Gigabyte UD5P quick review and time Trial

First I must express great thanks to MsiMax lending and sending his board to me for testing.

BIOS BUGS
Gigabyte reps if you read this these need attention asap
1. Ram clocking is hindered in f4d ( these appear to be fixed in a beta )
2. HT defaults at 2600 ( appears to be fixed in beta )
3. FSB clocking is hindered with X fire, max I could get 214 fsb
4. FSB clocking is hindered with single card, max I could get was 245 fsb ( I threw plenty of voltage at it )
5. DQS training should imho be default to disable and upon a failed OC attempt
6. Bank Swizzle should imho be enabled by default
7. TRFC is stuck on 1T option regardless of setting in bios CPU-z screenshots confirm this

Lets start with a picture show, and personal thoughts and opinions.

Let me state that seeing a board and having it on a bench to experiment with various configurations are 2 totally seperate things.

I try to be unbiased as possible and here's an example......As you can see from this pic tall ram sticks are an issue in the inner banks with a popular cooler like the xigametek 1284 EE. ( goldbrick it appears your ram still has a use after I killed it :up: )

http://members.cox.net/bostonreefer3/UD5P1.jpg

Ram with the more common heatspreaders are not an issue although still require the xigamatek to be unlatched for inner dimm installation...

http://members.cox.net/bostonreefer3/UD5P2.jpg

I took this one step further and tested the True 120 clearance.....

http://members.cox.net/bostonreefer3/UD5P11.jpg

As you can see with a True the inner dimm installation can be done rather easily.......the blame is now rather subjective.....

http://members.cox.net/bostonreefer3/UD5P12.jpg

PWM for memory

http://members.cox.net/bostonreefer3/UD5P3.jpg

PWM for cpu

http://members.cox.net/bostonreefer3/UD5P4.jpg

Loading up the available slots with different configurations to show Pros and cons ( the sound card is from another install and is just used as a prop for those that may have a Pci E sound card )

http://members.cox.net/bostonreefer3/UD5P6.jpg
http://members.cox.net/bostonreefer3/UD5P7.jpg
http://members.cox.net/bostonreefer3/UD5P8.jpg
http://members.cox.net/bostonreefer3/UD5P9.jpg

And to wrap it up a full board shot.

http://members.cox.net/bostonreefer3/UD5P10.jpg


I've got a clean install going in now so results later, be patient they will be done when they are done.

Well my thoughts, I'll take this over my 790GX anyday. In fact I don't really know what the issues people are having are......

I found 2 bugs....Bios defaults HT at 2600 and the command rate option for ram timings is broken and stuck at 1T even though set at 2 T in bios.......*edit* Bug with trfc not detecing by SPD, and also a bug with DQS resetting itself to enabled....default should be disabled for compatibility purposes.

This was about 30 minutes of playing with board after everything was installed, I focused less on CPU clock and more on breaking 1600 memory speeds and 2800+ nb speeds...as thats what people were complaining about. This was done on the existing bios.......Dirty, unoptimized.....haven't had any time to tweak anything, just find stable bench speeds.

http://members.cox.net/bostonreefer3/UD5P13.JPG

http://members.cox.net/bostonreefer3/UD5P14.JPG

http://members.cox.net/bostonreefer3/UD5P15.JPG

Beta Bios, Please do not ask for this, I am not supposed to even have it..........................Noteable fixes.....HT speed is loaded at default correctly, TRFC is set properly by SPD, better ram compatibility overclockability. Command rate still stuck at 1T regardless of setting to 1T or 2T in bios, DQS still defaults to enabled....

Max I can get out of this ram, I tried ocing with AOD and HT went 3 steps and then crashed......memory BSOD. 104mhz for free, no complaints by me and these are budget sticks.
http://members.cox.net/bostonreefer3/UD5P16.JPG

Had to downclock ram a little to match NB max. I'm sure I could get better NB speeds but on a time schedule this will have to do. Tossed some PI in there to show some efficiency, almost cracked sub 67,500. Keep in mind I opted to use my more tolerable scythe 80 cfm fan, I could get a bit more with my delta 150+ cfm fan.......not in the mood for it today though.
http://members.cox.net/bostonreefer3/UD5P17.JPG

720 BE retail benches, not optimized i'm pressed for time, wasn't worried about performance just passing, Wanted to bench at 3900+ but don't have time to fine tune it, quite sure its possible, just 1 bench failed so I scrapped the plan in order to get done quick, bed is calling I'm tired, have a headache and still have to crank out a prime stable........

http://members.cox.net/bostonreefer3/UD5P21.JPG

http://members.cox.net/bostonreefer3/UD5P22.JPG

Prime stable.........use my bios settings I posted.....add .025NB vid to those settings and .050 vcore (1.6v) although I would not run this personally for 24/7, just showing it is possible.....

http://members.cox.net/bostonreefer3/UD5P23.JPG

Are those the new 1600 G-skills that you bought? I had to RMA my patriots and I bought the same set, I wonder how are they going to perform with that board.

Chew*, It would be great, if you can include some bios shots for your stable trials. Thanks

I'm surprised no one has mentioned the RAM installation issue before, it was obvious to me before I even bought the board, but knowing I'd be running only 2x2Gb sticks in the white slots it didn't bother me even though my Scythe Ninja renders the blue slots all but unusable.

Quote:

---

Originally Posted by suphiceto

Are those the new 1600 G-skills that you bought? I had to RMA my patriots and I bought the same set, I wonder how are they going to perform with that board.

Chew*, It would be great, if you can include some bios shots for your stable trials. Thanks

---

Hmmm I can do that with the current bios.

Quote:

---

Originally Posted by chew*

Hmmm I can do that with the current bios.

---

Thanks, I have the same board and I had some issues with the few things. That is why I asked.

I do agree that the ram slots could have been pushed out a bit, luckily most users will either not be affected or will be able to work around it. Still though...

Intital results post 2...

That is very impressive Chev, I cant belive you were able to get NB to 2800+ on air.
Here is my voltages and I cant pass 2200mhz on NB.
NB Voltage= +0.1
SB/HT Voltage= +0.1
NB/PCI/PLL Voltage= +0.1
CPU NB VID=+0.1
What sort of Voltages did you give for these values? Are you also going to try with your PII 720? Thanks

As usual, Chew gives us MUCH better info that 90% of the "expert review sites"

:up::up::up::up::up:

Thanks as always

try tossing +.250 NB vid volts at the chip ;)

Hi, Bought this board and a 720BE along with Patriot PDC34G1600LLK. Had to manually set timings in bios to get rated speed but no complaints. No luck with 4th core (0904 EPMW) not even a crippled one. Using same Vcard out of 939 rig (gtx260/216SC) I gained about 4000 3D'06 points. Switched to HIS 4870 1G (750/900) and scored about the same.
Warning: Windows update killed both Ethernets! Had to uninstall Realtek drivers and reinstall drivers from MB disc.

Played L4D all night and Butter Smooth!!! I love how the ATI card's picture is much prettier than my Nvidia (3DMark06).
:up:

Edit: f3 bios

Quote:

---

Originally Posted by chew*

try tossing +.250 NB vid volts at the chip ;)

---

You are owesome! I didnt want to give that much of volt and never tried it before. I couldnt boot up with .250 and NB at 2800, but It seems like I can do +.225 and NB at 2600. It is doing prime 1024 ok right now. If it stays stable like this, would +.225 be safe for 24/7 ? I still cant get my cpu over xmulti 18 and +.175v. It seems to get hot and crush on me. I belive once I lap the cpu and HS, I should be able to raise that as well?

Quote:

---

Originally Posted by chew*

also a bug with DQS resetting itself to enabled....default should be disabled for cpmpatibility purposes.

---

How does DQS training hurt compatibility?

Quote:

---

Originally Posted by Quintero

How does DQS training hurt compatibility?

---

Disable it and run memtest and then enable it and run memtest, note bandwidth each time.....then you will see why it should be default off.... It drives memory harder.

Suphiceto, you need the find the happy balance between vcore, nb vid, clock speed, nb clockspeed, and temps.....increasing nb vid volts will increase your temps just as easily as vcore does.

Don't take this as the LAW because I myself still am not sure what is safe for 24/7, none of us have had the chips long enough but I would say that you do not want to go over 1.575vcore for 24/7 use, and no more than +.275NB volts for 24/7 use.

I will post a bios screenie of prime stable settings later however I do not reccomend my bench settings and thus will not post bios shots of those because they are way out of what I consider the safe range........

Quote:

---

Originally Posted by chew*

Disable it and run memtest and then enable it and run memtest, note bandwidth each time.....then you will see why it should be default off.... It drives memory harder.

Suphiceto, you need the find the happy balance between vcore, nb vid, clock speed, nb clockspeed, and temps.....increasing nb vid volts will increase your temps just as easily as vcore does.

Don't take this as the LAW because I myself still am not sure what is safe for 24/7, none of us have had the chips long enough but I would say that you do not want to go over 1.575vcore for 24/7 use, and no more than +.275NB volts for 24/7 use.

I will post a bios screenie of prime stable settings later however I do not reccomend my bench settings and thus will not post bios shots of those because they are way out of what I consider the safe range........

---

Thanks man, are you going to test your 720BE on that board?

Quote:

---

Originally Posted by suphiceto

Thanks man, are you going to test your 720BE on that board?

---

yes later tonight or tommorow, working on posting up max NB speeds and max fsb speeds atm then a bios screenie with x4 first.

Quote:

---

Originally Posted by chew*

Disable it and run memtest and then enable it and run memtest, note bandwidth each time.....then you will see why it should be default off.... It drives memory harder.

---

Thanks for the quick reply. I believe what you're saying, but can't quite find an underlying technical explanation that makes sense. AFAIK DQS training takes place during every initialization, and factors like voltage, clock frequency and temperature affect the settings. With the latter in mind, it seems like a good idea to re-calibrate the strobe delays immediately after running some memory-intensive tests. This should improve scalability, but I don't understand why it affects BW at the same frequency. I was under the impression that the controller calibrates the memory every time you switch on the computer, but I don't think this feature "interrupts" the memory after the initialisation. Perhaps the controller performs this calibration during every startup (even when DQS training is disabled), and continues to calibrate DQS throughout the session to account for temperature variations (when enabled) ?

New beta bios F4D

http://www.gigabyte.com.tw/Support/M...ProductID=3005

Quote:

---

Originally Posted by bro20000

New beta bios F4D

http://www.gigabyte.com.tw/Support/M...ProductID=3005

---

No thats the regular bios.........look at top screenies...thats why I showed the motherboard screenies in the first set of tests....and why its hidden in others....

Quote:

---

Originally Posted by Quintero

Thanks for the quick reply. I believe what you're saying, but can't quite find an underlying technical explanation that makes sense. AFAIK DQS training takes place during every initialization, and factors like voltage, clock frequency and temperature affect the settings. With the latter in mind, it seems like a good idea to re-calibrate the strobe delays immediately after running some memory-intensive tests. This should improve scalability, but I don't understand why it affects BW at the same frequency. I was under the impression that the controller calibrates the memory every time you switch on the computer, but I don't think this feature "interrupts" the memory after the initialisation. Perhaps the controller performs this calibration during every startup (even when DQS training is disabled), and continues to calibrate DQS throughout the session to account for temperature variations (when enabled) ?

---

Not quite sure to be truthfull, I just know it that I can't clock ram as high and bandwidth was increased in memtest when enabled on both this board and the 790GX. I also compared the added bandwidth in benches versus the ability to clock higher and the higher clocks got better results.

Thanks for the Beta bios teasers, nice results. You think that with the new BIOS, this board is up to par with the M4A79Ts?
For that matter, is the UD4P bios going to be receive a similar update as well?

Quote:

---

Originally Posted by Miwo

Thanks for the Beta bios teasers, nice results. You think that with the new BIOS, this board is up to par with the M4A79Ts?
For that matter, is the UD4P bios going to be receive a similar update as well?

---

I honestly have not tested the M4A79T, the M4A79 deluxe ddr II variant is on the bench but I have not put any hard work into it as of yet So i can't make a fair comparison.

The bios seems to need a little more work, Although I haven't tested 1T on ddr 3 boards before, on ddr 2 boards there is almost no benefit in PI 1m and it costs ram clockability.

With the 1T command rate fixed ram clocking will likely be better, the DQS issue is only on a failed OC attempt, I can shut it off so its not that big of a deal.

No clue on when bios or bios's will be available, as I stated I'm not supposed to have it so I have no inside connections with gigabyte etc am not in the know.

cool thanks. Hopefully you can get your hands no the M4A79T soon :)