***Asus Rampage III Extreme Owners Thread***

[shadewither]

Seems like some people are having problems with Bloomfield and GFLOPS in Linx and some aren't...so perhaps it's another problem?

Have you checked other benchmark results?
LinX/y-cruncher seems to count on some cycle-based timers, which might cause wrong execution time and thus wrong GFlops.
You could try y-cruncher, which run-to-run delivers a more stable peformance than LinX.
(and count realworld time against the reported execution time)

I clocked the 980X up to 4Ghz (sample needs a whopping 1.4Vcore lol) and got 57-58Gflops §6GB 1600C8 on Win7 64), I see Gautam with 980X at 4.7 having only 51Gflops...

It seems that LinX doesn't scale (well) with 6-cores.
So it might result in a very unstable GFlops because of resource contention, etc.

[Soulwind]

Kind of an off/on-topic question.

I'm using PC Probe II right now to monitor: CPU temp (overall temp), MB temp (which is the northbridge, I think?), IOH temp, ICH temp, and Pump RPMs (have the rpm wires plugged into cpu and power supply fan headers)

I'm also using RealTemp to monitor the CPU Core temps, since PC Probe doesn't seem to allow for showing actual core temps; and SmartDoctor to monitor the video card's GPU and Memory temps...

So:
1) Is there any way to monitor the voltage regulators/mosfets temps or memory temps (other than an external temp probe) ?
2) Should I be using some other monitor (like maybe Everest or something) instead of PC Probe/RealTemp/SmartDoctor ?

[Leeghoofd]

impossible comparison Alien Grey, this is not S775... calculations don't go just like that... which problem size, RAM used, clocks,....HT on/off,... too many variables...

Let's do a head to head with the 980X at 4Ghz and your quad at 4.5... at linX you might beat me, for the rest your going down... if you want to brag with ya LinX screens be my guest, I know where the raw power is at for my needs.

Performance clock for clock this is the better X58 board, if you want to believe it or not.

Again LinX does not show the performance clearly, again proving my point in it's value regarding the gigaflops. Maybe it's a good stress tester but that's really where it all seems to end... only users that have similar systems like Oc nub, using same components, OSses and co can compare...

[donmarkoni]

Quick stupid question, I have a 940 using the RIIIE, I have the multi at 23 in the bios and seem to not break 4gig with high voltage. I can run 3.8ht on at only 1.3v stable linx but cant get stable at 4gight on air near 1.38-1.40

Do you recommand multi 22? Is that possibly my problem?

Thanks

You should try 23x multi with C1E Support Disabled and Intel(R) C-STATE Tech Disabled.
If you want those Enabled (like me ) then go for higher BCLK and 21 multi.
Check here for an example.

So:
1) Is there any way to monitor the voltage regulators/mosfets temps or memory temps (other than an external temp probe) ?
2) Should I be using some other monitor (like maybe Everest or something) instead of PC Probe/RealTemp/SmartDoctor ?

Short answers:
1) No. I use MB's integrated external temp sensors.
2) Yes. I use Everest.

[Alien Grey]

impossible comparison Alien Grey, this is not S775... calculations don't go just like that... which problem size, RAM used, clocks,....HT on/off,... too many variables...

Let's do a head to head with the 980X at 4Ghz and your quad at 4.5... at linX you might beat me, for the rest your going down... if you want to brag with ya LinX screens be my guest, I know where the raw power is at for my needs.

Performance clock for clock this is the better X58 board, if you want to believe it or not.

Again LinX does not show the performance clearly, again proving my point in it's value regarding the gigaflops. Maybe it's a good stress tester but that's really where it all seems to end... only users that have similar systems like Oc nub, using same components, OSses and co can compare...

It really does because I've seen 90+ GFlops with a 980X running at 4GHz+ so you can say what you want but theres something wrong with the board or the BIOS. The problem is situated between memory and the CPU. The CPU can only calculate what it receives and clearly there's some kind of slowdown and the outcome is bad.

Clock for clock you should have better GFlops than my Q9650@4.05GHz that gives me 57+ GFlops. The score should be 1 or 2 GFlops better with a Quad.you get better results than using all available free memory.

Good luck with your Rampage Extreme III. I think that you soon are going to replace it whenever something new is available from the competition.

[Leeghoofd]

Well got UD7-9, XPower and REX III here. Think I know what they and what they don't do...

980X 90+ Gflops at 4Ghz, so at 4.5 it goes over 100 then ? what settings, how much ram,...

Till now all the new boards have bios issues : ranging from bugged dividers, unable to boot at high QPI speeds,... Usually things overclockers don't really care about as most vital parts are there and it still enables to kick ass for all 3 boards under LN2.

For the 24/7 users there are still bugs to be ironed out... Saddest till now is the cold boot issue that seems to pop with some users, which is very frustrating. for your pleasure I will put the UD9 and the REX III head to head tonite for a quick 4GHZ 980X run. I hardly believe I will see 50 on the REX III and 90 on the Giga... we will see tonite...

But plz explain me if it's so bad in LinX. How can it perform and outperform the competition in 2D/3D ? (1.2bios Xpower, F4A for the UD9 and 0704 REX III ) No I don't include LinXin my testsuite...

[donmarkoni]

I've seen 90+ GFlops with a 980X running at 4GHz+

You have seen it with HT off. With HT on it drops under 70 for unknown reasons.
4.5GHz -6C/6T - 96GFLOPS, 4500MHz - 6C/12T - 68GFLOPS

[Alien Grey]

Well got UD7-9, XPower and REX III here. Think I know what they and what they don't do...

980X 90+ Gflops at 4Ghz, so at 4.5 it goes over 100 then ? what settings, how much ram,...

The screenshot was posted here on the forums but I can't find it anymore but I know the 980X was running at 4GHz+ and the results were just fine for that speed at 6 cores.

Till now all the new boards have bios issues : ranging from bugged dividers, unable to boot at high QPI speeds,... Usually things overclockers don't really care about as most vital parts are there and it still enables to kick ass for all 3 boards under LN2.

It isn't only the new boards that have BIOS issues you can easily say that most of them are ASUS boards.

For the 24/7 users there are still bugs to be ironed out... Saddest till now is the cold boot issue that seems to pop with some users, which is very frustrating. for your pleasure I will put the UD9 and the REX III head to head tonite for a quick 4GHZ 980X run. I hardly believe I will see 50 on the REX III and 90 on the Giga... we will see tonite...

But plz explain me if it's so bad in LinX. How can it perform and outperform the competition in 2D/3D ? (1.2bios Xpower, F4A for the UD9 and 0704 REX III ) No I don't include LinXin my testsuite...

It's just like you say the board is good for benching but for 24/7 use it isn't and maybe ASUS can fix this if its a BIOS issue because they're going to need it.

At 4GHz you should have around 80-85 GFlops and it's possible you see that on the Gigabyte but I don't remember anymore what board they were using. All I remember is that the result was really good and that 6 cores wasn't a problem.

[zpaf]

My first success linx pass for my setup at 4.2 Ghz with ram at 1600Mhz.



and another one fail with ram at 2000Mhz.
try to up down voltages to cpu, qpi, ram but for some reason linx stopped always at step 10.

[shadewither]

The CPU can only calculate what it receives and clearly there's some kind of slowdown and the outcome is bad.

Nothing to do with MB/BIOS.

OS (and apps) decide which core to execute at; apps dictates how large the working set is.
If the working set is too large or the task scheduling is screwed, performance will become miserable.(two simple reason of performance drop)

Clock for clock you should have better GFlops than my Q9650@4.05GHz that gives me 57+ GFlops. The score should be 1 or 2 GFlops better with a Quad.you get better results than using all available free memory.

HT decreases Linpack GFlops by a lot.
Nehalem delivers higher GFlops than Penryn with HT off.

[Alien Grey]

Nothing to do with MB/BIOS.

OS (and apps) decide which core to execute at; apps dictates how large the working set is.
If the working set is too large or the task scheduling is screwed, performance will become miserable.(two simple reason of performance drop)



HT decreases Linpack GFlops by a lot.
Nehalem delivers higher GFlops than Penryn with HT off.

You've got a point there and that's why I say to use less available memory so you know for sure that it isn't going to use the page file. As for the OS I think everyone is using a 64 bit edition.

It doesn't matter if it's Windows XP X64 or Windows 7 X64 the results are the same for me. I don't know about Vista I never used it.


I think we all know that you get higher GFlops with HT off but I don't think that this explains the really low GFlops.

[Deboyzfun]

There is a Bios setting to turn this off.

From the Asus Forum --- To disable the standby power going to your PS/2 and USB ports, simply enter BIOS, and enable EuP support, then you should have the standby power turned off once and for all.

i was wondering what the EuP did

[Deboyzfun]

New bios, 0802, it´s not a beta, it´s a formal and it´s going to be official in a near future from Asus site.

No notes of what they have improved

Happy testing guys


I have added it to the OP also.

Wow what a treat just got 704, now another one to try!

[donmarkoni]

My first success linx pass for my setup at 4.2 Ghz with ram at 1600Mhz.



and another one fail with ram at 2000Mhz.
try to up down voltages to cpu, qpi, ram but for some reason linx stopped always at step 10.

Is it just Probe's inaccuracy in RAM voltage @ 2000MHz or is it actually set lower (1.61 compared to 1.65 @ 1600MHz)?

[Splave]

check the read point by the power button

[zpaf]

Is it just Probe's inaccuracy in RAM voltage @ 2000MHz or is it actually set lower (1.61 compared to 1.65 @ 1600MHz)?

It's real I set this voltage trying to find the best set but no luck.
I think there must be a balance < 0.5 between qpi and dram voltage.
Are there any other values other than cpu vcore, qpi and dram that can help to be stable at 2000Mhz ?

[OC Nub]

@OC Nub
Another thing to check: Is GFlops REALLY LOW?
1. CPU core temp should be several Celsius lower, if GFlops is really low;
2. Linpack GFlops calculation might be screwed.
For example, if you use SetFSB/Turbo-V to change frequency in Windows, the reported GFlops is wrong.

The CPU temp in everest or pc probe it hardly gets warm sometimes mid 40's c. I can tell as soon as it starts running that the gflops are going to be low. I have remounted the cpu several times, last night I took everything apart and looked the board over for any damage/scratches etc, it looks perfect. Also have tried several i7 chips, different PSU, 3 kits of memory.

@Leeghoofd,
I will try running some different tests to see if the CPU scores well, if it does then great I wont worry about gflops in LinX. I know you arent a fan of LinX, and I wasnt either until the LinX4.5ghz thread started. I normally prefer blend for stability testing, I could care less about wprime or pcmark05 but if you were getting horrible score in those Im sure you would be unhappy. If I wanted to post a run in the LinX thread I would want the gflops to match with the speed it is running, other than that I really dont care about what gflops it is showing. I was hoping to get my 920 to pass at 4800mhz, I am 75mhz away from that on the UD7 but even if the board does help to do that I wouldnt post the screen if gflops were real low.

[Leeghoofd]

Comparing the Linx runs for your pleasure now, OS Windows 7 64 bit, LinX 0.64 980X at 4Ghz ( 20 x 200) 3200 uncore, 6GB 1600mhz ram speed 9-9-9-24 1T TRFC 88 RTL 59-60-60. LinX tested at 5120 ram

UD9 with F4A bios, REX III still with 0704 :







UD9 is faster : 50 vs 46 at 4ghz, then I start to mess with the CPU features like I always set them for LN2 runs... Asus drops to 41 Gflops, UD9 to 43... SO that reflects about your results I think OC_Nub with the asus trailing behind 1-2Gflops. But if I run Wprime, Cinebench, 3dmark01/06 the Asus is a tad better than the UD9...

No idea how the one guy got 96Gflops on 980X at 4.5Ghz... beats me... and really doesn't interest me as I think it's more a software manipulation then anything else...

So I hope Alien sees that the Giga does NOT overpower the Asus with LinX. There's a small loss, but again that doesn't seem to reflect in any of my benchmarks that I always use as testsuite for the motherboard/CPU review...

[Leeghoofd]

It's real I set this voltage trying to find the best set but no luck.
I think there must be a balance < 0.5 between qpi and dram voltage.
Are there any other values other than cpu vcore, qpi and dram that can help to be stable at 2000Mhz ?

Back 2 back delay ?

[OC Nub]

Comparing the Linx runs for your pleasure now, OS Windows 7 64 bit, LinX 0.64 980X at 4Ghz ( 20 x 200) 3200 uncore, 6GB 1600mhz ram speed 9-9-9-24 1T TRFC 88 RTL 59-60-60. LinX tested at 5120 ram

UD9 with F4A bios, REX III still with 0704 :







UD9 is faster : 50 vs 46 at 4ghz, then I start to mess with the CPU features like I always set them for LN2 runs... Asus drops to 41 Gflops, UD9 to 43... SO that reflects about your results I think OC_Nub with the asus trailing behind 1-2Gflops. But if I run Wprime, Cinebench, 3dmark01/06 the Asus is a tad better than the UD9...

No idea how the one guy got 96Gflops on 980X at 4.5Ghz... beats me... and really doesn't interest me as I think it's more a software manipulation then anything else...

So I hope Alien sees that the Giga does NOT overpower the Asus with LinX. There's a small loss, but again that doesn't seem to reflect in any of my benchmarks that I always use as testsuite for the motherboard/CPU review...

Thanks for all of your testing ! That is good to know, I was concerned that something was really wrong with my board. I just finnished installing windows, will do some other testing besides linx.

[Leeghoofd]

Maybe something is still wrong mate, still not seeing 90Gflops. Only CPU feature I didn't see on the Asus bios is the Prochot feature...

No idea really why it does this, all the rest seems fine... Hope you guys can get at least the same clocks as on the UD7... the Xpower guys need one decent bios instead of the current ones Don't give up on the motha yet !

[Leeghoofd]

Last run, applied all settings of the Gigabyte : iso of the skews and amplitude at auto set them like in the screenshot, also spectrums enabled and the CPU features like in the screenshot. REX III went from 46Gflops to 53 lol... It's all in the settings...



Bios settings :







CPU features ( the more I disable the lesser the Gflops ) :



Hope it helps Nub...

[kwangkwang]

Maybe something is still wrong mate, still not seeing 90Gflops. Only CPU feature I didn't see on the Asus bios is the Prochot feature...

No idea really why it does this, all the rest seems fine... Hope you guys can get at least the same clocks as on the UD7... the Xpower guys need one decent bios instead of the current ones Don't give up on the motha yet !

Concur with my result, 53 Gflop @4.275 rock stable 5 hr Linx, 10 hr P95 Blend.

90+ Linx at this setting, never get that high

I also notice IBT is reporting higher than Linx GFLOP but both use linpack.
CPU temp for LINX same level as Everest CPU stress.
CPU temp for IBT same level as Everest FPU stress and P95 Blend (a tad higher temp)

Linx GFLOP is not accurately reporting to spec. IBT and some other program reports closer to spec GLFOP.

[OC Nub]

Leeghoofd, Thank you for the bios pics and for taking the time to help try and figure this all out. I really appreciate it.

Tried some 06 on both boards, same basic setup, fresh install on both. I have these in 2 seperate cases both watercooled so it is a pita to switch out all hardware. I have a pair of 5870's and 2 920 do's so this was run using 2 different cpu's and gpu's but all clocks and gfx settings were the same. Ran 06 3 times on each board and these are the best scores of the 3. I was mainly looking to see the cpu score.



Uploaded with ImageShack.us


Uploaded with ImageShack.us

[OC Nub]

Leeghoofd,
I had a lot of things disabled, set my bios up the same as in your bios pic all but clocks and voltages. Even set the cpu differential amplitude~700mv, im on bloomfield not gulftown but tried it. Worked like a charm. You are the man, again Thanks.



Uploaded with ImageShack.us