Fall IDF 2005 - Day 2: Pat Gelsinger Reveals Intel's Server Plans
by Anand Lal Shimpi on August 24, 2005 1:39 PM EST- Posted in
- Trade Shows
Looking at Server Platforms
Pat dove into a bit of platform discussion, looking at both current and future server platforms.
The first item of discussion were the soon-to-be-released Dual-Core Xeon processors running on Lindenhurst (DP) and Truland (MP) platforms.
The long-awaited Dual-Core Xeon parts will be shipping by the end of this year, and will both feature 2MB of L2 cache per core. The Dual-Core Xeon 7000 (MP) will feature a faster 800MHz FSB.
Pat ran a benchmark run-off between three server platforms: an Irwindale Xeon DP platform, with HT Enabled.
A Paxville Dual-Core Xeon DP platform, with HT enabled.
And a Paxville Dual-Core Xeon MP platform, with 4 CPUs and HT enabled.
The performance test of choice was Pov-ray, which the Paxville MP system completed in 26.14 seconds, compared to 41.22 seconds of the Paxville DP and the Irwindale DP platform did not complete by the time the slide was gone.
Facebook
Twitter
RSS
11 Comments
View All Comments
Guspaz - Thursday, August 25, 2005 - link
A big deal was made over Microsoft Virtual Server's ability to move a virtual server to another physical box with no downtime. This is nothing new. Xen can alreay do this WITHOUT any special instruction extensions. As in, it can do it now, on any hardware.In addition, VMware's bluescreen demonstration is impressive until you realize that it too is nothing new; you can do the exact same thing on pretty much any virtualization solution, including older versions of VMware running on older hardware.
nullpointerus - Thursday, August 25, 2005 - link
Intel's virtualization technology is impressive because it is done at the hardware level, not because it was never done before. Read the last three paragraphs of this:http://en.wikipedia.org/wiki/Hypervisor">http://en.wikipedia.org/wiki/Hypervisor
So Vanderpool and Pacifica can efficiently run multiple, unmodified operating systems at the same time.
The blue screen thingy was a proof-of-concept showing that the Intel chip was working.
islandtechengineers - Sunday, August 28, 2005 - link
There’s a good point. It could be an attempt to help get the word out with multi OS server machines. I’ve witnessed a group load of server machines that aren’t cost efficient as a multi server setup. If I remember correctly (I haven’t checked out VMWARE lately), vmware required a host platform to operate properly and this appears to be their step up. I think it’s a great idea to run their setup without an initial host / parent.tanekaha - Thursday, August 25, 2005 - link
HmmChanging processors improved the good hole hit rate from 20 to 70%
Why ??
Were the old processors getting the sums wrong?
I can`t get my head round this.
You`d think all ya needed was more time on old processors
I`d appreciate enlightenment on this
Tanekaha
Calin - Thursday, August 25, 2005 - link
Yes, they were using the old Pentiums with error in FPU... (end sarcasm)I don't know for sure, but I think using processors that have more horsepower they can use more samples from previous holes. Some of these wells must be holed at a kilometer or more deep, and if one uses more sampling points in calculation, the results can be spectaculary better (or just as good, as this depends of lots of other factors).
However, doubling the points might increase the computing load ten times or maybe more. So, the need for better computers.
ceefka - Saturday, August 27, 2005 - link
Nice, if you still have to exploit Iraq (end sarcasm).jamawass - Wednesday, August 24, 2005 - link
Will they have on chip memory controllers?ceefka - Saturday, August 27, 2005 - link
All I saw on my quick read were massive on die L2/3 memory: 4 - 16MB and they are holding on to Hyperthreading.knitecrow - Wednesday, August 24, 2005 - link
All these Code names are making my heard hurtLeper Messiah - Wednesday, August 24, 2005 - link
so true...i'm confused as all hell when I read about sossaman, and tulsa and whichmacallit...