Which of these old Intel CPUs is better?

[Shad]

I acquired some old hardware from work today which I'm hoping will be useful for building a NAS server. I have the following Intel chips but I'm not sure which one is best suited...

• Pentium 4 530J (part SL7PU) - 3.0 GHz, single core with HT (so OS sees two cores, right?), x86 only, 1 Mb cache, 800 MHz bus
• Celeron D 351 (part SL7TZ) - 3.2 GHz, single core (no HT), x64, 256 Kb cache, 533 MHz bus

I had hoped to use Server 2008 R2 but that hangs on whether or not an x64 Celeron is better than an x86 P4. The mainboard that will be used is an Intel D101GGC which is limited to 2 Gb of ram anyway. Planning on using a PCIE RAID controller with 8 drives attached, not decided on which one yet though.

Any thoughts???

[alsenior]

does not rally matter you have to use the celeron for r2 as it's 64-bit only

[Shad]

Correct but if the performance advantage of the P4 is significant I will use Server 2008 instead. The mainboard only supports 2 Gb of ram so that's one of the key x64 advantages out the window straight away.

[alsenior]

go for the p4. more cache means better performance in file serving tasks

[scaryjim]

The P4 will have much better general performance due to its higher cache and FSB - also I strongly suspect that newer versions of Windows Server are optimised for multithreaded performance so the HT should have an impact too.

There's a reason Celerons are cheaper than P4s, you know

[jordan427]

P4 definitely, celerons were awful!

[scaryjim]

... celerons were awful!

That's a bit unfair on the Prescott based Celerons, they were competitive at their price point (i.e. against the Thoroughbred-based Semprons).

Sure, they couldn't compete with Barton Athlon XPs, full Pentium 4s, and Athlon 64, but they were a lot cheaper than any of those chips! And they'd still beat the rear end of an Atom

[Shad]

Jim, it just so happens I have a Celeron D that's spare - you buy now yes?

[scaryjim]

Sorry mate, got 6 sitting in a box downstairs waiting for a bundle of cheap motherboards to come up so I can sell them on myself

Although I admit a 64bit one is intruiging...

[Syllopsium]

HT is faster than non HT in practically all scenarios, and even in those, it's not slower by much. Whether that's 200MHz worth of performance is a good question, but I'd still go for the P4.

[kalniel]

HT is faster than non HT in practically all scenarios, and even in those, it's not slower by much.

We're talking about the same HT that server people disable to get 20% performance improvements here? This isn't the new i5/7 HT, the p4 hyperthreading was terrible.

[shaithis]

We're talking about the same HT that server people disable to get 20% performance improvements here? This isn't the new i5/7 HT, the p4 hyperthreading was terrible.

Yup. One of the main applications that was always said to be a great showcase for multi-threading was SQL Server.....yet Microsoft made us disable HT on all our P4 era Xeons before they would look at any support cases we had.

[Syllopsium]

Can you provide some references for that? I thought the same thing too, but last time I went looking the actual results I saw showed minimal impact in non ideal cases.

PK

[kalniel]

Can you provide some references for that? I thought the same thing too, but last time I went looking the actual results I saw showed minimal impact in non ideal cases.

PK

http://news.zdnet.co.uk/hardware/0,1...9237341,00.htm

"It's ironic," said Ibbotson. "Intel had sold hyperthreading as something that gave performance gains to heavily threaded software. SQL Server is very thread-intensive, but it suffers. In fact, I've never seen performance improvement on server software with hyperthreading enabled. We recommend customers disable it when running Citrix and our software on the same server."

[Syllopsium]

Thanks. I've also found :

http://blogs.msdn.com/slavao/archive...12/492119.aspx
http://sqlblog.com/blogs/kevin_kline...ql-server.aspx
http://support.microsoft.com/kb/322385

The upshot is : it may help in some circumstances and hurt in others. Test your workload to be sure. If want to be safe, disable it.

For file sharing, I'd probably keep it enabled.

PK

[kalniel]

For file sharing, I'd probably keep it enabled.

As long as your load was low I'd agree. But if you get close to 100% CPU load I'd disable it immediately. We used p4 era Xeons for file servers and there was about a 20% hit if HT was enabled when the load approached 100%.