Pair of 1.6 GHz Xeons

[DocZavage]

Well, I guess the adapters are a non-issue for me because I just got a UPS confirmation from Orbit Micro!!!!!1!!11!!1:thumbsup::thumbsup::thumbsup::thumbsup::thumbsup:

Now I just need to order my motherboard.

 

[DocZavage]

Originally posted by: GLeeM
@DocZavage
Welcome to the forum!
...

@MichaelD
The ram came early so I tried it in my P4 3.0@3.3 to see if I could get a better OC. It corupted my logon in WinXP I didn't know that could happen!?!?

Thanks for the welcome

And you should try memtest86+ from http://www.memtest.org/ to see if your memory is corrupting windows.

 

[TerryMathews]

Originally posted by: lhotdeals

Originally posted by: TerryMathews

Originally posted by: lhotdeals
Hey guys, are the overclocks to 2.5/2.6 GHZ stable enough to use as a dedicated file/application server? If so I am so jumping in if they ever get more in stock, else I will have to go with dual oppy for a bit more $$

Nothing is absolute when overclocking; you might get that one dud. That said, almost always people seem to hit 2.5GHz at stock voltage at 2.8-3.0 at 1.6V and stable enough to run F@H.

Dunno what universe you're living in, but a dual Opteron setup is going to cost more than "a bit more $$" compared to this setup.

Also, if anyone's wondering: The PC-DL comes with two plastic adapters to allow the mounting of a stock Intel S478 heatsinks. I checked, the fit is perfect.

Dual oppy will cost 400 more but its for business use, so 400 is not much

I just don't want anyone to get the impression that a dual Opteron rig is close in price to this deal; it's not. At $400 more, it's more than double the price of a pair of these Xeons and a PC-DL and a pair of fans.

 

[MaskedAvenger]

My Cooler Master heatpipes have been Fed Ex'd from Orbit Micro today as well.

Regarding the adapters supplied with the ASUS PC-DL, they are too narrow for an un-modified stock Intel 478 heatsink. Maybe I'll dremel one up to see how they work!

 

[TerryMathews]

Originally posted by: MaskedAvenger
Regarding the adapters supplied with the ASUS PC-DL, they are too narrow for an un-modified stock Intel 478 heatsink. Maybe I'll dremel one up to see how they work!

?

I managed to connect the heatsink off a 2.4C P4 without any problems using the two plastic clips included with my rev 1.05 PC-DL.

 

[splice]

Originally posted by: MichaelD

@Splice
It's amazing that you're getting such good performance and temps with essentially your case fans configured incorrectly. Really. In 99.99% of setups out there, both desktop and servers, the front fans suck in (intake) and the rear fans blow out (exhaust) I guess your PS is the only exhaust in your case, then...what are your case temps? Your situation is just...so strange......that it's unusually incredible.

Hmm.. I've done some testing with both my Dual AthlonXP and this Dual Xeon setup. Both run cooler having the rear fans blowing in, the front blowing in, and having the 2x fans in the PS exhaust the air. In some ways it makes sense, you want to draw in cooler air in from down low and draw hot air out at the top.

Before my dual setups, I always ran front to back.

 

[MichaelD]

Thanks for the additional info, Splice. That's just wild...it goes against everying I've ever seen.

The way you're running, the fans in the PS are sort of acting like having a big 120mm fan on the top of the case...heat DOES rise. Hmm. This is very interesting! I wouldn't do it myself, but if it's working for you, then good luck! :beer:

Anyone else get their Xeons running yet? What kind of stable OC did you get?

I have to live vicariously thru you guys while I save up for a NCCH-DL.

 

[XeonTux]

2.8 GHz, 200x14, 1:1
1.55v (1.49v real world)
Corsair Value Select VS1GBKIT400 CAS2.5 2.5v
Asus PC-DL 1.05

NO BSEL mods, jumper does the 200FSB
NO VDIMM mods, Corsair Value Select doesn't need it

I'm running some extended stability tests with LV Xeons #1273 + #1274 now. Everything seemed cool with 1.50v setting (1.45v from PC-DL) , but XP froze up once. Not sure if it was because of HW prob or just being windoze. I bumped the voltage up to 1.55v like I had planned to do later anyways, to hopefully ensure a 100% rock solid system. In retrospect I wish I wasn't so quick to bump it to 1.55v (1.50v)

Temps get to 46c and the top-mounted IWT fans kick in, keeping it from getting any hotter. I think I will look into adapting 80mm adapters (yes the adapter needs to be adapted) and using a larger quieter fan

I had bought 4 procs, to pick the best two. Well they all do 2.8. The non-sequentially numbered ones will be for sale soon.

It is actually kind of strange that they all end up at 2.8GHz stable. I ran many tests on individual procs at various voltages and speeds. They had different characteristics. But none are stable at 3GHz. I had big hopes for #1274. The thing does 2.66 GHz 166x16 on default voltage, as in 1.25v from the PC-DL VRM. Heh, even had it once at 14x133 at a mind boggling 0.88v from the VRM. Ran windows 200x15 at 1.50v (1.45v), unstable. At 1.60v setting I thought would blow my mind, but I get same errors.


Has anybody gotten 3GHz+ from these? And if so, are you 100% positive it is 100% stable? Prime95? PIfast? memtest?

 

[Vette73]

^


I be willing to bet your ceiling is caused by your board, NOT the cpus.

I THINK? some people have got to the 3Ghz range over at 2cpu.com but that was on the NCCH-DL board, which has 2 4phase power VRM, unlike 1 4phase for the PC-DL

 

[Dubb]

mine do 13x200 stable...but much more than that and prime 95 errors out on one chip

windows is pretty stable at 14x200, it took alot of stress testing to get it to error out. load temps hit 54c in my crapily vented case...

 

[user1234]

Originally posted by: Dubb
mine do 13x200 stable...but much more than that and prime 95 errors out on one chip

windows is pretty stable at 14x200, it took alot of stress testing to get it to error out. load temps hit 54c in my crapily vented case...

correct me if I'm wrong, but 13x200 is 2600, or in other words 2.6 GHz, far short of 3 GHz. And if you're using this thing as a server you probably want to run it 100% stable, so better not take the risk with higher clock speeds if they cause prime95 to fail, because next it will be something critical.....

 

[TerryMathews]

Running stable at 16x165. Won't POST at 200MHz. I could probably work at it some more, but the tighter CAS timings I can run at 165MHz vs. 200MHz will probably make up for the difference anyway.

EDIT: nvm, I'm a big idiot. Now at 200x14 for a total of 2800MHz. 200x15 will start to boot and then bluescreen.

To those who have trouble booting at 200MHz, a few words of wisdom:
1. Swap the CPUs around. The initial burst of default multiplier * FSB (3200MHz in the case of the 1.6 LV Xeons at 200MHz FSB) only affects CPU1. If CPU2 hangs, it doesn't matter; it gets restrapped anyway.

2. The memory settings are confusing. When I was troubleshooting, I would jumper back to 133MHz, set the multiplier I wanted in the BIOS and set the memory to DDR333 and CAS timings at 3 8 4 4. I had bad luck running auto and leaving the timings on auto when testing new settings, seemed to hang the system a lot.

On a more general note, why can't these companies give us general ratio settings instead of specifying a speed? I'd much rather set the ratio directly instead of trying to remember what FSB I'm going to, to determine what the setting I've picked will imply the ratio to.

 

[XeonTux]

TerryMathews> On a more general note, why can't these companies give us general ratio settings instead of specifying a speed?

Aye! With some of the FSB BSELs (IIRC 166) it was downright confusing at first

 

[TerryMathews]

I've beat on this thing all day long @ 2.8GHz, can't believe how stable it is. Processors are running at a cool 40c (idle, not under load) under the newest Intel IWTs (Bought on eBay, from a 3.2GHz 2MB Xeon).

Could crank the FSB up more (probably), but I know from experience my Kingston HyperX 3200 2.5 isn't good for much over 200MHz. That combined with the low Vdimm voltage the PC-DL provides means I should stay where I'm at. Might tweak the CAS settings though - I'm at 2.5 8 3 3 (SPD settings).

If this deal pops back up, I can say with certainity it's the best bang for the buck right now. Getting a pair of 2.8GHz Xeons, a motherboard, and fans for $330 is simply insane. It's about even with the price of a P4 2.4c and motherboard, and pretty close to twice the performance if you multitask.

 

[xeonx2]

If anyone has a pair of these they want to sell, please pm or email me. A deal fell through and now I only have a PC-Dl with no procs.


Dan

 

[TerryMathews]

Originally posted by: xeonx2
If anyone has a pair of these they want to sell, please pm or email me. A deal fell through and now I only have a PC-Dl with no procs.


Dan

No offense intended, but I wouldn't sell you a pair of these chips for $300. This has got to be Hot Deal of the Year.

 

[user1234]

Can you change the subject to "DEAD" as this deal is long gone now (as it was from the start) ?

 

[MaskedAvenger]

How do you change the voltage on a PC-DL? Is it a jumper, cause I don't see anything in bios?

 

[Lurker1]

Originally posted by: TerryMathews

Originally posted by: lhotdeals


Dual oppy will cost 400 more but its for business use, so 400 is not much

I just don't want anyone to get the impression that a dual Opteron rig is close in price to this deal; it's not. At $400 more, it's more than double the price of a pair of these Xeons and a PC-DL and a pair of fans.

The Opteron setup would cost less than $300 more, and get you better real world performance. Check out Inforworld's article for the numbers.

Yes, in special cases the best Intel has to offer beats AMD, but not in real world applications. Also, the opteron setup would let you exceed 4GB, if that's a concern. These CPUs are not in the same league. (And these are the older, 32 bit only Xeons.)

 

[MesaMirageII]

Actually Lurker1... some of the exact reasons the test results in the link you quote show AMD in front of Intel are solved by these "older" Xeons being overclocked up to 200mhz FSB. If Intel is unwilling to bump the FSB, due to the marketing department wanting to keep a performance spread between Itanium and Xeon, we will just have to help correct that by bumping the FSB on our own.

Good luck with your expensive "Newer" AMDs performance vs my "older" cheap dually Xeons.


MesaMirageII

 

[Vette73]

^


I disagree abiu the reason for not upgarding the FSB.

Intel wants it to be 100% stable. So by keeping the FSB down juts a little that helps the chips. It also helps board makers as it is not that easy to find 100% tested PC3200 Reg ECC DDR. BUT finding ram that will run 100% at PC2100 or PC2700 was not hard. It has juts happened that MORE memory makers have produced wider ranges of Reg ECC PC3200 ram.


I also would rather have a dual opteron system, but the cheap Xeons I got and the fact I can reuse my standard desktop DDR ram makes it to cheap, for ME, to pass up.

 

[TerryMathews]

Originally posted by: Marlin1975
Intel wants it to be 100% stable. So by keeping the FSB down juts a little that helps the chips. It also helps board makers as it is not that easy to find 100% tested PC3200 Reg ECC DDR. BUT finding ram that will run 100% at PC2100 or PC2700 was not hard. It has juts happened that MORE memory makers have produced wider ranges of Reg ECC PC3200 ram.

I'll grant you that there wasn't much in the way of ECC buffered PC3200+ memory when these Xeons came out. That said, I'm curious about your statement that a lower FSB "helps the chips". Please inform me what part of a P4 cares if it's running quad-pumped at 100MHz instead of quad-pumped at 200MHz.

 

[Vette73]

Originally posted by: TerryMathews

Originally posted by: Marlin1975
Intel wants it to be 100% stable. So by keeping the FSB down juts a little that helps the chips. It also helps board makers as it is not that easy to find 100% tested PC3200 Reg ECC DDR. BUT finding ram that will run 100% at PC2100 or PC2700 was not hard. It has juts happened that MORE memory makers have produced wider ranges of Reg ECC PC3200 ram.

I'll grant you that there wasn't much in the way of ECC buffered PC3200+ memory when these Xeons came out. That said, I'm curious about your statement that a lower FSB "helps the chips". Please inform me what part of a P4 cares if it's running quad-pumped at 100MHz instead of quad-pumped at 200MHz.

Some chips can run a faster FSB then others. That can be based on the CPU, motherboard/chipsets, ram selection, etc...

So its not the CPU as a whole, but the WHOLE system that intel has to look at. Even the first batch of opterons AMD would only gurantee they worked with PC2700 ram. That way they knew 100% of their chips they sent out would work. Now that the chipsets are tested, MORE reg. ecc ram is out, slight update to the core (CO - CG) they, AMD, say the opterons will work with PC3200 ram.

1 crash with a NEW system equals a system admin not wanting to use that type of hardware again, or only in a very small manner. So Intel, and AMD, play it safe with their Server systems.

 

[TerryMathews]

Originally posted by: Marlin1975

Originally posted by: TerryMathews

Originally posted by: Marlin1975
Intel wants it to be 100% stable. So by keeping the FSB down juts a little that helps the chips. It also helps board makers as it is not that easy to find 100% tested PC3200 Reg ECC DDR. BUT finding ram that will run 100% at PC2100 or PC2700 was not hard. It has juts happened that MORE memory makers have produced wider ranges of Reg ECC PC3200 ram.

I'll grant you that there wasn't much in the way of ECC buffered PC3200+ memory when these Xeons came out. That said, I'm curious about your statement that a lower FSB "helps the chips". Please inform me what part of a P4 cares if it's running quad-pumped at 100MHz instead of quad-pumped at 200MHz.

Some chips can run a faster FSB then others. That can be based on the CPU, motherboard/chipsets, ram selection, etc...

So its not the CPU as a whole, but the WHOLE system that intel has to look at. Even the first batch of opterons AMD would only gurantee they worked with PC2700 ram. That way they knew 100% of their chips they sent out would work. Now that the chipsets are tested, MORE reg. ecc ram is out, slight update to the core (CO - CG) they, AMD, say the opterons will work with PC3200 ram.

1 crash with a NEW system equals a system admin not wanting to use that type of hardware again, or only in a very small manner. So Intel, and AMD, play it safe with their Server systems.

The portion I've placed in italics is wrong; it's the error I was trying to point out to start with.

A CPU doesn't care if it's running 1600MHz by means of 16x100MHz or 8x200MHz. The data IO path connecting the chip to the northbridge is the only part of the CPU that doesn't run at the CPU frequency, and it's running at a much lower frequency so it's all golden.

I've already agreed with you that up until recently there wasn't much in the way of ECC buffered RAM at fast speeds. OTOH, if Xeons have been able to work with the i82875 chipset all this time, Intel really fubared their chances to compete with AMD on the workstation level. A series of cheap Xeons, basically repackaged P4s at a slight price premium and dual-processor i82875 motherboards. That would have been very competitive with AMD's pricing on the Opteron 2xx chips.

And finally, we keep getting back to using these systems in a high-availibility environment. I disagree with you and a couple of other people; anyone who needs a system with uptime like your talking about wouldn't build their own. They would buy from Dell or IBM with one of their platinum on-spot replacement warranties.

Back to the topic at hand though, this is (was) by far the best SMP bang for the buck. I'm willing to run any benchmark that any Opteron user is, to prove my point.

 

[imported_gimper48]

This is DEAD DEAD DEAD let it die