of cores is better than running 4 cores at 4.5 GHz or faster. Especially with workloads such as servers and Big Data, 15 cores or more at 2–2.5 GHz can be far more effective than 4 cores operating at 4.5 GHz, while consuming less power per unit of performance.

Scalability and Reliability

  • A Xeon processor typically doesn't just have more cores. These processors also have much larger caches (37.5 MB in your case) and have ECC memory support. In addition, the processor you selected is designed for servers capable of scaling up to eight (!) sockets. These are not cheap features to develop, test, and enable.

  • Even if the processor doesn't have extra cores relative to a Core i7 EE part, the extra die space needed for the larger cache and advanced features, along with the extensive validation required to ensure that the processor will function reliably under extreme conditions, 24/7/365, in mission-critical business applications, significantly increase the manufacturing cost of the processor.

Total Cost of Ownership

  • Another major criterion in business applications is power and cooling. In consumer desktops, disproportionately higher heat output and power consumption are often acceptable compromises for an end-user who wants maximum performance on a gaming workload. However, business and datacenter computers run 24/7/365, often in large clusters, and therefore incur very high power and cooling costs. A processor that can do the job with less power and heat while delivering the same effective performance is going to end up costing less money in the long run. In these environments, total cost of ownership (TCO) is the real measure of the cost of any computing resource, not the sticker price.

  • In critical business environments, $8000+ each for processors which you can cram eight onto a motherboard and can use ECC memory for reliability is much better than an $1000 consumer 6-core part which is less efficient and cannot scale to more than one per board. When a business's ability to perform depends on its compute capabilities, these Xeon processors are easily worth the sticker price. That's how Intel is able to charge these prices.

  • Physical space is at a premium in datacenters, and lower density means less performance and less efficient cooling. Reliability is paramount, and crashes and errors due to memory glitches are not tolerable to any degree. It's much easier to understand why these processors are that expensive when you consider their intended use.

6

Adam isn't wrong, but I think he missed the biggest driver. The E7-8893 v2 is intended for octa-processor systems. Meaning they will have eight of these beasts in a single motherboard. (That's what the first "8" means in the processor number, based on the Xeon naming scheme.)

Since having 8 CPUs is such a rare thing, intended for such specialized applications (honestly I don't even know what), they are charging a ridiculous premium on them. Maybe there is extra hardware cost involved in making them, but it really could just be because Intel can get away with it.

Xeons in general are really designed for multi-CPU setups in rack-mounted servers, and that's why businesses will pay more for them, even those that will only handle dual-CPU. The consumer parts, like the i7 series, are intended for home use, and are always restricted to single-CPU applications.

Which one is right for you depends on your budget and your application. If you have a big company behind you, doing big company work on a big company budget, maybe you can get Xeon CPUs. And maybe -- just maybe -- you are in whatever narrow specialty needs octa-processors. But most likely you want the i7.

  • They could be the same price but if a consumer will pay more why shouldn't they sell it for more. It's intential product separation. Of course Xeon also has feature seperation. – Ramhound Feb 22 '15 at 23:06
  • There are many applications that are able to use even more than 8 CPUs. VMware Horizon, VMware Vsphere, VMware workstation, all of them support 16 CPUs with 16 Cores in each (256 cores in total), not so rare. – Conspiria Mar 11 '15 at 16:13
5

Most of these answers are more or less wrong. The main thing that distinguishes a Xeon processor from a desktop processor is scalability. To run a huge multiprocessor with 36, 72, or even 144 threads, you need the Xeon chip caches to talk to each other. L1 cache must talk to all the other L1 caches, L2 cache must talk to L2 caches, etc. In an i3, i5, i7 processor, this is all implemented inside the chip. On Xeon processors this feature brings the cache coherency signaling protocols to the pinouts of the chip.

By having the caches talk to each other, they can implement a cache-consistency (coherency) protocol so that data can be shared far more quickly and effectively. As a result, 30 cores can work together and make much faster progress on a problem. That feature is intentionally omitted from core i3, i5, i7 processors so they can charge the big bucks for the server chips.

Also, Intel intentionally cripples the i3, i5, i7 processors by limiting total addressable RAM to 32GB, generally (Xeons can often address 256GB or more). It would cost Intel < $0 to allow i7 processors to address 256GB of RAM, it's a feature like multiplier-lock which actually costs them extra money to cripple their chips in these ways. Dual, Quad, and even octa-processor(chip) motherboards always use Xeon chips because both of these features are critical for high scalability.

  • Welcome to Super User! That's a nice answer, keep it up. (I didn't vote because I don't know the area, but I definitely appreciate your effort.) – Ben N Jan 22 '16 at 17:01
4

To add to what RaveTheTadpole was saying, the systems which have 8+ sockets are used mostly in finance (think high-frequency trading) where they need to keep massive databases in RAM.

In this field, Oracle and RISC chips are still in vogue, because they can hold up to 96TB of RAM - while the Intel space can only hold 12TB (and that is with these newest chips).

Why does holding it all in RAM matter? Because if you have to go to non-volatile storage like RAID arrays or SSDs even - your competition has already made the trade at the best rate because their database was in RAM and you lost out on millions because you were 0.5 seconds late.

Other uses would be highly virtualized environments or possible video editing or 3D rendering though I imagine a GPU is better suited to that.

At this level, you might want to think of processors like those massive dump trucks you see that are 30 feet tall and can carry a cement truck in the back of them. Yes, they are bigger, but they are highly specialized, and you don't want one.

The power bill for an 8 socket system would be more than the average mortgage.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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