Gigabyte EX58-UD5/Extreme Discussion Thread

[vega22]

vega22, see post 921. It is just limitation of sensors reading in steps of .016 versus bios setting .00625.

Sorry ... today I'm more stupid than usual. I answered without having read all the posts pending. In addition, i just write wrong all my voltages: 1.244 and 1.228 should be 1.344 and 1.328 (i should really to use excel)
IMPRESSIVE WORK rge.
All my doubts were cleared up completely with your post. Thank you so much

[rge]

here is a good diagram of QPI and other connections, it is to the I/0 hub on desktop platform (equivalent of NB), but everything from NB/SB has to through QPI eventually to get to cpu.

http://2.bp.blogspot.com/_kttUtY0Z_z...tylersburg.gif

[Lukee]

easy tune 6 is not easy to to get working.

anybody using 6 stick and got 19x200 working?

I am using 6 x 2GB OCZ Gold 8-8-8-24 1.65v

200 x 21
Qpi 36x
Uncore 18x
Memory 8x

The key to get 8x memory to work with the 16x+ Uncore is to raise VDimm!!!!

I couldn't get

200 x 21
Qpi 36x
Uncore 16x
Memory 8x

stable at all. It would crash when I would try to run EQ2, but I saw a post of the OCing thread and tried it. I was able to get into EQ2 no problem and ran 2+ hours of Prime before I had to shut it down, so I still may need to test it a bit more, but to be honest, I don't game more than 3-4 hours so I'm ok with that atm.

Now, I am testing 8x memory and 20x uncore. (15 minutes prime and going)

Read this post

http://www.xtremesystems.org/forums/...postcount=1335

btw, I am using the Extreme F5a bios

[annaconda]

I bought the 6GB OCZ GOLD PC 1280 DDR3 memory and it did not work with my Gigabyte X58 UD5 Motherboard. I had to return it, is any body had this problem, i mean DDR3 1600mhz works for them?

[SteveRo]

rge - I also tried qpi on slow - yes very painfull to watch 3dmark06 run.
My score dropped from 24900 to 3000 - qpi on slow is correctly named - slowwwwwwwww.

Lukee - what vdimm are you at now? - 1.65 - I see it.

[Xello]

rge - I also tried qpi on slow - yes very painfull to watch 3dmark06 run.
My score dropped from 24900 to 3000 - qpi on slow is correctly named - slowwwwwwwww.

This is what i was talking about before then, slow mode doesn't really serve a useful purpose that i can see. It was included i think so we could find max stable bclks, but what use is it if those bclks aren't going to be stable at the 18x multi

It's interesting though that around 222 is the max stable bclk we're finding, because @ 12x that puts QPI to 2664mhz, well below the threshhold of stability. With all other bottlenecks removed, then, what is actually keeping bclk from going higher? It must be something to do with the cpu and the bclk itself, and if that's true i think we won't see any higher than that. Then again why should we need to, the multipliers alone are enough for everything else to reach any desired clocks right

[SteveRo]

Xello - yep, bclock (capability of all of the good x58 mobo's) is not the limiter.
Seems like the primary limiter for me is heat - hitting 90s and 100 with regularity!
Getting a bit uncomfortable...

[Hornet331]

I don't think so, there's no way i could have been stable @ 222mhz otherwise. RGE is probably right, it sets 12x.

Someone with more time on their hands should investigate 12x vs 18x QPI and the difference it makes in performance

Search the post form benchzowner regarding this topic.

maybe the "fixed" it in later bios revisions, but slowmode is usually only a state when the cpu boots up and not inteded to be used in normal operations anyway

[lowdog]

GRRRRR, I bought the Gigabyte X58 Ext and an Asus P6T - Dlx, i7 920 and 3 x 2gb Corsair TR3X6G1600C8D.

Asus board runs like a dream, I can shut it down put more ram into it reboot without issues etc etc just like all my previous Asus boards.

The Gigabyte is a NIGHTMARE, first boot up was fine, flashed to the F4m bios without issues, made some bios changes rebooted no problems. Went to install os and realised I didn't have my ide cable connected for cdrw....no issue powered down and turned off the power and pluged in the ide cable, powered on and no post, just a quick fan spin up, few leds flashed then noting. Cleared cmos = nothing.

Hrm....striped the mobo down of everything and left the battery out and the jumper on the cmos pins overnight. Next morning removed the jumper, battery back in, cpu in and 1 stick of ram in + graphics card and it booted up fine, so I think okay, powered down, turned off power, put 2 more sticks of ram in and connected hdd and cd rom, power up and same trip, no post just a quick fan spin, ledf flash briefly and nothing....rince and repeat for 3 days now. Seems like anytime it's shut down it just won't re post.

Any ideas????, remember all components work fine on the P6T-D.

This might be my first and last foray into the realms of Gigabyte.

[Lukee]

rge - I also tried qpi on slow - yes very painfull to watch 3dmark06 run.
My score dropped from 24900 to 3000 - qpi on slow is correctly named - slowwwwwwwww.

Lukee - what vdimm are you at now? - 1.65 - I see it.

I raised VDimm to 1.84. I'm pretty sure that I can run it at a lower voltage, but I haven't started tweaking it yet.

[safan80]

I raised VDimm to 1.84. I'm pretty sure that I can run it at a lower voltage, but I haven't started tweaking it yet.

I'm not willing to raise the Vdimm that high yet since the ram is the most expensive part.

[Lukee]

I'm not willing to raise the Vdimm that high yet since the ram is the most expensive part.

Hehe.. I'm not willing to keep this 24/7, just doing a bit of testing and this was a starting point. I'll start bringing it down a bit after testing.

[safan80]

Hehe.. I'm not willing to keep this 24/7, just doing a bit of testing and this was a starting point. I'll start bringing it down a bit after testing.

let me know your voltages when you get them down and you can pass 8 hour prime blend.

[Lukee]

Ok, Vdimm, down to 1.74 atm. still working on it. lol

[lowdog]

Update;

If the cmos is cleared by removing all components of off the mobo and removing the battery and placing a jumper on the cmos pins for at least an hour minimum then replacing all components then the board boots fine.......however if I then shut down and try to power on the system again there is no boot and I have to go through the whole procedure again.

Bigest POS board I have ever owned.

[safan80]

Update;

If the cmos is cleared by removing all components of off the mobo and removing the battery and placing a jumper on the cmos pins for at least an hour minimum then replacing all components then the board boots fine.......however if I then shut down and try to power on the system again there is no boot and I have to go through the whole procedure again.

Bigest POS board I have ever owned.

I had the same problem with my evga X58... the X58 bioses just are not up to par yet.

[SteveRo]

My #1 ggbt x58 mobo Christmas gift list item - lower (lower than x36) qpi multipliers.
Actually more multi's all around!

[Hornet331]

Update;

If the cmos is cleared by removing all components of off the mobo and removing the battery and placing a jumper on the cmos pins for at least an hour minimum then replacing all components then the board boots fine.......however if I then shut down and try to power on the system again there is no boot and I have to go through the whole procedure again.

Bigest POS board I have ever owned.

it seems you have the same cold boot bclck hole i have.

150mhz bclck works with cold boot.
151-199mhz bclck works with out a problem, but board resets after clodboot
200mhz bclck board cold boots.

I can live with that cause i intended anyways to go for 200mhz bclck, thought i need a bit higher QPI voltage is still only @ 1,35V + i have 12gb ram that puts addidtional streß on the mem controller.

[Chri$ch]

My #1 ggbt x58 mobo Christmas gift list item - lower (lower than x36) qpi multipliers.
Actually more multi's all around!

There is a lower multiplier..



but if you think you can reach higher BCLK then can i tell you that it doesn´t help

[Xello]

There is a lower multiplier..



but if you think you can reach higher BCLK then can i tell you that it doesn´t help

You can reach higher stable blck actually, 222 for me, but what use is that on such low QPI

[Bjynne]

You can reach higher stable blck actually, 222 for me, but what use is that on such low QPI

What is the highest stable bclk, without running qpi at slow mode?

[rge]

My highest stable bclk is bios dependent. My max bclk is 216 with qpi 36 or qpi slow mode using F4j or F4m. Though I know some got higher bclk with slow mode.

It is bugging the crap out of me why my max bclk is 216 in bios F4j and F4m, but 178 in bios F4 and F5a. I have tried everything using F4/F5a from red-lining every voltage, manually setting all timings same as F4j, etc but no go, there is some setting/code that I cant get to in F4/F5a. If its a hidden setting, I want to play with it If buggy code, then I will wait for another bios update.

I dont like the thought of not being able to use further bios updates, if it is some setting/code change that will persist in future bioses via one of the modules, hope not the case. There are some things I dont like about current bioses that have been improved, like the slow scrolling, F4j loads like molasses.

Pic is modules and what changed from F4j to F4m to F5a, usually a change in size denotes which modules changed. Going from F4m to F5a, the _en_code.bin is just english language, etc so that change is not issue, that is just the pink/red etc text changes, with the bugged, disappearing cursor on F4 and F5a when on red/pink voltage, even the changes I like have sloppy bugs.

Either the t1group or awardext.rom contains the bugged code or setting that kills my bclk, it was changed from F4j to F4m and again with F4/F5a. Havent had the nerve yet to try switching a F4j module to F5a...but if GB future bioses have improvements I want, but still prevent my bclk above 180 from one of the modules, I will try switching some modules. Hopefully with dual bios I wont kill my board...but if so I will buy a different brand. Bioses have always been weakspot of GB, unfortunately.

Anyone with an educated guess on which module most likely the culprit given pic and above?

[ReverendMaynard]

# 0 is the culprit, although it didn't really require and educated guess

[rge]

# 0 is the culprit, although it didn't really require and educated guess

Actually I did not even see the #0 code changed...lol, glad I posted that up, saved me a lot of trial and error time, thanks!

And that makes most sense.

[Xello]

What is the highest stable bclk, without running qpi at slow mode?

216, same as RGE, on F4j, which makes about 3.8ghz. I'd love to know if there is anyone who can hit the 4ghz (222x18) stable.