IBM Think Center S50: The Ideal Corporate Small Form Factor
by Wesley Fink on October 21, 2003 10:55 PM EST- Posted in
- Systems
IBM Think Center S50: The Test
One of the decisions made when designing the IBM Think Center S50 was to remove any possibility of upgrading the on-board Intel Extreme Graphics 2. Without any kind of AGP port, there is no effective way of upgrading video except by using a PCI graphics card, which are rare and getting rarer.The lack of a video upgrade option is not a real issue where the S50 will be used, but it makes it impossible for us to mount our standard ATI 9800 PRO graphics card for a full suite of Benchmark tests.
IBM supplied our Evaluation Unit with a 3.2GHz P4 and a single 512Mb Infineon DDR333 CAS 2.5 DIMM. To be consistent with our other benchmarks, our limited testing used our standard 2 x 512Mb PC3500 Mushkin Level 2 memory. CPU-Z confirmed the Mushkin was running at DDR320 at 2-2-2-5 timings. Just to double-check, we also ran the Content Creation benchmarks with the stock Infineon memory. Performance in the benchmarks was a couple of points lower in either test with the stock Infineon memory. Reported results are with 2 x 512mb Mushkin at DDR320 2-2-2-5.
| Performance Test Configuration | |
| Processor(s): | Intel 3.2 800FSB Pentium 4 |
| RAM: | 2 x 512MB Mushkin PC3500 Level II DS 2 x 256MB Corsair 3200LL SS |
| Hard Drive(s): | Maxtor 120GB 7200 RPM (8MB Buffer) Western Digital 120GB 7200 RPM (8MB Buffer) |
| Bus Master Drivers: | Intel INF Update v5.00.1012 Intel IAA for 875P RAID not installed for consistency of Test Results |
| Video Card(s): | On-board Intel Extreme Graphics ATI Radeon 9800 PRO 128MB (AGP 8X) |
| Video Drivers: | Intel 82865G Graphics Controller 6.13.10.3510 ATI Catalyst 3.7 |
| Operating System(s): | Windows XP Professional SP1 |
| Motherboards: | IBM Think Center S50 Soltek Qubic EQ3401M Biostar iDEQ 200T @ 200.8 MHz Shuttle XPC SB65G2 (865PE) @ 200.5 MHz Asus P4C800-E (875P) @ 200.5 MHz ABIT IS7-G (865PE) ABIT IC7-G (875P) Gigabyte 8KNXP (875P) |
Recent performance tests on Intel 875/865 boards used 2 x 512MB Mushkin PC3500 Level II Double-bank memory. Previous tests of Intel motherboards used 2 x 256MB Corsair 3200LL Version 1.1.
All performance tests with the ATI 9800 PRO 128MB video card were run with the AGP Aperture set to 128MB with Fast Write enabled. Resolution in all benchmarks is 1024x768x32.
Additions to Performance Tests
We have standardized on ZD Labs Internet Content Creation Winstone 2003 and ZD Labs Business Winstone 2002 for system benchmarking. We are no longer reporting SysMark2002 results as part of our standard benchmark suite.As much as we appreciate the superb engineering that went into the design of the IBM Think Center S50, the actual performance of the machine was very disappointing. Content Creation benchmarks usually remain similar whether we are testing with on-board graphics or a $400 video card. However, the S50, while running a faster 3.2GHz while other machines ran 3.0, still posted the lowest performance in Content Creation of any Pentium 4 that we have recently tested. It appears that IBM's decision to use DDR320 (DDR333) instead of DDR400 coupled with other design decisions have combined to lower performance significantly. Stability and compatibility matter most on the Corporate desktop, but we doubt most IT departments will be happy to see the IBM Think Center with an expensive 3.2GHz P4 actually performing 10% to 20% slower than other 865 machines that we have tested with a 3.0GHz P4.
We suspect that this could be a significant issue to corporate purchasing and IT groups, since our performance benchmarking only tested Content Creation and General Usage — suites with Word, Excel, Power Point, and Multimedia Content, those applications that are the bread-and-butter of the Corporate PC.
Facebook
Twitter
RSS
20 Comments
View All Comments
mindless1 - Sunday, February 8, 2004 - link
It's a shame the article uses proprietary SWF images, instead of the industry standard formats which everyone can use. Is the author getting kickbacks from Macromedia?Budman - Wednesday, November 5, 2003 - link
asdadadUtterman - Friday, October 24, 2003 - link
I have deployed around 1500 IBM thinkcentre S50's nationwide and they are really great systems to work with. Out of the 1500 systems that I worked with, I only had problems with 5 of them. This review is pretty dead on with everything about the S50. I find they are great systems to use for an office environment, but anything that needs a lot of performance prob. should look at something higher end.Shalmanese - Thursday, October 23, 2003 - link
It has always been my private opinion that Content creation is only related to typical buisness usage in that it has the same range of applications. However, saying that a 10% difference in content creation will translate into a 10% difference in real world buisness usage is like saying a 50% increase in memory bandwidth will lead to a correspondingly large increase in bandwidth intensive applications. The dillema is that any benchmarking utility that simulated TRUE desktop performance would be of no use as a benchmarking utility. Over the span of an 8 hour workday, the difference between a fast and slow computer may be 30 seconds worth of extra wait time if that. Also, the S50 is offered at every speed from a 2Ghz Celeron to a 3.2Ghz P4. Obviously, they would ship you the most expensive model to review but the vast majority sold are going to be in the mid range.Anonymous User - Wednesday, October 22, 2003 - link
The lack of CPU cooling fan concerns me. I've got a couple hundred SFF Dell OptiPlex machines throughout my office and I sleep better knowing the CPU has its own active cooling. Too many machines are shoved into areas that don't provide enough airflow for effective convection cooling, and we all know what that eventually leads to.And as for the 2-3 year lease deals, I think one of the good things to come out of the "dot bomb" era is that companies are demanding more from their equipment. Short-term lifespans may be great for the business models of technology companies but those of use in old-economy industries have a different idea of what a machine's expected lifespan should be. Mind you we're not using Commodore 64s for anything, but a four-year-old P3/500 runs Windows 2000 and MS Office just fine for your typical office worker. And that is where a business-oriented machine really shines- it's much easier to keep a fleet of old OptiPlexes or HP Vectras running smoothly than a hodge-podge of no-name machines.
sprockkets - Wednesday, October 22, 2003 - link
The only thing I care about is how a stupid 3.2ghz 82 watt minimum processor can be cooled with such a dinky heatsink and quiet small fans. It like all other OEM cooled processors run at around 80c. They won't let you see the temperature for obvious reasons, since while a processor can run at 80c it will last much longer at cooler temps. And if you want stability you need to keep stuff cool.Also a 3db increase in sound is not twice as loud. A 10db increase is. But that doesn't take into perception of how annoying something can sound. For instance, a computer I have at home is just a little wider than a pci slot, so it has it's cd-rw drive vertically. The ps fan on it is on the outside and is only 60mm and 10mm thick. It gets very loud due to it getting rid of the heat in the system. In fact, it's loud whether I have a 800mhz duron or 1800+ Tbred. But then I took it to a reception hall and put it on stage I could barely tell it was on. So while it may sound quiet in a open building in a business, it will sound very loud at home. But that's just me and one observation.
Anonymous User - Wednesday, October 22, 2003 - link
Interesting review. I just picked-up a stack of Small Forms for my company…the IBM small form was one I rejected, basically because it looks so bad….amongst other rejected systems including HP/Compaq, Dell(which is now absolutely prohibited from conducting business with my company – different story) and Micron.I ended up, at $600.00 per machine, with Gateway E4100s. Celeron 2.4s with 256 dual-channel DDR, Intel MoBo and basically the same chassis….only not as tacky looking as the IBMs. They’re also completely silent.
Anonymous User - Wednesday, October 22, 2003 - link
I implement machines for small to mid-size companies in the Pacific Northwest. My recommendation and what I am hearing in the field is that there is very little need to renew leases from 2-3 years ago. User's themselves are seeing little to no benefit for getting the latest and greatest. People want their jobs, not more PC's brought in. There really aren't many apps that take advantage of the speed for most cases. People aren't rendering here or playing games, folks.Having said that we buy mostly small form factor PC's, and we buy mosltly HP. We used to buy Dell but saw their support go absolutely downhill in the last 2 years. Without Support why buy from these companies at all? Anyway, now we're just considering getting shuttles or vanilla brand. The only parts that really fail anymore are HDD's (and do they ever fail, the failure rate is about 10% easy across all manufacturer's of IDE, which is a lot)
We can save a customer roughly $200-$300 by just getting no-name brand boxes like shuttle, etc, over IBM/HP. It's something we're seriously considering.
Anonymous User - Wednesday, October 22, 2003 - link
Regarding the mysterious lower performance of the IBM S50, I think the clue must be the memory speed because 320MHz memory is bizarre when dual 400MHz is the design for this processor. This means one of two things...1) Asynchronous mode - if the S50 uses asynchronous memory timing, we all know that this reduces performance. We've seen many tests where 266 MHz synchronous is faster than 333MHz asynchronous. Also, to my knowledge, the P4 3.2 is made for dual channel 400MHz (800MHz effective) so running a 320MHz DIMM in single channel (320MHz) or dual channel (640MHz) mode WILL DEFINITELY hurt performance since the long pipelines in the P4 and the very high 3.2GHz speed are very dependent on avoiding any kind of wait on memory (a lower speed P4 such as 2GHz would be affected less). If true, this would be a double oversight.
OR
2) Synchronous mode - if the S50 uses dual channel memory in synchronous mode and they limit the speed of each bank to 320 MHz, then the fixed multiplier of the CPU must result in an actual CPU clock speed of 2.56 GHz rather than 3.20 GHz. If true, this is a simple case of underclocking.
Either way this is a strange decision by IBM.
--charlesz (waiting on my AnandTech password).
Wesley Fink - Wednesday, October 22, 2003 - link
As stated in the review, stability and trouble-free operation are MUCH more important to IT Departments than performance will ever be. We did not test the IBM S50 as we would an Enthusiast machine since that is not appropriate. In fact, we ONLY ran Content Creation and General Usage benchmarks because these are made up of the kinds of applications Corporations normally use on their desktops. The IBM was at least 10 to 20% slower in those benchmarks than any other 865/865G we have tested. That is significant enough that we think it WILL interest some, if not all, IT departments.The IBM deserves the praise we gave on Engineering and low noise levels, but someone should be questioning the dismal performance we found in Corporate applications suites. If my IT department specified 3.2GHz P4s I would certainly expect to see performance in that neighborhood, and not performance more typical of a 2.6GHz CPU.