1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Universal abit IP35 PRO Review

Discussion in 'Hardware' started by Johnny Bravo, 22 Jun 2007.

  1. Johnny Bravo

    Johnny Bravo What's a Dremel?

    Joined:
    1 May 2006
    Posts:
    22
    Likes Received:
    0
    [​IMG]

    Intel has once again lifted the performance bar with the introduction of the P35 chipset bringing with it native support for 1333FSB based cpu's like Yorkfield and Penryn.

    Universal abit has released their gauntlet of P35 boards the IP35, IP35-E and the board being reviewed here, the IP35 PRO.
    The PRO as the name suggest is the top end of the P35 range of boards available from abit, tricked out with a number of nice features:


    • 100% Japanese solid state capacitors
    • uGuru Interface
    • Full RAID0/1/5/10 ACHI via Intel Matrix Storage Technology
    • Full heatpipe/heatsink coverage for PWM northbridge and southbridge
    • External CMOS clearing switch
    • Dual Gigabyte LAN
    • Onboard power, reset and debug code readout
    • HDMI header (for audio)
    • 2 x eSATA


    But enough of the details let’s get down to business:

    First Impressions

    [​IMG]

    The packaging is reminiscent of the AB9 series of motherboards and its last incarnation the QuadGT. The IP35-PRO and QuadGT share some similarities too. Looking at the two boards side by side is it evident that alot of the PCB placement has been carried over, one big exception is the PWM which had made a return to analogue (which will no doubt please some folk). The IP35 PRO has also gained a PCI slot and second LAN socket, albeit at the expense of an onboard mounted firewire connector.

    [​IMG]
    [​IMG]

    In the box we find the usual assortment of cables, I/O shield drivers and manuals. I was a bit perplexed to find an absence of a RAID disk, it would seem that you have to create your own from the driver CD - though this can be down from boot and won't require a previous installation of windows or PhD to get working.

    A closer look at the boards reveals a few other important chips:

    [​IMG]

    This is the "new" PWM controller chip the Richtek RT8802A which is also on the AB9 boards. The power delivery is a 4 phase design, interesting as the AB9 was a 5 phase design.

    [​IMG]

    This chip is the Analog Devices AD8402 2-Channel digital potentiometer, those on you with a keen eye will be able to see that is it situated near a bank of resistors at the cpu socket. It's purpose is to allow "on the fly" adjustment of GTL level voltages which can significantly improve FSB overclocking especially on Quadcore cpus. For more reading on this I recommend reading this article by Freecableguy over at thetechrepository.

    [​IMG]
    [​IMG]
    [​IMG]

    Other areas of the board are the same as the QuadGT - the MCH (northbridge) is fed via a PWM power circuit as is the DDR2 while the southbridge is a linear design. VTT is supplied via the MCH power supply and in the BIOS you will find that the available VTT will rise with increased MCH voltage.

    One final touch that is lost on many other competitors’ boards is the fan headers on this and other recent abit boards are supplied via power MOSFETs. This means that you can directly attach high power fans like deltas directly to the motherboard without fear of blowing the header, very handy indeed especially if your doing water cooling - you would be able to control not only the fans but the pump as well via uGuru, a nice touch indeed!

    BIOS

    [​IMG]

    uGuru reappears as the main hub of overclocking on the board. While their are those who aren’t too keen on the idea that only uGuru is able to adjust clockspeeds ( I myself have found it troublesome especially on the QuadGT) Looking at the program as a whole it is an amazing piece of software. You are given full control over and monitor all voltages, fan headers and clockspeeds all from one program - with the ability to create profiles accordingly. In the BIOS we find a good spread of available voltages as shown below. The memory divider has proven to be a bit confusing for some so here's a table of the various dividers and straps:

    [​IMG]

    This is from another forum so I've yet to confirm it all. You will also notice the addition of the GTL voltage levels given in %. 67% seems to be quite good for most people so is a good starting place, don't make rash adjustments here as you can quickly loose stability at higher FSB speeds.

    [​IMG]

    Also an interesting point to note is that the 12v 4/8pin connector for the PWM and the 12v line for the 24pin connector are now independently monitored in uGuru, useful if you want to see if your PSU supply for the cpu is drooping badly under load.

    Moving onto the chipset configuration it should be noted that the CAS has been limited to 4, 5 and 6. As of yet no reason has been given as to why CAS3 isn't available but apparently it is in the works for the next BIOS update. Also abit have opened up the sub timings with this board which is a Godsend for all the memory nuts out there.


    System Setup

    CPU - x6800
    HSF - Thermalright ultra-120 Extreme and Intel stock
    Memory - Cellshock PC2-8000 4-4-4
    VGA - Radeon 7000 PCI and Nvidia 8800GTX
    Raptor 36GB Samsung DVD-combo PS/2 mouse keyboard

    [​IMG]

    Initial Observations

    Voltage readings (as always) are a little off from those reported via BIOS/uGuru the table below is a guide only and may be representative for your chosen cpu voltage levels.

    [​IMG]

    Notice that the droop under load for an x6800 is 0.005volts, impressive enough. The MCH seems to be more strained under load swinging up by 0.012volts which has the knock on effect of swinging the VTT voltage - 0.017volts. VDDR is fairly stable dropping only 0.02 volts under load.

    The PWM is also quite cool running with temperatures rarely topping 50oC under all testing. This is without any directed active cooling as the fan for the Thermaltake ultra 120 wasn't not direct path of the PWM heatsink. More surprising was the coolness the MCH heatsink ran at, even under higher voltages. My initial fear was a poor thermal contact but even after applying paste in lieu of the thermal pad the temperatures were still in the 30s-40s.

    The onboard IDE and eSATA is provided via the JMicron JMB363 chipset. This has gotten its fair share of flak over the last year, some of it well deserved. I found with my testing both using an IDE drive and a CD-combo drive that it was fully functional and detected the drives correctly. HDTach showed that the IDE drive had sub performance compared to a native connection but this is to be expected, as I doubt anyone will be running their OS off this chipset. CD drive performance was roughly on par with that of a native connection. The eSATA performance was not checked due to lack of compatible hardware.

    However not everything was sunshine and roses. The CD-combo drive that I normally use has a PATA interface, and with the advent of none IDE spec’ed boards I decided to purchase a PATA-2-SATA adaptor to allow it to connect via SATA. This has never proven to be a problem until now. On multiple occasions the computer would hang at the "scanning for SATA devices" and it quickly became apparent that the CD-Combo drive was to blame. I've yet to confirm this problem is widespread or an isolated case, but it seems to be eased somewhat by connecting it to the odd numbered ports (I currently run HDD on SATA1 and CD-combo on SATA3). However be aware that "high" FSB speeds (500MHz+) seem to increase the chances of it occurring as well.

    Another oddity that I came across was a curious memory incompatibility. I have used two different sets of RAM when testing this board, my old trusty Crucial 10th anniversary sticks (Micron D9GMH) and my newer Cellshock PC2-8000 4-4-4 sticks (Micron D9GKX). Strangely the Cellshocks don't seem to be quite at home with this board. Rated to run at 500MHz 4-4-4-12, the sticks would hang at the post at these rated speeds. Conversely the Crucial sticks had no problem and run stably. I will have to wait until others with Cellshock memory and this board can run comparative tests to see if it is just my memory or a problem with the board.

    Results

    No doubt many of you will have read other reviews of this board and seem impressive FSB speeds reached with 550FSB+ being achieved. Unfortunately I cannot boast the same sort of results as it would appear the x6800 I have is somewhat FSB limited, peaking at 510MHz. Putting it under phase improved this score somewhat to 530MHz but nothing noteworthy. I hope to get a hold of another cpu to try again and maybe pull some bigger numbers.

    Firstly the maximum FSB for this cpu on air:
    [​IMG]
    LINK

    Dropping the FSB down a little to 500MHz and run a few memory benchies. Firstly some stability at 500MHz 4-4-4-12 timings:

    [​IMG]

    Now some stability testing with 4BG of RAM. This is kind of inaccurate as XP can only see 3GB of the memory but the IP35 PRO is able to handle it with ease. It would seem that the memory controller is quite robust on the P35:

    [​IMG]

    Stepping things up a notch now we run the memory (Cellshock) at the tightest timings allowed in the BIOS. I've included both memory benchmarks and stability testing:

    [​IMG]
    [​IMG]

    Changing the memory dividers we are able to ratchet up to 400:600 fully stable:

    [​IMG]

    Strapping in the 8800GTX and running on the tightest timings I was able to get a 3DMark01 score of 66444, not too shabby for a card running at stock:

    [​IMG]

    Having spent little over 5 days with this board I must say I've been impressed with its stability. I was always annoyed by the QuadGTs inability to apply saved BIOS profiles correctly regularly hanging on reboot. The IP35 PRO has yet to suffer the same fate, and is remarkably resilient when overclocked too far. Usually simply turning the power off and back on will allow the board to load the previous "good" configuration.

    Apart from the problems with the disappearing CD drive and the Cellshock incompatibility I have been impressed with the performance stability and features of the board. Over the following weeks I hope to delve into the heart of the board a little more and see how to really get the most from it. Expect vmods to follow shortly ;)

    John
     
  2. Highland3r

    Highland3r Minimodder

    Joined:
    25 Jul 2003
    Posts:
    7,558
    Likes Received:
    16
    Nice review mate :D

    See you shelled out for the cellshock in the end then? :D
     
  3. Guest-16

    Guest-16 Guest

    Awesome review, John, and very informative. Looking for a job? :D

    Never had a problem with the JMB363 chipset before, and we never really test the performance from optical drives. If they work and burn at the speed that should that's generally good enough. It's only performance nazi's that care if it's 3mb/s slower copying the contents of a dvd to a hard drive or something. It's always booted the xp or vista cd just fine from a range of drives we use. Not all PATA to SATA adapters are made the same either, some never used to work properly even when the IDE port was natively included. I'd thoroughly recommend a plextor SATA optical drive any day of the week though, it's a fantastic investment.

    I talked to Abit at computex and they are labelling the clear cmos button on the rear of the case, and they made a good point that having it on the rear means you dont have to open the case to get to it, but in some cases I'd rather do that than have to fiddle around over the top and at the back of the case all the time. They need another uGuru panel, perhaps something 3.5" but one that looks good with it on the front.

    If people want performance they'd use SATA drives to have a 1:1 native connection not a pipe through the pci-e bus on a shared IDE channel :)

    As for floppies, vista doesnt need them and very few people will need them for xp that dont already have a pc elsewhere. Plus floppies cost money and i'd rather have all the pci brackets and sata cables any day of the week. When will people include S/PDIF brackets ffs? Why aren't there 6 USB ports on the rear?

    How did you record the voltages? And your VDDR goes up a bit, not down ;)
    Having a GTLRef is nice, and it's far better than DFI's "number" on the ICFX3200T2RG, but I'd still like a value or something. Maybe with an EFI BIOS we'd get a graphic to help understand where best to position it. I'm assuming the percentage is a percentage of the actual voltage through it so it knows where to look for a signal?

    The whole board uses solid state caps, which is good, but it uses non-ferrite chocks unlike the Gigabyte and if marketing material is to be believed, that's better. The MCH has a nice big heatsink on it and it's no where near as power hungry as the nforce 6-series or pending x38 which is some 35w.

    I dont see why people would want analouge pwms too, unless the current generation of digital suffers from vdroop, but i've always preferred digital from the accuracy of application and tidyness on the board.

    Taking pictures: from experience take them out the packets and put them on white sheets of paper ;)

    As you have an AB9 Pro there, and I assume QuadGT too, take pics of the boards side by side to make it easier to understand :)

    We've not got CPUs that'll do >500FSB. The most we've had is like 495ish on a Striker from our X6800s. We've got a 1333FSB CPU that should do more and when I come to look at this board in a few weeks we'll see how far it goes. It's very hard to find something that'll do all day everyday stable over 500 on air anyway.

    As for memory timings: 49.5ns is quality! I'd much rather have 1:1 at low latency than pump it up to 600 and drop it down to 5s. The Intel chipsets have exceptional memory performance if you keep it all tight nit.
    What test are you using on Everest exactly and is that the trial version of ultimate?
     
    Last edited by a moderator: 22 Jun 2007
  4. Guest-23315

    Guest-23315 Guest

    I'm impressed with that board.

    Some nice FSB's there, and some nice Heatpipes aswell!
     
  5. BUFF

    BUFF What's a Dremel?

    Joined:
    25 Mar 2005
    Posts:
    912
    Likes Received:
    1
    Perhaps (probably) P35 is like 965 where CAS3 seems to give worse results than CAS4?
    Nice overview anyway. :blah:

    As I'm sure that you know there is another panel imminent but again it's 5.25".

    Previous iterations (e.g. DFIs & QuadGT) tend to run very hot & also quite a few complain about them being noisy. There is another generation of digital PWMs though which are supposed to improve things.
     
  6. Neogumbercules

    Neogumbercules What's a Dremel?

    Joined:
    14 Aug 2004
    Posts:
    2,464
    Likes Received:
    29
    Pretty awesome review for a forum post.
     
    Last edited: 23 Jun 2007
  7. Johnny Bravo

    Johnny Bravo What's a Dremel?

    Joined:
    1 May 2006
    Posts:
    22
    Likes Received:
    0
    Whoops should adjust that statement. Voltages were recorded via a DMM on specific voltage readpoints that are as close to the relative component. AFAIK the percentage in the BIOS is indeed a relative percentage of the GTL voltage.

    There was a problem many had experienced where the digital PWM ICs generated a lot of heat. This was due to two things
    a) The temperature is reported more accurately (and consequently higher)
    b) The thermal contact to the ICs was less than stellar. Even doing something as simple as folding the thermal pad in two dropped the temperatures considerably. In the end people just got fed up and said they didn’t like it.

    Care to post it out to me :D

    Ultimate edition, just running the memory benchmark program.
     
  8. Guest-16

    Guest-16 Guest

    The next gens used on the DFI and Abit MAX boards piss away a ton less heat and are meant to be loads better according to techies.
     

Share This Page