Niagara II vs Niagara I OR UltraSparc T2 vs UltraSparc T1
So Im sure everyone, well maybe not "everyone", but a lot of people have heard the announcement of the new UltraSparc T2 (Niagara II) processor.
I currently am working with this Processor on a secret new platform. I've also worked with the UltraSparc T1 (Niagara I) on this system (T6300), codenamed St. Paul. (Am I allowed to post about system codenames!?). Anyways, the St. Paul, err the T6300 hosts the Niagara I chip. So whats the point of this post.
Well the Niagara II chip boasts "Unprecedented Throughput" and people have asked me how does the Niagara II chip compare to the Niagara I chip. Well , the answer is a LOT better, but I wanted some feedback of applications readers might use to test such "throughput", mention, point, suggest some tests for me to run and I will try to get some numbers in my spare time to really show the performance increase.
So its up to you readers!
As a sidenote, so I was talking about system codenames and there always seemed to me this secrecy about the codenames...well only for systems not released yet, but if you dig, you can find..take for example this code snippet at opensolaris
139 # add stpaul links to individual components
140 #
141 s none usr/platform/SUNW,Sun-Blade-T6300/lib/.......
Posted at 09:45PM Aug 08, 2007 by Eric Shobe in Code | Comments[4]
Today's Page Hits: 18
Right now most customers are on V20z's or the like, but my guess would be that you want to show off the multi-FPU and memory bandwidth. How about a shell script that use ImageMagick to resize and convert 8 TIFF images at a time, into JPEGs? That should show a significant improvement, and is something that is often done in web environments.
Posted by Patrick Giagnocavo on August 08, 2007 at 10:22 PM PDT #
Thats a good idea Patrick, Ill get some high-res TIFFs do some converting, maybe convert some mp3s to wavs, then compare them with the N2. I wouldnt mind finding a good pure multi-FPU test....
Posted by Eric Shobe on August 08, 2007 at 10:27 PM PDT #
Good point.
With ImageMagick, we had not so good results on T2000
with 16 parallel Threads.
X4200 Opteron results are much better for ImageMagick image conversion than the T2000 results. Interesting to see this on T2.
Posted by Thorleif Wiik on August 09, 2007 at 03:13 AM PDT #
Seems like Sun is shifting positions a little with respect to target applications for Niagara. I hear Sun is looking to promote Niagara-based servers for database workloads ... If so, seems the old standard TPC-related benchmarks should be tried.
I know its off-topic but I am still trying to discern whether Solaris can schedule two different processes (in my case forked from the same parent process) onto two different HW threads running on the same core at the same time?
Any thoughts,
Posted by Peter Thawley on May 07, 2008 at 10:22 AM PDT #