Gigabyte GA-P35-DS3P

[phoronix]

Phoronix: Gigabyte GA-P35-DS3P

Gigabyte has long been a very big name in computer hardware. They manufacture almost any sort of hardware you can imagine. High quality computer cases, motherboards, video cards, laptops, CPU coolers, you name it they have it (they even have a cell phone!). Like ASUS, they are able to diversify and cover a massive segment of the industry, and they are able to do this without sacrificing quality and performance. An impressive feat indeed. Not so long ago, Gigabyte made a massive splash in the enthusiast/overclocking world with their release of the board known as the DS3. The 965P-DS3 was one of the best overclocking motherboards ever to be released on the market. Not only was it brilliantly designed, but very importantly, it was brilliantly priced. Everyone could afford it because it was not only better performing, but also cheaper than the competition. Overclocking was BY FAR the easiest we have ever encountered in all our years of experience. This is also partially because almost all of the Core 2 Duos are simply beasts. The P35-DS3P that we will be looking at today carries the same DS3 mark. This board is really the successor to the incredible 965P-DS3 and should be held to the same standards of quality and performance.

http://www.phoronix.com/vr.php?view=11001

[Kano]

I have got a similar board (GA-G33-DS3R), but never had issues with network in dual boot, maybe your WIN driver was too old? It does not like to boot from grub with USB cdrom, isolinux boots. One thing with OC, when you for example would lower the multi then WIN shows correct speed, but when you use LINUX then you see a speed which would be shown when the original multi was used. You could try that if you like to.

[Panix]

I have got a similar board (GA-G33-DS3R), but never had issues with network in dual boot, maybe your WIN driver was too old? It does not like to boot from grub with USB cdrom, isolinux boots. One thing with OC, when you for example would lower the multi then WIN shows correct speed, but when you use LINUX then you see a speed which would be shown when the original multi was used. You could try that if you like to.

I'm looking at getting a GA-G35-DS3R board and was glad to read a review on the GA-G35-DS3P which should be similar in setup and application regarding Linux.

Actually, there is a thread in the Ubuntu forums devoted to the network issue with the GA-G35-DS3R/P boards. It seems it's universal but a minor thing. More of an inconvenience if anything. It sounds like a decent board to use with Linux which is good because I've been considering the board for a while because of the price. I only need a PSU and GPU besides the MB.

[Kano]

G35 can only combined with ICH8(R) instead of ICH9(R), but the difference is rather small. You need a Linux distribution with kernel 2.6.22+ and relative new Xorg. Ubuntu Gutsy would work out of the box, but if you want something Debian Etch based I could help you too with a Xorg backport.

[Panix]

G35 can only combined with ICH8(R) instead of ICH9(R), but the difference is rather small. You need a Linux distribution with kernel 2.6.22+ and relative new Xorg. Ubuntu Gutsy would work out of the box, but if you want something Debian Etch based I could help you too with a Xorg backport.
I meant I'm considering a GA-P35-DS3R board, sorry!

Does that mean it's more likely to be supported out-of-the-box? I can use debian-based distros like Ubuntu/Kubuntu/Etch etc. or Fedora? Maybe kernel .20+?

[Kano]

ICH9R alone is no problem with current kernels. But the G33/G35 would be interesting too if you want to try the intel onboard drivers. Usually the price is the same, just firewire missing and the serial port needs a slot bracket.

[Panix]

Is there any issue with sound with these motherboards? I went to the Realtek site and although Linux drivers are listed, none of the Realtek (audio) chipsets/codecs listed there are ALC889 (Gigabyte P35 boards usually have that one). ALC888 is listed so an Abit IP35 Pro mb (has the ALC888) probably wouldn't have an issue?

[Michael]

There were no sound issues with the GA-P35-DS3P during testing, it worked out well.

[Panix]

Thanks for the quick confirmation.

I read someplace that Linux (well, one distro is claimed to, anyway) reads the ALC889A codec as ALC885 but the concern there is an issue is alleviated to discover the sound works without additional tweaking. Thanks.

[dj_segfault]

It seems this board is out of stock at almost every vendor. For instance, newegg is estimating they'll have them 10/01. Do you think they're transitioning to rev 2.0? Or is there something wrong with this board that they're recalling them?

Thanks.

[ivanovic]

Short questions because I want to buy a board of the same series, just the "plain" version with only one pcie x16 slot (either the DS3 or the DS3R):

1) Does IDE on this board work out of the box or are there some problems with it? I would have to use IDE for an optical drive and a hd.

2) What about sensors support? Does lm-sensors support this board already with all sensors or what is the current status?


BTW: I would really welcome a test with a vanilla kernel in those mainboard reviews. I have no idea how many patches and such are applied to the kernels shipped with some distributions and personally I prefer to at least have the option to switch to a vanialla kernel. In such it would be nice to have a test if everything works nicely with vanialla and attaching a config file that does support all the stuff, too.

[Kano]

jmircon is supported in current vanilla kernels. CONFIG_PATA_JMICRON option.

[]--Freeman--[]

Hi,I have this motherboard (GA x38-DQ6) and I have a question about the raid. Does Ubuntu Gutsy support ICH9R chipset? I would like to set raid 5 but I don't know if it's possible in Ubuntu Gutsy. If it's possible, whick program I have to use so as to install ubuntu gutsy in a raid array? Is there some guide?


Thanks in advance!

[Kano]

Raid 5 should be possible with Kanotix Thorhammer, if you try that be sure to click on the IRC icon and ask for instructions, dmraid installs are possible but for example you can not use gparted to partition it. Also best to use some other tricks like GRUB4DOS when you want to do a dual boot system on raid (and install bootloader into partition). No Ubuntu kernel has ever had the needed dmraid 45 patch and the included dmraid of course would not load the extra module even if present. U is in that way pretty stupid ;)

[Syzygies]

I bought a GA-P35-DS3P for my second build, in no small part because of this
review (the OSx86 community also loves this board), so this seems a handy
place to put some initial build gotchas.

[1] Only a PS/2 keyboard can be used to access the BIOS, to turn on USB
keyboards and mice. Then, a USB keyboard can be used e.g. with the Ubuntu
Server install disk. Otherwise, one determines that one's USB isn't fried by
using the Ubuntu Desktop install disk, which defaults to booting to live disk
if it doesn't hear from you. One may need to keep the PS/2 keyboard handy for
future BIOS work, as it is picky about USB keyboards even after changing this
setting.

[2] My Corsair TWIN2X4096-6400C4DHX 2 x 2 GB memory (2.1v 800 4-4-4-12, but
800 5-5-5-18 in SPD) failed memtest+ (don't use the copy on a Ubuntu disk, get
the latest from www.memtest.org) with my f1 BIOS. The latest BIOS is now f2.
The download is a Windows .exe, so unpacking it gave me my first productive
use ever of my virtual Windows on my Mac. I couldn't update from a USB flash
drive, contrary to the documentation. Formatting an internal Sata drive to a
single FAT32 partition, and putting the unpacked BIOS files on it, did the
trick nicely and very swiftly. I had far less trouble with my memory in the f2
BIOS at default settings, 1 error in an hour rather than hundreds. I then
manually set 2.1V, 4-4-4-12, and as I write I'm an hour into testing with no
faults.

TIP: When you take your shiny new SATA drive (Raptor, anyone?) out of its OEM
pouch, DO NOT PASS GO, DO NOT COLLECT $200. Immediately attach it to an
already working machine, format it FAT32, and copy on the updated BIOS files.
NOW install the drive in your build. Flash the latest BIOS as soon as you can,
to save hours of your life, and while the mobo is still returnable. It will go
fine.

(In my case, the memory instability crashed my first install, leaving the hard
disk in an unsettled state, causing the BIOS to hang on reboots until the
drive was disconnected. So removing it to give it a clean temporary format
killed two birds with one stone.)

If this post saved you a few hours of your life, allow me to rant a moment. I
haven't checked the language rules for this forum, so you'll just have to
imagine a calm, exceedingly tenacious guy turned so furious spewing
obscenities that his brains are hemorrhaging out his eyeballs.

What gives with [1] and [2]? Perhaps in Taiwan a Windows-centric attitude
isn't seen as the equivalent of a whites-only drinking fountain, but this has
me steamed.

The big picture is what really bothers me here. When a GB of flash memory goes
on sale for $8, which could hold THREE decent Linux distributions, where does
this industry get off with 1980's BIOS code that's in the MS-DOS reading
group?

I mean, there's room for code that reasons
[I]
"Huh. We've got USB devices disabled, but it's been MINUTES, and Marge, I
haven't heard a peep out of that PS/2 port. Oh yeah, the hard disk seems
corrupt, should I just freeze like I'm as stupid as an Adirondack deer? Oh
yeah, those USB ports keep getting insert events like someone's going around
trying all of them, thinking he blew the wiring. Should we turn one on as a
courtesy? Keep moving it, keep him guessing. Oh! Oh! Marge... Ya know that
stealth OSx86-ready micro-iSight next to the mobo LEDs? The guys got the cover
off, see that case intrusion sensor, and he's LOOKING at us. Oh Marge,
jeesh!!! he's got a sledge hammer and he looks mad. He's raising it.. Marge,
turn on the USB! turn on the USB!"

I expect nothing less than BIOS code like this. Why are we all lemmings that
accept less?

[Kano]

Usually you can flash from usb stick pretty easy as well. Just extract the files, you can do that with linux using "unrar x motherboard_bios*exe".

[Syzygies]

Usually you can flash from usb stick pretty easy as well.
Just extract the files, you can do that with linux using "unrar x
motherboard_bios*exe".

Thanks for the Linux extract pointer.

Any idea why I couldn't flash from a USB stick? Is there a BIOS setting?

On the other hand, this BIOS is fussy about USB keyboards, it may be fussy
about USB sticks. Is _IS_ a great mobo once one gets past all this, I'd
buy it again.

Given how there are posts all over the web of people who take the warnings
seriously on not risking a BIOS flash, while they suffer the consequences of
an old BIOS (are we all afraid to pull those tags off of pillows, too?), and
how flaky USB sticks can be in the best of times, I stick to my
recommendation:

A part of the standard build procedure should be to put the latest BIOS on the
first hard disk before assembling the hard disk into the case.

[My memory went 10 hours with no errors in Memtest+, after manually setting
+0.3V (from 1.8V to 2.1V) and 4-4-4-12. There was only a single error with the
F2 BIOS and auto-detect. From my experience I'd say to manually set the memory
no matter what, and to be sure to run Memtest+ overnight.)

[Kano]

Maybe your usb drive is in super floppy mode, thats not supported, you need partitions on it then it works fine.

[Syzygies]

After various issues I swapped out a Corsair TWIN2X4096-6400C4DHX 4 GB kit for

OCZ DDR2 PC2-6400 Reaper CL4 4GB Edition
http://www.ocztechnology.com/products/memory/ocz_ddr2_pc2_6400_reaper_cl4_4gb_edition

which is working flawlessly at rated 2.1V, 4-4-4-15 settings; I'm in the midst
of a multi-day mprime (prime95 for Linux) torture test. I do plan to
overclock, but I'll relax the memory as necessary so scientific calculations
can go months without errors; this is my need.

In the end, Corsair tech support was great, and their memory was probably
fine, but I like the potential for the new memory, so I took an opportunity
to swap.

What I uncovered as the probable explanation for my problems is an issue
anyone using this board should understand, whether or not they ever plan to
get into overclocking: Like it or not, using performance memory with this
board requires manually changing BIOS settings as if one is an overclocker.

And here's the rub: One can run afoul of Gigabyte's "helpful" and poorly
documented automatic BIOS settings. Specifically, "Performance Enhance"
defaults to Turbo, not Standard, going to manual memory timings does not
by itself disable this feature, and the shaky English makes it sound like
only a yahoo would settle for Standard. In fact, one must explicitly
disengage Turbo, according to Gigabyte phone tech support, and this is
documented nowhere on Gigabyte's site or in the manual.

Some good explanations of what Turbo is mucking with can be found on these
pages; the last article is an awesome read:

http://www.clunk.org.uk/forums/5254-post75.html
http://www.clunk.org.uk/forums/6154-post134.html
http://enthusiast.hardocp.com/article.html?art=MTM5NSwyLCxoZW50aHVzaWFzdA==
http://www.anandtech.com/mb/showdoc.aspx?i=3208&p=7

There's a theme here, if one reads criticisms of this board. One has to
simply know what various settings are before one changes them, e.g. +0.3V
is pretty obviously setting the memory voltage to 2.1V if one already knows
that the standard is 1.8V. Anyone buying memory should have picked this up,
but the list goes on...

In short, if one has any troubles with this board, a proper course for
navel-staring is to contemplate each and every BIOS setting in turn. The
particular issue to watch out for is automatic performance enhancements that
might be guessing wrong or interfering with manual changes you're required
or want to make.

This is literally how I resolved my issues. I was trying to relax the
memory in any way possible, and downgrading "Performance Enhance" from its
Turbo default (GA-EP35-DS3P F2 BIOS) to Standard did the trick. Only then
did I research what I had just done. The English suggesting I was settling
for performance only a loser would accept didn't stop me, because I was
trying to slow things down as a test.

I nevertheless love this motherboard. Ultra Durable speaks for itself, I
wanted this grade of parts. (I became disenchanted with making headphone amps
after realizing all they'd do for me is make noise, but I remain fussy about
parts.) And the BIOS is extremely configurable for experts. My Q6600 at full
load draws less power and runs cooler than my previous Q6600 build on idle (!)
using an Intel "Bad Axe II" motherboard (there are many factors in this, but
this motherboard is the biggest one). The Intel mobo has actually given me
zero trouble, but it's unexciting, like an old Chevy.

[Kano]

Well the Turbo default is not so good for manual OC, but what I dislike more in the BIOS is that it always wants to OC the ram when raising the fsb. Of course it is possible to control that manually too, just the auto mode is only good for high priced OC ram.

[Syzygies]

My build is the following:

Gigabyte GA-EP35-DS3P motherboard
Intel Core 2 Quad Q6600 Kentsfield 2.4Ghz Processor (SLACR, G0 stepping)
OCZ Reaper HPC Edition 4GB (2 x 2GB) 240-Pin DDR2 SDRAM DDR2 800 Memory
Antec P182 Gun Metal Black ATX Mid Tower Computer Case
Thermalright Ultra-120 eXtreme CPU Cooler
5 x Scythe S-FLEX SFF21F 120mm Case Fans
CORSAIR CMPSU-620HX ATX12V v2.2 and EPS12V 2.91 620W Power Supply (modular)
Western Digital Raptor WD1500ADFD 150GB 10,000 RPM SATA 1.5Gb/s Hard Drive
LITE-ON 20X DVD±R DVD Burner with LightScribe Black SATA Model LH-20A1L-06

Finally, the other shoe dropped, and overclocking went from mysterious to
obvious. I'm running 3.00 Ghz cpu, 1066 Mhz 5-5-5-15 memory, all at stock
voltages except the "stock" increase to 2.1V for the memory, which can take
2.2V without voiding the lifetime warranty.

My core temps stay below 60C, and that's with the cpu cooler fan on
motherboard speed control; one could attach two fans in a push-pull
configuration running full tilt, to get further on air.

I view this as a very conservative build, what one might do if one "isn't"
overclocking (!). I plan to leave this on 24/7, often at full load doing math
computations. I had to start raising the CPU voltage slightly at 3.30 Ghz, and
I gave up as the core temps crossed 60C, with no end in sight. People have
easily reached 3.6 Ghz on air with rigs like this.

Selecting parts for this build involved overcoming pet prejudices, such as
"There's Corsair, then there's everyone else" for memory, or "Ooh! Ooh!
Doesn't the copper Ninja look like the best heat sink ever? Object lust!"
Saying this on any forum seems to be bait to get people to chime in on their
pet prejudices. I'm only willing to believe tough comparative reviews, and
there are many out there. Before overclocking, this rig ran cooler at full
load than my previous Q6600 rig runs at idle. That's how one gets a margin for
overclocking.

This passes stress tests that are harder than 24 hours of mprime (Prime95 for
Linux). My favorite is to run 400 builds from source of the GHC Haskell
compiler, 4 or 8 builds at a time, which takes days. A friend tried this with
a $10K workstation he had built for him (8 cores, 64 GB ram), and apparently
the builder had guessed the power supply needs seat-of-the-pants. My stress
test blew his power supply, literally curling smoke into the room. So this is
a "real world" stress test. My rig passes this test.

The short course for conservatively overclocking this board:

=> Fix the PCI Express Freq at stock, 100 Mhz
=> Increment the CPU Host Freq to raise the CPU clock 0.1 Ghz at a time
=> Find a Sys Mem divider that keeps the Mem Freq <= 800 Mhz

... and nudge one's way up, stress-testing the setup until voltage increases
are needed for stability, then back off. Now relax the memory timings, and
separately see how high one can push the memory clock, by changing the memory
mulitplier.

The overriding principle is to disable any automatic help that the BIOS
offers. Nothing I tried with its help worked, but manually overclocking this
board is a piece of cake, even for a newbie like me.

Note that I'm "forced" to manually control all voltages, because the automatic
setting won't give my memory the needed boost. This is ok, increasing voltages
is a slippery slope, where components get hotter, needing more volts, which
makes them hotter... For conservative overclocking, see what you can do with
stock voltages.

Here are my BIOS changes:

CPU Frequency (Ghz) 3.00
CPU Host Clock Control Manual
CPU Host Frequency (Mhz) 333
PCI Express Frequency (Mhz) 100
Performance Enhance Standard
System Memory Multiplier (SPD) 3.2
Memory Frequency (Mhz) 1066
DRAM Timing Selectable (SPD) Manual
CAS Latency Time 5
DRAM RAS# to CAS# Delay 5
DRAM RAS# Precharge 5
Precharge delay (tRAS) 15
System Voltage Control Manual
DDR2 Overvoltage Control (+V) 0.3
CPU Voltage Control Normal
Normal CPU Vcore (V) 1.275