Sign in with
Sign up | Sign in
Your question
Solved

Overclocking the Gigabyte GA-P55-UD3R...excellent!

Last response: in Overclocking
Share
September 28, 2009 7:06:19 AM

AMD had a beautiful parade a couple years ago, we called it K8. In virtually every conceivable comparison, it would beat the Intel P4 chips into the ground. Those days were sweet for AMD, but out of nowhere Intel began a surprise attack on AMD dropping the Conroe bomb, then the Penryn bomb, then the Nehalem bomb, and now the Lynnfield bomb...or is it more of a "miss"?

For years AMD had been the enthusiast CPU of choice. AMD was not as openly opposed to overclocking as Intel had been, and for good reason, their chips OC'ed relatively well and didn't create as much heat as Intel's. Since Intel released Conroe into the wild a little over three years ago...the enthusiast landscape has shifted dramatically. Intel's newest architecture was an enthusiast's delight. And since then, Intel has been more openly embracing the overclocking community.

So, I'm trying to figure out where Lynnfield fits in. Nehalem already holds the performance crown, AMD doesn't have anything that can compete for performance. So, AMD has had to compete with pricing, dropping the prices of it highest performing CPUs to be even slightly competitive from a price/performance standpoint. This leaves Intel in a bit of a predicament at the low end, because they have been forced to compete against AMDs top CPUs with older hardware. Enter Lynnfield, a CPU designed to compete at a lower price point, yet still have the performance to beat AMD's best.

So, I'm here today to talk about Lynnfield, right? And I'm an extreme overclocker, so I'll be using an i7 870 to test the platform right? And I'll be using one of the most expensive motherboards on the market, because it will give me the most potential for overclocking, right? Oh, and let's not forget the LN2....right? No, no, no, and....no!

Gigabyte Japan asked me to look at their entry level board, the GA-P55-UD3R. At first I was hesitant...what am I gonna do with an entry level board like the UD3R? I'd probably kill it within the first 5 minutes right? So, I decided to do one for the people...the gamers...you! So, the first part of this review will be for the GA-P55-UD3R with an i5 750, and WITH AIR COOLING ONLY! In this case I'm using a Prolimatech Megahalems cooler with the new LGA1156 compatible bracket.

First off, the board:

Model
Brand - GIGABYTE
Model - GA-P55-UD3R



Features

  • Features Ultra Durable 3 technology with 2oz copper PCB design
  • Innovative Smart 6 technology for Smarter PC management
  • New Dynamic Energy Saver 2 technology enables best energy efficiency
  • Support ATI CrossFireX for ultimate graphics performance
  • Features high speed Gigabit Ethernet connection
  • XHD technology accelerating hard drive performance with ease
  • AutoGreen technology Greening your PC via Bluetooth cellphone
  • Patented GIGABYTE DualBIOS technology delivering highest level failure protection
  • Home theater quality 8-channel High Definition Audio
  • Compatible with Windows 7 to deliver the best operation experience






    Physical Spec
    Form Factor - ATX
    Dimensions - 12.0" x 9.6"
    24 Pin Power connection



    Rear Panel Ports
    PS/2 1
    10 x USB 2.0
    2 x eSATA 3Gb/s
    S/PDIF Out - 1 x Optical, 1 x Coaxial
    6 x Audio Ports



    Onboard Audio
    Audio Chipset - Realtek ALC888
    Audio Channels - 8 Channels

    Onboard LAN
    LAN Chipset - Realtek 8111D
    Max LAN Speed - 10/100/1000Mbps



    Supported CPU
    CPU Socket Type - LGA 1156
    CPU Type - Core i7/i5



    Chipsets
    North Bridge - Intel P55

    Storage Devices
    PATA - 1 x ATA133 2 Dev. Max
    SATA 3Gb/s - 8
    SATA RAID 0/1/5/10

    Onboard USB
    Onboard USB - 4 x USB 2.0



    Memory
    Number of Memory Slots - 4×240pin
    Memory Standard - DDR3 2200/1333/1066/800
    Maximum Memory Supported - 16GB
    Dual Channel Supported



    Expansion Slots
    PCI Express 2.0 x16 - 1
    PCI Express x16 - 1 (@x4 bandwidth)
    PCI Express x1 - 1
    PCI Slots - 4

    September 28, 2009 7:08:17 AM

    So, my first step was to run it through a basic battery of CPU benchmarks at "stock speed" to get some baseline measurements.

    Super PI 1M
    This program calculates the digits of the number PI out to 1 million digits, it is single threaded.



    Super PI 32M
    This program calculates the digits of the number PI out to 32 million digits, it is single threaded.



    PiFast
    This program that crunches through the digits of the number PI, it is single threaded.



    wPrime
    This program calculates prime numbers. It allows for two tests, to calculate the first 32 million, or the first 1.024 billion prime numbers. This program is great for multi-core and/or multi-processor systems, and scales very well with each additional core.



    With the new Lynnfiels CPUs, "stock speed" has different meanings in different tests. The "stock" multiplier of this i5 750 CPU is 20x, but in multithreaded benchmarks it runs at 21x, and when I ran Super PI or PiFast, it ran at an impressive 24x!!! That took the "stock" speed of 2660MHz to either 2793MHz or 3192MHz depending on the test...wow! 20x is only used when the CPU determines that the load is too heavy and thermal throttling kicks in to prevent the use of the higher multipliers.

    You've probably noticed that this CPU is very similar to the i7 920, they both have the same "stock speed" of 2660MHz, they both have the same amount of L1, L2, and L3 cache, and they both have 4 physical cores. But there are some important differences as well. The 920 has a triple channel memory controller, the 750 only dual channel. The 920 uses Hyper Threading (HT) to enable the operatining system to treat the CPU is if it has 8 cores. On the other hand, the 920's turbo mode has a maximum multiplier of 21x and is therefore more limited in it's overall clock speed for single threaded applications, this would help the 750 make up for some of the performance lost with the lack of triple channel memory.

    The final and most important difference between the two CPUs is the price. At only about $200, the i5 750 is significantly less expensive than it's i7 bretheren (about $280). Not only the CPU but the entire platform is designed to fill a lower price bracket. So, while an entry level 1366 board, like the GA-EX58-UD3R runs about $200, the P55 varient shown here sells for only $140. Also, you must factor in a set of DDR3 memory, since the Lynnfield arcitecture only supports 2 channels of memory, you could get a 2x2GB kit for as little as about $70, whereas those who chose to get a platform based around Nehalem would need a 6GB kit to be competative, which would drive the price up to about $110 or more. Altogether, a new setup based on Lynnfield (using the price estimates shown above) would only cost about $410 whereas the i7 rig would set you back about $590, or an increase of about 44%. So the question is, does the i7 platform perform 44% higher?

    Since my main goal these days is to push HW to the brink with extreme cooling methods with high scores being the only end in sight, I wasn't sure how to proceed with this more entry level HW. So I decided to review this motherboard to bennefit the target audience...the budget friendy crowd looking to maximize performance/cost...hopefully that means YOU! As I stated before, I will only be using air cooling for this portion of my review, and I won't be using any LGA1156 i7 CPUs either.

    For this portion of the testing, I wanted to find the very best performance I could with air cooling, that means using a powerful fan, in this case a 252CFM Delta strapped to the side of the Megahalems. I tweaked the OS a bit for each bench, and adjusted the voltages to whatever was needed for the highest clocks for each test. Using the same tests used with the stock speed testing, here are the results...

    Super PI 1M
    9th fastest i5 750 on HWBot (1st with air cooling)



    Super PI 32M
    3rd fastest i5 750 on HWBot (1st with air cooling)



    PiFast
    3rd fastest i5 750 on HWBot (1st with air cooling)



    wPrime 32M
    4th fastest i5 750 on HWBot (1st with air cooling)



    wPrime 1024M
    3rd fastest i5 750 on HWBot (1st with air cooling)



    Next up, I've got some 24/7 settings for you everyday overclockers...later, I'll be putting this thing under some LN2
    m
    0
    l
    September 28, 2009 7:08:50 AM

    Right off the bat, I was impressed with the ability of this board and this CPU to attain stable overclocks with low voltage. With a stock voltage of 1.1V, it doesn't seem like there is a whole lot to work with. I've pushed 45nm CPUs to 2.0V and beyond with liquid nitrogen, so limiting myself to only 1.1V seemed really silly. But I wanted to see the potential for this chip with absolutely zero voltage increases. Here is a shot of the BIOS voltage page for the following run:



    The result? 3.45GHz 14 hours Prime95 stable sounds pretty good to me with only 1.1V.



    * temps only reached 60C due to a high ambient temp of 32-34C during this run.

    So the next step was to see what I could get with a bit extra voltage. I wanted to give you an idea of what this setup is capable of with decent air cooling for a 24/7 based setup. With that in mind, made the following changes in the BIOS:



    The result? I was able to reach 4.2GHz 30 minutes Prime95 stable. I know some of you believe in 24 hours or some such nonsense for true "stability testing, but I apologize, I don't have the patience for it these days. But, I think it's quite obvious that it wouldn't be hard to do with this board and CPU.



    By the way, I did not use the Delta fan for these tests either, I swapped for a much quieter 1600RPM fan ;) 
    m
    0
    l
    Related resources
    September 28, 2009 7:21:04 AM

    OK, time for the fun stuff :D 

    This board turned out to be a real blast when pushed to the extreme with liquid nitrogen.







    As you can see, I have the Kingpin Cooling F1EE liquid nitrogen pot mounted on the board. With the same CPU/memory used at the beginning of the testing, I tweaked and tuned this chip to find is maximum potential.



    Current i5 750 SuperPI 1M world record!



    This is currently i5 750 SuperPI 32M 2nd fastest time in the world. The first place title is still held by Massman, despite my higher clock speeds. He obviously has skills in tweaking SuperPI 32M that I lack.



    Current i5 750 PiFast world record!



    Current i5 750 wPrime 32M world record!



    Current i5 750 wPrime 1024M world record!

    WOW! Obviously this board has what it takes to compete with the best is the overclocking department!!!

    So, let's wrap this up. I set out to answer a couple different questions during this testing. Does this chip compete with the AMD Phenom II series from a price/performance standpoint? And, is the i7 920 worth the 40-50% price hike? Unfortunately, I have neither the time or the means to do that completely. But, I think we can safely draw a few conclusions with this testing. Yes, the platform definitely has what it takes to compete with the best from AMD, and being that cost is about even between the two, and the fact that the Intel stuff overclocks a bit better, this is hands down the budget minded enthusiasts platform of choice right now.

    Which brings us to my second question. There is no doubt that Nehalem is more powerful than Lynnfield, but is it worth a 44% price increase? In this case I'm comparing the i5 CPU, which lacks Hyper Threading capabilities. If multitasking or heavily threaded applications care common in your typical usage, you would probably benefit from Hyper Threading. The good news is, this does not mean you need to go all the way to Nehalem, thanks to the i7 860. The price advantage is not as great as with the i5 series, but the multitasking/multithreaded performance should be just about on par with Nehalem, and it'll still be less expensive. For the rest of you, the i5 is exactly what you've been waiting for, a quad core CPU with zero compromises for single threaded applications, thanks to aggressive turbo modes.

    Finally, lets discuss the main subject of this review, the Gigabyte GA-P55-UD3R. I'm very impressed with the performance capabilities of this entry level board. Carrying on the UD3R series tradition of top shelf overclocking capabilities for a fraction of the price, this board carries the UD3R legacy into the current generation. Sure, it does have a few limitations; it only has 4 fan headers on board, and it lacks SLI support. But if you are looking for more features, your shopping in the wrong price bracket. This board is designed for the budget enthusiast...max overclocking potential and only essential features...with that goal in mind this board is an excellent choice. Gigabyte also offers a full range of models to address some of the features lacking here. For instance, if SLI support is essential to you, the GA-P55M-UD4 can be had for only $10 more.

    Lastly, when pushed to the extreme, this board did not disappoint. It obviously was built to be a beast of an overclocker. Gigabyte has spared no expense for the enthusiast community, giving this board the best of Gigabyte technology to allow for some incredible overclocking potential! Highly recommended for those who do not need SLI.


    Please feel free to post comments, suggestions for future review, and any questions about the board.

    A big thanks to Gigabyte Japan for the chance to show off this awesome product, cheers!
    m
    0
    l
    September 28, 2009 8:05:11 AM

    nice . i just downloaded that app as i have just been doing bios overclocking. thanks for the indepth report
    m
    0
    l
    a b K Overclocking
    September 28, 2009 8:25:07 AM

    What did it cold bug at?
    m
    0
    l
    September 28, 2009 8:34:40 AM

    Cold bugs are not set at any particular temp with the Nehalem and Lynnfield architectures. The cold bug fluctuates depending on voltages.

    I only cold-bugged once in Windows and that was at -85C or so...other times I was benching fine at -90 to -100C.

    Cold boot bug was -35C to -65C.

    Typically, the higher your voltages, the worse the cold boot bug gets on Nehalem & Lynnfield chips, and that was true here as well.

    My trick for dealing with cold boot bug is to boot with low voltage (1.45V here), then boost the voltage via software once in Windows. ;) 
    m
    0
    l
    a b K Overclocking
    September 28, 2009 8:50:49 AM

    Wow dude! That sucks....

    Must have been painfull having to warm the pot up to -65 when booting.

    What was the bug while in windows?
    m
    0
    l
    September 28, 2009 9:07:45 AM

    again great report. can i please ask you a question regaurding setfsb?

    when i alter the FSB using the same generator settings as you used.

    it seems to drop the multiplier down from 21x to 20x then 19x depending on the fsb.

    then i tried with 2 cores and it dropped from 24 to 23 ect as i increased the fsb.

    it was reducing the multiplier even before the levels that i know it will run at.

    is there a way to stop this. ive disabled pretty much anything in bios that could allow this.

    any idea mate?

    btw im using p55-ud3. cheers
    m
    0
    l
    September 28, 2009 9:59:44 AM

    overshocked said:
    Wow dude! That sucks....

    Must have been painfull having to warm the pot up to -65 when booting.

    What was the bug while in windows?


    Like I said, the only cold bug I hit in Windows was at 80C...attempting high voltage VTT settings. I found best OCing at VTT 1.3V or so, and at that voltage, I did not find any cold bug above -100C. I would have tested colder if I had time...which I didn't.

    slickncghia said:
    again great report. can i please ask you a question regaurding setfsb?

    when i alter the FSB using the same generator settings as you used.

    it seems to drop the multiplier down from 21x to 20x then 19x depending on the fsb.

    then i tried with 2 cores and it dropped from 24 to 23 ect as i increased the fsb.

    it was reducing the multiplier even before the levels that i know it will run at.

    is there a way to stop this. ive disabled pretty much anything in bios that could allow this.

    any idea mate?

    btw im using p55-ud3. cheers


    I saw this recently with my W3580...it was a bug in CPU-Z...try an earlier version and see if it works right?
    m
    0
    l
    September 28, 2009 11:27:46 PM

    cheers mate. upgrading from 1.5.1 to 1.5.2 fixed it.
    m
    0
    l
    September 29, 2009 3:41:34 AM

    Pictures are back up!
    m
    0
    l
    September 29, 2009 5:04:25 AM

    any reason you are running 110 speed for pci (for the air superpi tests)?

    im just comparing mine to yours and i have been running with 100. your fsb results are a good 10 more than mine.

    i know different chips have different effects ect. my ram isnt the most stable and you have much better cooling but im just curious whether this would have an effect. and why you are running it that high

    your results are quite good on air for such low vcore
    m
    0
    l
    September 29, 2009 1:51:58 PM

    The PCIe has been integrated onto the die of the CPU. So theoretically, PCIe clock speeds that are 1/2 of Bclock speeds MAY increase stability. I did not do enough testing to really say that with conviction, but it seems logical, and it worked well for me :) 

    For instance, when I did the stability tests, the PCIe speed was 100MHz ;) 
    m
    0
    l
    September 29, 2009 11:36:03 PM

    thanks for answering my questions mate
    m
    0
    l
    October 1, 2009 5:19:56 PM

    While I was working on the earlier benchmarking, I could not get CPU-Z to validate any of my speeds. I tried again tonight with the leftover 3 liters of LN2 that I had, just enough to get me a bronze cup for my suicide shot. Really wish I could have done it in XP the other night though, as I was able to reach 5290MHz at that time :( 

    m
    0
    l

    Best solution

    October 7, 2009 4:39:26 AM

    I have a question since I have the same board with my i5 750. How did you select a multiplier of x21? I was only able to select 20 and I wasn't able to type in 21 either. I'm up to 180 x 20 right now. It's at 1.225v with Load Line Calibration enabled. I still haven't decided if I want to run it higher. The most intensive application I'll run on this rig is Photoshop CS4 and Lightroom. Overclocking is fun but I'd also like it to be stable. It's running Prime95 blend test overnight as I write this.

    Thanks for your help.
    Share
    October 7, 2009 8:21:28 AM

    Go to "advanced frequency settings", then find the sub-menu for "Advanced CPU core features". Ensure your screen looks like this for best OCing potential.

    Intel Trubo Boost Tech. Enabled
    CPU Cores Enabled All
    Everything else Disabled

    The "Turbo" mode is what enables the x21 multiplier.
    m
    0
    l
    October 7, 2009 2:23:04 PM

    Stay tuned people...I'll be updating with i7 870 results within a few days :) 
    m
    0
    l
    October 8, 2009 10:31:16 AM

    In case anyone wanted more info on the cooler I user, here is a shameless spot for my favorite air cooler to date.

    I used the Prolimatech Megahalems unit...it's one of the best air coolers on the market...it performs somewhere in between the Thermalright Ultra 120 Extreme (aka "TRUE") and the TRUE Copper. And it's reasonably priced. It also has the best mounting system of ANY air cooler to date.

    Prolimatech - Megahalems CPU Cooler - 775 / 1366 / 1156 - Rev B
    Shown with older 775/1366 bracket


    New black version:
    Prolimatech Mega Shadow - Deluxe Edition 775 / 1366 / 1156
    Shown with new 775/1156/1366 compatible bracket


    They also have the 775/1156/1366 hold-down bracket available for those who already have the original version:
    Prolimatech Megahalems Socket LGA 1156 mounting kit
    m
    0
    l
    October 8, 2009 4:32:47 PM

    OK...wanted to provide a brief update with more stability testing on the i5 750, as well as my first thoughts on the i7 870 that arrived today.

    First I'd like to point out something interesting I found when I swapped CPUs today. Apparently, "socket burn" may have effected me as well. Time will tell, but I believe I do show some indication of an issue here...hopefully it does not get any worse...I'd really like to know if this is a mobo/CPU/design problem:











    The board is still fully operational at this time, and does not seem to be suffering any negative performance effects. (See i7 870 results below)

    So, a quick update to my stability testing on the i5 750. After hooking up my basic custom water cooling system...I was able to achieve slightly better results. I was stable for about 15 minutes at 4.45GHz, but impatient as I am, wanted to try to hit 4.5GHz...so I pushed up the slider a knotch and crashed it immediately. Luckly, I got a quick screen here:



    Since I knew I was at the edge of Prime95 stability, and I wanted to let it run a bit longer for you diehard stability testing guys...I backed it off a hair and did about 100 minutes at 4.4GHz...not to shabby for a $200 quad :D 



    When I got home from work tonight, I found a nice new retail i7 870 waiting for me...thank you Gigabyte Japan While I was hoping for some sweet cherry picked ES or something...what I got was a retail CPU off the shelf...a quick glance at the box showed this thing is a very close relative of the golden i5 I've been playing with :D 



    So, out came the 750, and in went the 870....right off that bat, I'm impressed. Stable at 3840MHz WITH ALL STOCK VOLTAGES forced in the BIOS (auto overvolting is normal when Turbo mode is enabled)



    A quick run at 4.2GHz shows that this one should also be a pretty decent clocker...I'll be putting it under my F1EE in a few days.



    To be continued....
    m
    0
    l
    October 13, 2009 8:53:06 PM

    Sorry to report, but it looks like the i7 870 I recieved doesn't fair so well with extra cooling. I could not even get it stable enough over 5GHz to run SPI1M. It appears that this CPU is only strong in more typical setups like 99% of you out there, and in that respect it truly shines.

    I'm dissappointed I was unable to do better with extreme cooling, but the best Lynnfield based CPUs seem to be very elusive. In my humble opinion, extreme users are best off sticking to LGA1366.





    For the rest of you, this platform is what you've been waiting for. If you're not a heavy multitasker or power user, than the 750 should work wonders for you. If your regular usage includes more intensive applications or if you multitask a lot...than the i7 860 is a perfect choice for you. It would be difficult for me to reccomend the i7 870 to anyone but Bill Gates. If you have money to burn, by all means. But for the rest of us, the i7 860 will be fine.

    A few people have asked me about fitment for the Promitech Megahalems. Here is a quick guide:















    I think the moral of the story here is that if you plan to use the Megahalems on this board, then do not get memory with tall heat-spreaders.

    If you do have tall memory heatspeaders and you're inside a case, then you're probably best off using the fan in a "pull" orientation.

    If you plan to use the top PCIe slot, you're better off mounting the HSF in a horizontal orientaion, or mount vertically with the fan on top in a "pull" orientation.
    m
    0
    l
    !