Comment thread for DFI Infinity P965-S Dark review

[sswilson]

Originally Posted by pscout

there are lots of factors that affect how linearly cpus/mobos oc.

here is an interesting read ... and there are a few others on that site
Secrets of the i965 based Asus P5B and chipset strap uncovered. - The Tech Repository Forums

When someone claims stability because it is orthos stable, i always ask have they run smp folding at those clocks. Cuz in my experience with the c2d and especially c2q's, smp folding is the best stress tester for non video intensive use. Well, at least for me it is the only relevant stability test since i build all mine to fold. With presslers, orthos was a pretty good indicator of stability, and still is pretty good with c2d's, but with c2q's it isn't even close (4xp95 is my orthos equivalent for quads). SMP tests more than just cpu and memory ... it is heavy on interprocess communication and cache usage.

So i keep a copy of some of the more demanding SMP work units so i can use them to test oc's with.

I realize i am somewhat atypical tho, since i use virtually no intense video, and anybody building gaming/video compostion rigs would need to use video stressing tools, possibly in addition to smp like testing. If video is the bottleneck that you oc for, then orthos/4xp95 may be sufficient.

9 Min 38 sec / frame 2610............ :)

[pscout]

Originally Posted by sswilson

9 Min 38 sec / frame 2610............ :)

yes ... I know you know all about folding ... and nice frame times too

I just posted the folding reference for the other readers ... and to promote our HCF folding team a bit

[Eldonko]

Depends on what orthos test you use. If you run small fft I think you can conclude that your CPU oc is stable. I find SMP is good at picking up RAM errors mostly.

I fold too, SMP FTW! Just sucks I cant use it with Vista 64bit and have to boot an XP OS to fold. :/ Hopefully they upgrade it at some point.

[sswilson]

Managed to get over the 450 FSB stable wall I ran into.

Currently running (2 hours orthos stable... haven't run it all night for a full 8 hour run) FSB 452 for a cpu-z recorded freq of 3619.

Had it up to a windows bootable FSB 459, but even the 452 is taking crazy volts on CPU/NB/SB to get it stable so I'm fairly certain that's as high as I'm going to push it.

Something I have found strange compared to my S939 OC experience is the reaction to errors in Orthos.

I've found that it'll run fine for 2 hours + (so I know it's already hit max temps) but then sometime overnight it'll just reboot the system rather than reporting an error.

I've also noticed this "rebooting rather than orthos error" on lower OC's which just needed a slight vcore bump to overcome.

Is this the norm for C2Ds? A64s would either freeze or report an error in Orthos.... I can't recall an unstable A64 OC ever rebooting the system.

[Lithotech]

I had a fairly rare opportunity to compare the Dark against a pretty close match -- the Gigabyte N650SLI-DS4 -- using the same cpu, ram, vid, psu, etc.

This build here was originally spec'd with the P965-S Dark:

Case Pics

My experience with the board was somewhat of a letdown, we RMA'd it for an exchange to the Gigabyte N650SLi-DS4. On the other hand, it's benchmark numbers were impressive in a number of tests, at stock speed beating out the overclocked Gigabyte!

The board was purchased through NCIX in Vancouver BC. It's unfortunate NCIX didn't have another in stock, we would have tried another. Alas, they were out of stock. Why do these places wait till stock is completely out before getting more in?? Gah, it's truely inconvienient, and seems to happen to me all too often. It seems I am constantly sacrificing one part for another, because the one I want isn't in stock.

As of June 23, I see they are in stock again.

I also see NCIX has added a disclaimer to the product page:

This item is warranted by the product manufacturer only

I have to say, how much I don't like to see this.

Have there been significant enough problems with this board to warrant this? And not just isolated things? Or is DFI instigating the policy, not NCIX's decision? Before I saw this, I would have bought another one of these; now? No way. Actually, only one way; if I bought it used off another person and it was verified to function correctly.


Anyways, I got one of the last ones they had, at the time.

It was a bad squealer.

This Dark had a couple other issues, but these were possibly fixable with future bios releases or more tweaking, and also possibly due to other components or combinations:

• It would not post at 400+ x 6 no matter what, but neither will the Gigabyte so I could resign this to the CPU being a dud for higher FSB's.
• It would not run the ram stable at 1066mhz with the cpu at stock 266/1066 -- no matter what I did; following guides as per how to set that up using the NB strap, and tweaking everything else I could try in the limited time I had the board (nearly two weeks). I could only conclude that at least with that particular board/ram/cpu combo, the NB strap function simply did not work.
• The board functioned perfectly otherwise, except that it made an unacceptable squeal at load or idle, stock speed, three kits of ram and numerous settings, 2 psu's, 2 vid cards, ran outta stuff and time.

To be fair, the board was capable of some nice scores compared to the Gigabyte with similar config and speed. In fact, as you'll see in Everest screenshots below, some of the benches at stock speed exceed that of the Gigabyte's higher overclocks. You'd expect nothing less from a DFI, but the disparity is enormous in some cases when you consider the cpu speed differences. I am still inclined to suspect something funky going on with Everest, I must find time to study some of the other benches.

The Gigabyte incidentally, was able to run 1066 memory with the cpu at stock 266. In linked or unlinked mode, without a hitch, right out of the box. It's the first thing I tried. Orthos stable with 5-5-5-9 timings and only 2.1v on the ram. It does have it's own quirks, but not appropriate to list them all here.

There were a few things about the Dark that had me recommend it in the first place, and we were fairly determined that whatever we got would have to have some of these features. As with any new tech, there's bound to be growing pains, and we accept that. I told him right up front to give me a month with this build, we were prepared for delays and extensive burn-in and testing. In the end it was 6 weeks, but it was ready in a month and sat on my out-bench for two weeks until he had time to let me install it.

Some of our desired features: all solid caps, digital pwm, extensive bios settings, under $180 or so.

So the DFI Dark and the Gigabyte N650 DS4 are actually a pretty comparable match. Their main difference bieng not really so much the chipset, but one being Crossfire and the other beijng SLI. They are both all solid caps, digital pwm, great bios, and almost exactly the same price (at least at the time).

Of the smaller differences, one ironic perhaps, the CMOS Reloaded missing on the DFI is present in an even better form on the Gigabyte with their Dual Bios and Bios Setting Recovery gives you a total of 16 savable BIOS setups, 8 each for the two separate cmos chips. Additionally, you can run a different version bios on one chip, letting you test out new bios releases and still instantly boot back to the old original. One major major caveate to this is the cmos chips are soldered to the board. I really like this ability, and almost insist on it for my own systems. It is starting to appear in more and more mobos, Biostar's TForce have 50 slots? An Asus M2NPV I have has 2 slots. Oh yeah, baby, at least two slots belong on every mobo!


Aside from the squealing, and moving on from trying to run stock speed with 1066 ram, the motherboard was a very educational experience to tame. I spent a lot of time digging up and reading all sorts of info on this board, there is a lot more out there on it than the Gigabyte. However, with the Gigabyte I didn't need to dig up any info, what little I would have needed, I already learned from working with the DFI! But seriously, I printed out enough material on the DFI to fill a small 3 ring binder. I think I added maybe 3 sheets to it that were Gigabyte related.

The Dark is a little slow to boot, compounded by the double boot, but that is pretty much par for the course in decent mobos -- the cheaper they are, the faster they boot. The bios always fun to work with and plenty responsive on the desktop. Smart Guardian always a friendly face with all the appropriate voltages and info -- Gigabyte's EZTune has one voltage that may be the NB; I have no idea since it isn't labled; and in the BIOS all it tells you for volts is OK... people have been complaining about this for eras, yet they still don't fix it.

The P965 chipset being so mainstream, seems very compatible with other diagnostic software. The 650i has been around a while, and Gigabytes entry is a late release, yet Everest can't identify the mobo, and there was one or two other apps that couldn't delve the thing exactly right.

The Darks NB heatsink may get a little hot without some active cooling, but nothing like the copper one on the Gigabyte! This big block of black aluminum will never actually burn you.

With the Dark I found I had to run way more NB volts than most other owners, and had to add it much earlier. Here she is at 360fsb for 3.2ghz, ram was Buffalo Firestix 6400, at 1080mhz 5-5-5-15 and 2.2V -- the NB is at 1.55 iirc, and reads 1.53 in XP.



This shot quickly taken with the NB at probably 1.65 and reading 1.63, and a toasty 63C -- I cut that back right away, doh!



Here's the Everest screens. The "Unknown" is the Gigabyte heheh, the rest is obvious. Interesting to note the Gigabyte is strong on memory read, but very weak with Photoworx and one or two others. It's the Photoworx test that the Dark at stock speed beat out the overclocked Gigabyte.

[Lithotech]

I think you may be interested in this; a memory and BIOS Worksheet for the DFI P965-S Dark!

Updated June 24, 2007

Notes:

• I find this worksheet suitable for other motherboards and BIOS's, I simply use the first column to note the different applicable settings.
• I took the liberty of naming three of the new (to me at least, lol) functions with acronyms. None of these acronyms are commonly used by anything else that I know of, and make sense enough. I'll change them if they end up being referenced with something else.
• For tRFC, tWR and tRRD, the Infinity BIOS names them at REF to ACT, Write to PRE, and ACT to ACT respectively. Not terms I am familiar with, so I added the old more commonly known terms to the far right of each respective row, and greyed the text back to 50% so the end boxes in these rows are still useable.
• Feel free to distribute as you see fit, but please upload and link to your own webspace.

Link to PDF:

Clicky for PDF



Thumbnail to larger printable 300dpi JPG:




Small 72dpi JPG preview:




Enjoy! :D
.

[sswilson]

Thanks for the comments, and the worksheet (beats the hell out of the rough hand "drawn" one I was using... :) ).

Yeah, I noticed the "warranty only by manufacturer" as well and figured that was the last nail in the coffin for this board.

I'm guessing that they just don't get enough stock to be able to both sell it and deal with the squeeling issues.

Mine really isn't that bad, like I mentioned in the review I can hear it if I set up my system for lowest noise, but beyond that I don't notice it.

I'm getting another board to try out that will verify what the fsb limits of this cpu and mem are, but as it stands the best I'm able to get out of it stable is still pretty well what I listed in the review, and I haven't found the magic mix to get it to play nice with the mem set to 667 which would get me the best ratio for my cpu max.

[Eldonko]

sswilson, are you sure its not a mem issue causing reboot?

[sswilson]

This has happened with two different sets of mem, and like I mentioned.... at lower OC's it goes away with a slight vcore bump......

Still a lot of playing to be done with this board.... I just got my hands on a tech station and a used E6600 so I figure I'll rip my WC gear out of my case and transfer it all over to the tech station as a test bed and play from there..... :) (getting tired of opening the case to get at the c-mos reset button.... :) ).

Has anybody tried the beta bios yet? I know it's got a keyboard c-mos reset setting on it and wonder if it's worth trying...... :)

[Lithotech]

Originally Posted by sswilson

Still a lot of playing to be done with this board.... I just got my hands on a tech station and a used E6600.

Has anybody tried the beta bios yet? I know it's got a keyboard c-mos reset setting on it and wonder if it's worth trying...... :)

Wooo, nice feature to add to a bios update!

The playing never really ends, does it? I mean if you did actually get finished with testing and experimenting with every setting and combination, you just swap the cpu and/ram and start over lol.

One thing about DFI, it's always a nice bios to work with no matter what the model - genie ftw!

Pondering... considering the wall you are hitting, you have nothing to loose by trying new bio's, can always go back if it works out worse. Only real problem is with a new bios it's almost like a new cpu/ram combo, and to give it a chance you sorta have to start over.

Either way, it's fun. :D That's why we call is "play" and not work.

.