I’m trying to help someone in academia buy a unix workstation. The task involves intensive number crunching on very large (14 GB) data sets using packages like Maple, MatLab and SAS.
Here is a quote they got from their computing department:
There are two main UNIX servers:
- Sun Microsystems V40z with 2 Dual Core 64-bit AMD 852 CPU’s and 16GB RAM
running Red Hat Enterprise Linux AS (64-bit).
- Sun Microsystems V40z with 4 Dual Core 64-bit AMD 880 CPU’s and 32GB RAM
running Red Hat Enterprise Linux AS (64-bit).
Can you translate the above into English?
I appreciate any help.
[My translation after the jump.]
I didn’t know until today that Sun Microsystems makes Linux boxes powered by AMD and RHEL. Anyone want to comment on them?
Sun Microsystems is the vendor. This means that there will be a surcharge due to it being a large commercial vendor. Normally Sun tries to sell the Sparc line and their (very reliable, but not useful to them) Solaris operating system but this is not the case.
My first caveat is I thought the Sun Fire V40z is a server, not a workstation. This will be too loud to keep in the office. It needs a rack in a data-center somewhere or something like this.
AMD 852 and AMD 880 are the 800-series AMD Opteron chips. The difference is the first is not a “dual-core” chip and the second is. So the first configuration is almost exactly like [the computer I spec’d two years ago] only it goes 2.6 Ghz instead of whatever that was (I forgot but I think it was the Opteron 24x series “Sledgehammer” chips at around 2.2Ghz. The 852’s are part of the 85x series known as “Athens” Opteron. I explain the difference below, but it’s negligible.).
The second configuration on the AMD 880 is a different series known as “Egypt” series and are designed like the new Intel chips with “dual core” which means two processors per chip. One question I have is, given the mistake in the previous configuration,is this four chips two “dual core” chips? If the former, then it should really outshine the Mac Pro configuration mentioned earlier because there are 8 effective processors there with double the ram and dedicated memory bandwidth to each chip. If it is the latter, it should do worse because this processor is 2.4 Ghz instead of 3.0 Ghz, the memory bandwidth is 1Ghz instead of 1.33Ghz, and two processors aren’t enough to really cause the AMD’s memory advantages to shine (you need four or eight).
Also be aware that some software licenses charge based on number of CPUs they are executed on. Make sure that this isn’t the case or the costs are worth it.
Red Hat is just a vendor of a very good Linux. Basically Red Hat Enterprise Linux is like any other except they sell a support contract with it. Obviously, there is some surcharge for that.
A handy list of terms
- assembly language used by these chips. As opposed to PowerPC or Sparc
- chip vendor, competitor to Intel
- AMD’s name for server chips (means multi-processor capable)
- 24* or 84* series aka “Sledgehammer”
- Chips built using the 130nm process, The memory bandwidth is 800mhz. The max theoretical speed is the Opteron 850 @ 2.4Ghz. The max number of processors is 2.
- 85* series aka “Athens”
- Chips built using the 90nm process. The memory bandwidth is 1Ghz. The max speed is the Opteron 856 @ 3.0Ghz. The maximum number of processors is 8.
- 86*-890 series aka “Egypt”
- The same as the Athens only dual core. Each core has its own L2 cache (2GB instead of 1GB total) but shares the 1Ghz pipe to the main memory. The maximum number of processors is 8. The max clock speed is the Opteron 890 @ 2.8Ghz. These are like the Intel chips but each chip has its own memory controller which means that the memory won’t bottleneck at 4 or 8 processor configurations like they do on the Intel
- chip vendor. Inventor of the x86 assembler. Market leader.
- Core aka “Yonah”
- Replacement architecture to the Pentium 4. Actually just a rebranding of the Pentium M (mobile) arcitecture. It was necessary because the Pentium 4 had very high clock speeds but shitty performance and Intel wanted to avoid Mhz comparisons between the “Core” and the Pentium 4 architectures.
- Core Duo
- Noteboo and desktop chip with dual core design. This is the chip that was used in the first generation Mac Book and Mac Book Pros.
- Core 2 Duo aka “Conroe” (desktop) and “Merom” (notebook)
- The end of “Pentium”. Identical to Core Duo’s but uses a 65nm process with higher theoretical speeds at less power. This is the chip used in the lastest Mac Book and Mac Book Pros. It’s about 15% better CPU performance than the Core Duo.
- Intel name for Pentiums or Core Duos that are server chips (multi-processor capable).
- Dual-Core Xeon
- The server version of Core Duos. There are some early iterations that aren’t technically Intel Core microarchitecture.
- 5100 series aka “Woodcrest”
- The first Dual-Core Xeon based on Intal Core microarchitecture. Technically think of it as the multi-processor Core 2 Duo (65nm manufacturing process). This has a 1.33Ghz memory bandwidth and is the chip used in the Mac Pro. In my opinion, it’s going to outperform the “Egypt” in the 1 or 2 processor configuration but lose in 4-way or 8 way configurations.
- 5300 series aka “Cloverton”
- Recently released (November 2006) Quad-Core Xeon. Think of it as a Woodcrest with 4 cores instead of 2. I expect the next release of the Mac Pro to be using this chip. It should have been released by now
I appreciate any comments or corrections.
One thought on “Terminology soup in x86”
I misunderstood. The V40z’s are the shared servers in the department vs. making a purchase of a desktop for dedicated use.
In related news, apparently my Dell Precision workstation at work has one or two quad-core Xeons in them. Right now, I’m using it for testing Explorer and doing SSH. Haha! What a waste!
Maybe I should figure out something real to do with it.