Sign in to follow this  
steve8

TestBed4 ideas

Recommended Posts

i think u only have two real choices currently.

sata1 w/ ncq

ich6® via 915g,915p, or 925x or 925xe

intel

or preferably:

sata2:

nforce4 ultra or nforce4 SLI

these are both mainstream native sata interfaces and dont suffer from any bandwidth or latency limitations of the pci bus.

for intel suggest a pentium4 3.2Ghz 540J (1MB and nxbit) or wait for the pentium4 640 (2MB cache EMT64 along with nxbit and enhanced speedstep)....

or for the amd platform the athlon 64 3500 939 90nm

1GB of DDR pc3200 (2x 512MB)

both platforms also give u pci-e 1x and in some cases 4x for an additional drive controller.

the intel platform would use a pci-E 1x based GbE controller, while the nforce4 would have a chipset GbE controller w/ external PHY...

get a cheap pci-e based video card, maybe an ati x300SE.

thoughts?

Share this post


Link to post
Share on other sites

0.o

Testbed4 would imply that he's going to use it as the main vessel for testing review samples for this site for the next few years.

Edited by squalish

Share this post


Link to post
Share on other sites
0.o

Testbed4 would imply that he's going to use it as the main vessel for testing review samples for this site for the next few years.

195865[/snapback]

Or that he's making suggestions regarding it :)

Any chance the admins could make editting posts an indefinite thing?

Share this post


Link to post
Share on other sites

You know, Xbitlabs used their testbed 1 until 2004: PIII600 on P3B-F (440BX) that came out in 1998. No SCSI, SATA or ATA-66, 100, or 133 onboard so support was provided by various PCI Adaptec and Promise boards. While this may have left a lot to be desired, it really did offer a stark contrast to SR by showing how performance might look when CPU and bus availability was severely limited by other traffic--and when drivers were badly written you saw the CPU utilization soar! Now they have moved to... 2.4 on i865.

I think that SR needs to use a cutting-edge platform that will resemble how a power-user's will look in the future, and the AMD platform will probably be a lot more stable for the forseeable future. Intel is rolling out new chipsets for their promised dual-core chips that have been moved up to Q2 2005(!) http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2329&p=3

A DK8ES or K8WE(S2895) on the other hand, would look to be pretty standard way into 2006 and long after dual-core AMDs arrive.

One suggestion: buy the cheapest processor(s) that will run in this cutting edge platform because Xbitlabs' benchmarks suggest that storage has probably not been CPU limited for some time.

Share this post


Link to post
Share on other sites

CPU power is not important unless you are planning to test the latest software/hostbased RAID adapters. The RAIDCore BC4852 and HighPoint RocketRAID 1820A have very high CPU usage in sustained write transfer tests on large RAID 5 arrays. With both adapters the maximum write transfer rate of an 8 drive Raptor WD740GD RAID 5 array was limited by the CPU power of the 1,4GHz Opteron 240 in my test system. CPU usage scales linearly with (write) transfer rate.

Share this post


Link to post
Share on other sites

i agree with bfq9000. we want testbed4 to have some longevity, so it should be a fairly cutting-edge system. we also want its performance to be representative of what SR readers will see on their machines.

i think the clear choice is an AMD/nforce4 system. IMHO, AMD will be the platform of choice for enthusiasts in 2005. nforce4 will be the AMD chipset of choice, especially for our purposes since it supports s-ata ii natively. it wouldn't make any sense for testbed4 to support first generation s-ata only, since most of the drives released during its lifetime will likely support s-ata ii. of course, a controller card could always be added to whatever platform, but that would introduce other variables which (being the good scientists that we are) we want to avoid. so native s-ata ii support is a must, in my opinion.

Share this post


Link to post
Share on other sites

My personal opinion: the hardware should be stable and up-to-date, with quality disk controllers and PCI implementation (though I suspect that with any good quality motherboard, these issues require no real consideration). It should contain options, if possible, for potential future technologies such as PCIe. Other than that, I think the hardware is rather irrelevant.

I'm much, *much* more interested in how the software model for the testbed will change.

Share this post


Link to post
Share on other sites

Testbed 4 is here and alas, it is a Dual-Xeon platform. Intel will be replacing the current chipsets with dual-core, multiple FSB ones this summer so not sure how those will compare, performance-wise. Especially given they will have ICH7, not ICH6: http://www.digitimes.com/news/a20041025A1001.html

On the bright side, the K8WE is finally trickling into retail: 2x Dual-core capable, Dual *16-lane* PCI-E, 3 PCI-X, single PCI, 4 native SATA-II NCQ ports, and two gigabit ports. ~$450 or ~$600 w/SCSI.

http://www.gamepc.com/labs/view_content.as...nderk8we&page=1

Share this post


Link to post
Share on other sites

There's always something better round the corner... at least ICH6 has SATA2 and NCQ. I don't think Active Management Technology has much bearing on benchmarking, and there's not much else new in ICH7, according to the article bfg9000 linked.

Just as well, if Testbed 4 is already assembled!

Seriously, it would be great to get some info on how testing will be done on Testbed 4, before any testing is done, so that our resident benchmarking experts can do a little peer review.

Share this post


Link to post
Share on other sites

I think, good machine for testbed4 would be dualcpu opteron with nforce4 pro chipset. 2ghz should be enough.

Future is going to be dual core. But when you look at the first benchmarks, dualcore opteron has the same performance as the dual cpu opteron system.

So therefore I think dual cpu opteron system will be good choice for future.

Second, chipset. I think nforce4 pro will be good, because:

1. it has ata-II with ncq support.

2. many users in the world will buy nforce4 desktop chipset which has the same sata controller, so they can compare directly

3. nforce4 supports pci-e, i think in future, threre if there are going to be any new controllers, maybe SAS, threy're going to use pci-e

4. the board will have pci-x (or pci-64?), so you can use contemporary high performance controllers

Another suggestion is, to test contemporary new gen drives (wd300, 320jd, diamax10, hitachi 500gb, barracuda8, maybe new sata-II samsung and hitachi) on testbed 3, since testbed3 still provides valuable information about performance and is pretty up to date.

Then when the new testbed4 will be operational, you can test these drives again, and see how big will the difference be. I hope that with new testbed, you will update your benchmarks (and benchmark patterns) to more actual too.

Share this post


Link to post
Share on other sites

First, the ICH6 is SATA2.

To determine whether Intel or nForce4 has better SATA2, someone has to do proper benchmarks. Preferably with Microsoft upcoming AHCI driver too.

The chipsets will differ in PCI arbitration and bridging, which nobody ever tests.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this