Intel Pushes Back with Xeon 5100

conq writes "BusinessWeek has a piece on Intel's newest chip, the Xeon 5100, which many consider might be the chip that will llow them to stop losing ground to AMD. From the article: 'During the presentation, Intel ran the now-standard comparison test against AMD's highest performing chip, handily beating the system in a speed test. And in a jab at AMD execs, who handed kill-o-watt meters to analysts at the outfit's recent technology day, Intel execs used the same device to measure the new Xeon 5100 system's performance — gauged to be 7 watts better than that of the AMD-based system.'"

Comments

No.

[HoneyBunchesOfGoats]

Intel has by far the largest fabrication capacity of any chipmaker in the world. Both IBM's and AMD's fab capacities are much lower (AMD has used IBM's fab to help meet demand). IBM's inability to produce high numbers and high yields led to the Intel switch. Remember the delay in introducing the iMac G5? Apple had the design ready, IBM couldn't produce the chips. Result: months go by without any iMacs to sell. More than anything technical reason, IBM was bad for Apple's bottom line.

great for the market but...

[Fortun L'Escrot]

from the article (and missing from the summary):

That said, the results are skewed by the fact that Intel is producing chips using the 65-nanometer process, vs. AMD's 90-nanometer process. Typically, as more chips are packed onto smaller dies, performance improves dramatically. AMD is not scheduled to begin building chips on a 65-nanometer process until later this year.

and finally the really important bit:

"This is going to make AMD's life more difficult than it has been in the last two years," says analyst Nathan Brookwood, head of Insight64, a Saratoga (Calif.) consultancy. "If AMD can respond to this within a relevant time frame, the customers it has recently won may stick with it. If it can't respond with something that can provide the same level of performance and excitement, it could be very problematic."

besides intel's new server chip, what is really going on is summarized here:

"NOT JUST ABOUT THE CHIP." In negotiations with big-name customers, Intel also appears prepared to make price concessions to win more system designs than AMD. That sets the stage for what could turn into an all-out price war. "These chips tend to have high price tags and high margins," says Mercury's McCarron. "This is going to change the pricing dynamics. We simply haven't seen server chips sold below a few hundred dollars a unit, but now we're seeing them sold at less than $150. It's a different world than it was a few years ago."

while i am trying to help avoid any fanboyism that might follow, the above three paragraphs summarize the entire article quite nicely. enjoy.

Re:Still feature limited

[default luser]

Which is why these 64-bit Linux benchmarks show that Woodcrest scales as good as (and sometimes better than) Opterons at 4p. The vast majority of x86 servers are in the 4p range. Even Opterons have a worse-than-expected scaling issue past 4p, anyway, if you bother to look around to find the benchmarks.

The Optron's scaling issues beyond 4P is not "worse then expected," because it is entirely expected of the architecture.

The high-end Opteron has 3 HT links. This means it can work with up to 8 sockets "gluelessly," but it really performans much better with 4-socket systems. The architecture for a 4-way Opteron server uses the extra HT link to reduice the number of hops, so only one case has two hops.

But you can imagine that the 8-way configurations have a much higher average number of hops between processors, PLUS much more data flowing over the same HT links. No, the K8 Opteron is not really designed well for 8-socket systems.

But K8L IS designed for 8-socket systems.

Take a look at a page on this in the K8L preview article on Real World Technologies. Adding a 4th HT link will really make a difference.

4-socket K8L systems benefit because they take advantage of the 4 HT links to provide 1-hop latency to all sockets in the mesh, and can now have external I/O hooked up to ALL processors.

8-socket K8L systems take advantage of two things: the extra HT link is beneficial, and the advanced mesh created by splitting up the HT bus widths means MUCH better performance for 8-way systems.

Woodcrest is impressive as hell, but I will tell you one thing: there's no way in hell it's going to scale well beyond 4-socket systems. This is for the same reasons that have been holding back performance on 4-way Xeon syetems (reduced bus speeds with 4 processors on the bus, too much traffic). The Dual-Independent Bus allows Intel to scale well to 4-way, but no higher. K8L will allow for glueless scaling to 8-way, and will still provide a a cheaper solution than Intel's Dual-Independent Bus for 4-way chipsets and motherboard designs.

Woodcrest: good processor but not sufficient ?

[this great guy]

Just to recap things, the Xeon 5100-series, aka "Woodcrest", is the very first released processor family that is based on the new 8th generation, Intel Core Microarchitecture, technically inspired from the 6th generation (PPro, PII, PIII), instead of the 7th generation (P4). As a side note, Intel has been using the "Core Solo" and "Core Duo" denominations for some processors but this is just a marketing usage of the term "Core", because such processors are NOT based on the Intel Core Microarchitecture. Anyway, Woodcrest is the first to represent this all-new Intel Core Microarchitecture that is supposed to save Intel from the very competitive K8 design (Opteron, Athlon64...).

So, Woodcrest seems indeed to be a very good processor, as shown in this preview (the less-biased, more technically accurate I have been able to find up to this day). Intel claims that Woodcrest is "80% more performant at 35% less power" compared to the original dual-core Xeon processor, and most benchmarks seem to confirm this claim. It may seem technically impressive, but in fact considering the very poor design of the original dual-core Xeon processor, such an improvement HAD to be expected and was almost a prerequisite for Intel to even start thinking about taking back Opteron's market share.

Here is a quick fact list I have assembled from my own research and from the review linked above:

At equal clock frequencies, Woodcrest is about 5-15% more powerful than Opteron on traditional workloads (common x86 and arithmetic instructions), and much more powerful (30% and more) than Opteron on multimedia workloads (mostly SSE, SSE2, maybe FPU I am not sure).

At equal clock frequencies, Opteron is still much more powerful (30% and more) than Woodcrest on memory-intensive workloads due to its integrated memory controller (leading to better latency) and ccHT links in SMP cases (where memory throughput increases with the number of ccHT links).

At equal clock frequencies, Woodcrest consumes less power than Opteron, but Woodcrest's memory (FB-DIMM) requires more power than Opteron's memory (DDR400). So overall, a Woodcrest-based system consumes about as much power as an Opteron-based system (as shown in page 3 of the review).

At equal clock frequencies, Woodcrest is cheaper than Opteron, but Woodcrest motherboards (socket 771) are more expensive than Opteron motherboards (socket 939 and 940) and FB-DIMM memory is twice the price of DDR400. These pricing differences are so large that Opteron is still preferable to Woodcrest in most cases: Opteron is cheaper for any single or dual-cpu server config with 4 GB or more of memory, Opteron is cheaper for any entry-level server config (about $1500 and below) whatever the amount of memory is, Woodcrest seems to only make sense when the high-end processors (Xeon 5140, 5150 and 5160) are used with NO MORE than 4 GB of memory (else Opteron's cheaper memory has a price advantage).

Of course, in the high-end server market (4, 8 or more processors), Opteron is still the clear technical leader because Intel STILL hasn't switched to a CPU interconnect similar to HT and STILL isn't using an integrated memory controller.

In conclusion, I would say that when comparing only the processors, Woodcrest is superior to Opteron in many aspects (such as instruction throughput), and Opteron beats Woodcrest in other aspects (such as memory accesses). But when comparing a whole Woodcrest-based system versus an Opteron-based system, other factors come into play (such as price and scalibility), which make Opteron superior to Woodcrest in a lot of cases.