*4-17-2003 UDPATE* First Reviews Ready. Intel 865/875 (Springdale/Canterwood)

[bmg]

Based on the gaming benchmarks in the 8KNXP review it looks like the Abit IC7-G underperforms the others by 1-2%.

 

[wutou]

Originally posted by: iamTux
Thanks Evan for clearing this one up. I was quite sure it was just some rudiment of former PE tests.


I was referring to the new FSB 800 CPU: 2400 = 18*133 vs. 12*200. In your Test of the ABIT BH7 from April 6th you mention:

We're happy to report that the ABIT BH7 can indeed reliably operate at 800MHz FSB. In fact, the BH7 was able to reach just over 218MHz FSB (that's about 872MHz FSB for those of you that are mathematically impaired ).

I was thinking the FSB of 218 Mhz would maybe be the limit of this particular Mobile P4. Now you have a P4 FSB800 for testing which limits are far obviously far beyond (matched 240MHz FSB in the Gigabyte 8KNXP test). This looks much more like the actual limit of the chipset rather than of the CPU.

Can you check back and see how far a 845PE can be pushed? Maybe even the ABIT BH7 from your test if it's still available. This would be very informative. Thank you.

I second that.

 

[YJ]

Just bought AOpen's AX4C Max (Canterwood board) for $199.

- 4 Serail ATA 133 with RAID (0,1,0+1)
- GbE Lan
- IEEE 1394
Tons of feature.

Works great with my old 533 MHz P4 CPU.
Now all I need is to wait for the 2.8 800MHz from Intel.

 

[MegaCharger]

Hi I'm new here

Based on the scores of the benchmark, I don't see any major difference between all the mobos. I will be getting a new PC in May, probably with a canterwood, radeon 9800 pro, and 3.0ghz. I'm looking for stability here, so which one is the best bang for the buck?

 

[jaeger66]

Originally posted by: MegaCharger
I'm looking for stability here, so which one is the best bang for the buck?

Unknown, since right now only a few are available and are artificially expensive. Which ones are a good value won't be known until supply is in full swing, a few weeks maybe.

 

[WizardNJ]

I finally have stability on my AOPEN AX4C-Max with 3.0 GHz/800 MHz fsb cpu and 2x OCZ PC3200 Extreme Low Latency ram.

Originally had the Corsair XMS3200LL matched pair and they would not run at 400 MHz @ 2.5 volts with memory "by SPD" in bios. The OCZ is now 100% stable at default 2.5 volts, 400 MHz and "by SPD" in bios. I did notice the latency is slower than the Corsair. I did try to slow the Corsair down to the same latency as the OCZ but it still would not run.

My recommendations are to avoid the Corsair XMS3200LL memory on i875 chipset motherboards.

Dave

 

[Fallen Kell]

WizardNJ, I tried getting in contact with you on your other thread, are/were you using 2x TWINX1024-3200LL = 2 gigs RAM or were you just using 1x TWINX1024-3200LL = 1 gig RAM? Getting conflicting information about where/when the problem actually happens. Corsair is currently saying the problem is happening when using 4 modules of TWINX RAM (i.e. 4 dimms). Can you confirm that you had a problem with 4 dimms or was it worse and the problem may also occur with 2 dimms?

 

[Wurrmm]

That OCZ memory problem could be just the AOpen mobo, or your mobo, the bios, or the memory. Anands test used Corsair so suggesting that people avoid them is un warrented IMO. Corsair make great memory as well as OCZ. There is little reason for the problem.

 

[cobrasvt]

Actually the problem is with Corsair's low latency modules (all TwinX modules and the stand alone LL modules). We found this same problem when using the fast memory controller of the SiS655 chipset as well. A lot of the guys running 655 boards were reporting that the boards weren't booting properly and locking up upon post with the LL memory set to AUTO in the bios. Upon further tests done by Prom and BigToe over at Asusboards found that setting the memory to manual and changing settings to 2-7-2-2 or 2-8-2-2 fixed the problem. Only problem was was that you needed an older stick of slower memory in order to get into the bios to change settings. The problem doesn't find its way onto Granite Bay as most memory can run aggressive timings running at DDR266. Its only when you get into the 400+ range where the problems warrant. After reading the blurb from Corsair on the problem and recommendation, I felt as if they were saying to stair clear of all LL memory for the Canterwood. Corsair will probably go back to the drawing boards now and revise thier SPD timings so that TwinX and LL memory can be used with Canterwood.

To be safe, I called Googlegear this morning about swapping out my TwinX LL for XMS 3500 and they said not a problem. Should be here on Monday.

Reread his message as well because if you notice, he changed FROM TwinX LL TO OCZ and now the problem is solved! Ala, Corsair's recomendation stands true

 

[WizardNJ]

The Corsair TwinX 256-3200LL (two matched 256 meg modules for dual channel operation) would not run at default 400 MHz, 2..5 volts and default 2,6,2,2 timing. Bumping the VDIMM voltage to 2.7 volts allowed booting to the OS and everything ran fine until a reboot or if the PC was shut off for a lengthy period of time. At that point the cmos would be corrupted.

I swapped out the Corsair for the OCZ 3200 EL (two 256 meg matched modules for Dual channel DDR) and can run 400 MHz, 2.5 volts 2,6,3,2 stable. The RAS to CAS latency is 3 compared to 2 on the Corsair but everything is 100% stable. The Corsair would not boot at 400 MHz, 2.5 volts, 2,6,3,2. I'll give up one clock cycle on the RAS to CAS timing to ahve everything stable at default voltages

Dave

 

[Wurrmm]

That is fascinating.....Hmmm....I hope the problem is resolved before I get my system.

 

[WizardNJ]

Between Pastor Jay, PrometheusN, and myself we'll have all the kinks ironed out before you get the system

Dave

Originally posted by: Wurrmm
That is fascinating.....Hmmm....I hope the problem is resolved before I get my system.

 

[Wurrmm]

It is also interesting that Anands review uses Corsair twinX PC3200. I wonder what they have to say. And to think, I almost ordered a TwinX1024 PC3200LL kit today.

 

[sumrtym]

It was my understanding that Intel was being shall we say....iffy...on DDR 400 and only recommended DDR 400 at CAS 3 as guaranteed to work? That could be why you're seeing the OCZ at 3 work and the Corsair not.

 

[Wurrmm]

Hey Evan.....I was wondering if you have anything to say about this Corsair thing and your review??? Did you run into similar problems???

 

[mechBgon]

Evan, why isn't there an i850E/PC1066 platform in the tests? You know it's faster than i845PE/PC2700, and you do have one to test with, unless you guys got rid of it or fried it...

 

[wutou]

Evan, in your new Asus P4C800 review, you said the asus board implements CSA. But it is not what I have heard. Can you confirm this?
Thanks a lot.

 

[First]

Originally posted by: mechBgon
Evan, why isn't there an i850E/PC1066 platform in the tests? You know it's faster than i845PE/PC2700, and you do have one to test with, unless you guys got rid of it or fried it...

Here's page 8 of our 8KNXP Ultra/P4C800 Deluxe review:

Unfortunately we were unable to find a SiS 655 motherboard (and that includes the Gigabyte SINXP1394 used in our 800MHz FSB preview) that could run an 800MHz FSB CPU, with or without HT enabled. We were also unable to find an E7205, 850E, or SiS R658 motherboard that could run these same 800MHz FSB processors. There's a possibility that your motherboard vendor will release a BIOS update to enable 800MHz FSB support on these boards, but we've been told this is highly unlikely, so don't hold your breath. Therefore, users should not expect their current SiS 655, SiS R658, E7205, or 850E motherboard to run an official 800MHz FSB Pentium 4 processor.

Wurrmm, Corsair just informed me two hours ago about their take on the situation with TwinX LL modules:

1. Essentially, Corsair considers the current crop of 875P boards to have immature BIOSes (which is almost always true of any motherboard that has just been released). Intel has (or is about to) release a new BIOS that corrects the SPD detection and frequency detection issue with LL TwinX modules with their 875P board. Apparently other motherboard makers will follow.

2. My contact at Corsair would actually prefer that the LL TwinX SPD wasn't programmed so low (CAS2/2T/6T/2T), and instead at CAS2/3T/6T/2T. This would have solved a lot of nForce2, SiS 655 and now 875P problems, but of course LL TwinX modules are all programmed at CAS2/2T/6T/2T so it's too late. The vast majority of nForce2 motherboard makers have released BIOSes that correct the LL TwinX issues, and I think some SiS 655 motherboard makers have to. Now 875P board makers will as well.

You should also know that you should only have an issue with LL TwinX modules if you're leaving your memory timings at SPD instead of programming them yourself. Program your memory timings (CAS2/2T/5T/2T if you want) and save it to CMOS and you'll be just fine.

I have personally never experienced issues with LL TwinX modules, but that's because I never use SPD anyway.

wutou, the ASUS P4C800 review mentions that this board does not utilize the 875P's separate Gigabit bus (CSA).

 

[Wurrmm]

"3COM 3C940 Gigabit LAN (no CSA)"

"The LAN chip is powered by 3COM 3C940 Gigabit LAN chip, which unfortunately does not utilize Intel's own CSA bus technology. As you can see from our Intel 875P chipset article, CSA does indeed aid in Gigabit performance. We honestly don't know why ASUS opted not to include CSA expect to say that it may have decreased ASUS' time to market and/or costs to produce a non-CSA solution."

That is what it says on the review.

 

[Scroatdog]

I really don't understand what the difference is CSA versus no CSA. I read a little bit about it, and understand what it does... I THINK. It keeps the LAN from robbing bandwidth from the PCI bus, right?????

Let me pose this question...... if you were running SATA drives in say, a RAID-0 config, well, let's just make it easier and say just SATA..... wouldn't it be fair to say that you wouldn't see a performance hit because your hard drives are NOT running from the PCI bus???? Since the SATA does not go through the PCI bus?????

Please educate me. I am itching to buy a board to go with this 3.0 Ghz 800 Mhz CPU sitting on my desk. I hear goot things about Abit, but where is it??? One thing I did not like about the ASUS was the proximity of the rather large Northbridge heatsink in relation to the CPU socket. That would make installation of a Zalman 7000Cu or other large aftermarket HSF next to imporssible to mount........

 

[wutou]

Sorry my bad. I guess I just read too many reviews.

 

[Wurrmm]

Thanks Evan!! Time to get that newegg order in. TwinX1024 PC3200LLPT $299. Hoooaah!!

 

[Fallen Kell]

Evan, thanks for the info from Corsair. Hopefully this will correct the issues...

 

[jaeger66]

Originally posted by: Evan Lieb


2. My contact at Corsair would actually prefer that the LL TwinX SPD wasn't programmed so low (CAS2/2T/6T/2T), and instead at CAS2/3T/6T/2T.

Funny, I recommended the same thing to both Corsair and Mushkin and they said that they used to do it that way but they got too many returns from people thinking they'd been ripped off.

 

[jaeger66]

Originally posted by: Scroatdog
I really don't understand what the difference is CSA versus no CSA. I read a little bit about it, and understand what it does... I THINK. It keeps the LAN from robbing bandwidth from the PCI bus, right?????

Yes.

Let me pose this question...... if you were running SATA drives in say, a RAID-0 config, well, let's just make it easier and say just SATA..... wouldn't it be fair to say that you wouldn't see a performance hit because your hard drives are NOT running from the PCI bus???? Since the SATA does not go through the PCI bus?????

Yes, but ONLY for the SATA ports integrated into the ICH5. SATA from something like an SImage or Promise chip is PCI. Still, a Gigabit LAN controller eats up almost all the PCI bus by itself when running full speed. And then you have sound cards, capture cards, SCSI cards, TV cards...it just depends on how heavily loaded the PCI bus is on your system.