Gigabyte MA790FX DQ6

[Lightman]

WOW, This bios F3 final is much better than F3B..
Your score is very nice Lightman.
I can will test more, now i can run 2525mhz stable, can not do it before.

Thanks!

That is what I can do under x64bit OS, my best from XP MCE was 14380 points at 2640MHz and that was with low clocked NB. I think I will give it another go with high clocked NB

[KTE]

Today I was playing with NB and L3Cache speeds

Some interesting findings...

Very nice mate. How'd you get NB that high?
Can I have a look at your voltage settings at those NB speeds and which voltages you can control in AOD please?

I am sure MSI board is different. If the next BIOS is not as good then I will probably try your board.

[Lightman]

Very nice mate. How'd you get NB that high?
Can I have a look at your voltage settings at those NB speeds and which voltages you can control in AOD please?

I am sure MSI board is different. If the next BIOS is not as good then I will probably try your board.

KTE 2.5-2.6GHz NB at default voltages (1.165V)! We can't adjust that setting from the BIOS yet....
My CPU simply likes high NB clocks! I set multi to anything I'd like in up to 2.8GHz speed and viola!
Looks like half of my Phenom silicon is really good and other half is not so brilliant....

Here is AOD SS showing max. I can set for NB voltages...

My 24/7 setting at the moment (except NB voltages which are default ). Read values from left column because this beta is + F3 BIOS is misreading real freq. under XP x64. It works fine for Vista x32 though.

PS. Max NB I was able to bench was 2840MHz

[Syz`]

Max score I can grab is 13.8k 3DMark06.

That's with the Phenom @ 2.6Ghz - stable, btw and the two Radeons at 850/2500.

CPU is limiting them so badly.

[KTE]

Thanks for that Lightman.

KTE 2.5-2.6GHz NB at default voltages (1.165V)! We can't adjust that setting from the BIOS yet....

You have SB but no CPU VCore in AOD? That's good in a way but bad in a way too.

Here is AOD SS showing max. I can set for NB voltages...
http://img401.imageshack.us/img401/4...ancetabnv8.png
My 24/7 setting at the moment (except NB voltages which are default ). Read values from left column because this beta is + F3 BIOS is misreading real freq. under XP x64. It works fine for Vista x32 though.

I hate to be the bearer of bad news, but are you sure about this all?
CPU-Z has a known bug with flawed NB/CPU readings if you throw the NB too high in BIOS. We've been dealing with it for around 3 weeks now. What happens is, the system will downclock everything to keep your NB low near 2GHz. AOD as far as we know will report the most accurate values as the actual real frequencies. It happens because Phenom 9500/9600 cannot run more than 9x NB multi.
To verify all this, run EVEREST Ultmate and Crystal CPUID. They use various methods to detect the speeds. See if your EVEREST/CCPUID is reading the same as AOD (mem/ht/cpu), if they are, then what I'm saying is true and if not, then not so which is a first. I have a feeling your system is actually downclocked to keep NB near 2GHz and so the core/mem is actually underclocked and those AOD frequencies are correct. Because they only appear that way if Phenom takes a safety measure and actually underclocks despite the BIOS values you choose when they're too high (override).

I've attached an image of what the downclock looks like with a buggy CPU-Z and the correct AOD. Remember, NB actual high speeds will need more volts.

High NB Clocks? Follow this... >>>

Me/Tony/Sammi/lukija battled and reported through this around 10-15 days back.
Follow these in order to get a good grasp of what I'm mentioning and test;

1/ See what Sammi said for starters: http://www.xtremesystems.org/forums/...&postcount=200
2/ This was my reply (pics): http://www.xtremesystems.org/forums/...&postcount=202
3/ For instance, this is how the downclocking looks in AOD: http://www.xtremesystems.org/forums/...1&d=1197286076
4/ lukija's advice and tests: http://www.xtremesystems.org/forums/...&postcount=217
5/ And my reply of what I discovered: http://www.xtremesystems.org/forums/...&postcount=218

Let me know what you get, it's interesting. Hope it helps.

.

[bro20000]

what i dont get is what is the green button in the right hand corner in the new version of amdoverdrive! What does it do!

[KTE]

what i dont get is what is the green button in the right hand corner in the new version of amdoverdrive! What does it do!

Read the link in my last post where Sammi talks.

[Lightman]

@KTE I did another test session to figure out that NB speed.

All software is showing same low CPU clocks as AOD is when running NB above 9x multi... Thing is performance is where it should be for 2.4GHz Phenom! I did couple of SPi tests, Cinebench4D, 3DMark2001/2006 and WinRAR with various NB multi.

It looks like non of the software is correctly reporting values.... AOD is showing wrong CPU/memory speeds, CPU-Z is showing hard to believe NB clocks.

Still I can report slightly different results when changing NB clock. For example WinRAR is faster at NB 9x than at 10x by 30-80KB/s but 11x multi gives same results as 9x???
That run of 3DMark from my previous post where I got 13488points NB2736 is higher than todays run 13408points with NB2052 (9x228).
On top of that Cinebench xCPU scaling is at best 3.09x when NB 9x contrary to 3.12-3.14x when NB 12x....

Differences are small enough to be counted as a measuring error but still there is a pattern....

If only we could test L3 Cache bandwidth using some software

[KTE]

Run POV-Ray 3.7 with high NB clocks as you have above please with the readings for what AOD/CPU-Z shows and I'd like to see what results it gives you.

Because I'm aware of its perf scaling from 1.8GHz uptil 2.85GHz CPU 2.16GHz NB/HT, 518 4-4-4-4 RAM and can tell because I've ran it with downclock too.

[Lightman]

Run POV-Ray 3.7 with high NB clocks as you have above please with the readings for what AOD/CPU-Z shows and I'd like to see what results it gives you.

Because I'm aware of its perf scaling from 1.8GHz uptil 2.85GHz CPU 2.16GHz NB/HT, 518 4-4-4-4 RAM and can tell because I've ran it with downclock too.

Done!


NB1617MHz

NB2079MHz

NB2541MHz
x64 Hope it helps

Because this is raytracing engine we can't see much of difference between different settings. Still pattern is clear and results are repeatable...

BTW 3DM2001 scores 2000+ points lower with 1617MHz NB and 800+ lower with 2079MHz NB compared to 2772MHz NB

Edit:

NB1617

NB2541

NB2772 was 39500+ but I forgot to save a pic

[KTE]

Thanks. I don't really know what to make of that Lightman, honestly.

Since in 32-bit which is slower than 64-bit for this app and Cinebench, when I ran CPU 2420MHz / NB 1980MHz / HT 1100MHz / RAM 440MHz 4-4-4-4-24 2T single channel, the results were a little faster than all those: 1642.89 PPS.





Also, synthetic benches won't be accurate here and that's why I'm not running them because they will obviously give grand results (they did with my 1.8GHz-3.1Ghz Nb results too). Real life ones will count fully though because they show the real hardware power. POV-Ray scales from 1498-1650 PPS from 2.2GHz to 2.42GHz on CPU MHz with Phenom.

I need some time and after about 2-3 hours I'll be able to downclock mine to the same MHz to see what difference it gives.

The problem with all this is, its already an established fact that CPU-Z bugs out like this when you put the NB multi high and when it does other better apps that read core registers will still read the correct values, not one app, all of them. Sammi himself thought these high NB values above 9x multi are bugged readings when I showed them him (2.6-3.1). Tony also could not get above 9x NB multi in all his testing but bugged results and there's a few more that have tested them and cannot (always get this bug). So I would be very weary. I thought the same things you are mentioning now around 3 weeks back in the K9A2 thread before I spent much more time thinking>testing and worked it out.

In fact, I guarantee you no matter which board you use, be it XP or Vista, if you change NB multi to above 9x for 9500/9600 you will get these bugged CPU-Z readings and AOD/EVEREST/CCPUID will show everything as downclocked.

Do something. Most software runs off the most accurate method of calculating time which is based off CPU MHz readings (i.e. Super Pi). It uses Windows Performance Counters. Check what Windows says your CPU is booting up at. Those are BIOS reported values to it.

[Lightman]

KTE I presume you were running SSE2 version of POV-Ray for that score... Guess waht, I get almost same score as your when running that executable! Score 1651pps - NB2079(9x) CPU2424

Edit:
I set my system to almost same setting as yours...


As we can see, despite 4MHz lower core speeds I'm getting higher score compared to my previous run...

BTW Windows System Properties says 1.82GHz for that run....

I know - next thing will be to measure time required to finish a task using some precision external clock (Might be atomic or at least quartz )

[KTE]

This is getting ... interesting.

So you are saying CPU-Z is correct and all else is wrong. Can you validate any of those CPU-Z results with high NB?

If your NB is actually running high, you should have very high EVEREST Mem bandwidth and low latency, sub 47ns. I've tested the high NB between 1.8GHz and 2.2GHz at same everything to see this effect. So, what do you get?

Thanks as usual.

[Lightman]

This is getting ... interesting.

So you are saying CPU-Z is correct and all else is wrong. Can you validate any of those CPU-Z results with high NB?

If your NB is actually running high, you should have very high EVEREST Mem bandwidth and low latency, sub 47ns. I've tested the high NB between 1.8GHz and 2.2GHz at same everything to see this effect. So, what do you get?

Thanks as usual.

I will do that later.

This morning a ran exactly same setup as yours and POV-Ray scored 1644PPS (NB1980).

Everest latencies and bandwidth are improving slowly... Thing is that when L3 cache is clocked higher than CPU profits are smaller that you would hope simply because bandwidth can't be utilized efficiently.

Now I need to go to work
Will be back investigating this Phenom(en) later today .

[KTE]

Sadly, with regret I have to inform you Lightman, very unfortunately, you will be divorced from this this exicting joint venture soon and have no one to carry on alongside with.

I have to deliver the 9500 I have to the buyer in 4-5 hours.

Thanks.

[KTE]

Dearest Lightman

I have attached a utility to measure your CPU Clock speed according to the CPU registers (directly), thanks to the works of Jan.

Please run the BAT file in there in the "NB overclocked mode" and it will tell you your real CPU frequency.

As an example, my current setup running WCG.



.

[KTE]

I'm busy away for a a few but I'll get back soon. Meanwhile I just quickly want to say that I've ran many tests and you're right Lightman. The core/mem speeds are still the same in both conditions, props.

[Lightman]

So I spent over 1 hour watching my stopwatch and rendering this scene at two different settings...
I've ran this calculation intense scene to eliminate any doubt.


NB1980

NB2640
Times reported by LW are correct!

KTE, your small software is showing as others do (except CPU-Z)...
But images above and all other tests I did point to conclusion that CPU-Z is correct this time, at least in regards of CPU clock.

Hope that now we all have less clear mind than before

BTW are you planning to get BlackEdition?
Also I heard from good source that AMD rev.B3 will officially reach 3GHz+ in 2H2008 plus 45nm chips will have lower latencies to caches


Merry Christmas to ALL!!

[KTE]

Yeah I did something similar, mine were 2860MHz NB and 1980MHz NB @ 2420MHz CPU and 1980MHz HT. Many many tests, real tests. The video decoding/encoding proved it that CPU/Mem/HT clocks are the same. CPU-Z doesn't detect properly. Try and validate it with high NB, it'll fail like this: http://valid.x86-secret.com/show_oc.php?id=286546

All other software is reporting correctly, it is reading everything from the core registers themselves as it should. However, the core itself is telling its frequency as low values for some reason and causing accurate software and OSes to all error. Meanwhile, the actual frequencies remain unchanged.

About the NB speed. There is only one software where I've seen any perf. difference with high NB readings: CPU-Z latencies, which I don't know if its synthetic (based on CPU-Z readings) or real-time. I will ask Franck about it. Sandra does not show NB speed gain whatsoever, neither anything other including all popular benchmarks be it synthetic or real-time and I ran them all that I could, even WCG benchmark. Performance was exactly the same, even in encoding/decoding, compression//decompression and encrpytion/decryption. In fact, decoding encoding 13 videos which took 15 minutes 56 seconds, took exactly the same time for both, absolutely same.

Keep up your tests. Yes I'll probably have BE soon, haven't contacted my uncle to ask if he has any yet and can't until Monday when he comes back. Only Intel can change that, if their quads hit in early Jan, because then I'll want to play with them first instead.

[bingo13]

I downloaded new BIOS F3 (Final) and it works fine!
I can enable Errata for K10 (default is DISABLE ) and it has many more improvements and very helpful things like showing effective freq. next to memory adjustment or NB multi. setting!

Big thanks to GBT for that BIOS!

We had asked Gigabyte to change the terminology on the settings to something that 90% of the users would understand. Also, memory performance is finally back up to par and NB speeds should be stable up to 2.2GHz. There are still a couple of rough spots with 2GB memory modules but that will be addressed in F4 along with additional usability options. This BIOS also works with Crossfire X.

[Lightman]

We had asked Gigabyte to change the terminology on the settings to something that 90% of the users would understand. Also, memory performance is finally back up to par and NB speeds should be stable up to 2.2GHz. There are still a couple of rough spots with 2GB memory modules but that will be addressed in F4 along with additional usability options. This BIOS also works with Crossfire X.

Thanks for clarification Binog13!

I have couple of questions to you ....

What those HT link options hidden in Advanced Chipset do? (IH Flow Control and the rest)

Have you cracked x64bit stability issue? Any news on that front?

Is NB voltage adjustment still in plans for future implementation?

Thanks for any hints

[linhvndiy]

Lightman: How to get that high NB?
I can do now only NB: 9x235 and cpu 10.5x235.
Can not set NB x10. And any higher Htt will make mainboard reset clock to df.
Many thanks. May be my cpu suck.

[Lightman]

Lightman: How to get that high NB?
I can do now only NB: 9x235 and cpu 10.5x235.
Can not set NB x10. And any higher Htt will make mainboard reset clock to df.
Many thanks. May be my cpu suck.

I'm doing this in two steps

1. Set HTT220-228 and NB x10, if it boots to Windows then OK
2. Back to BIOS and set desired HTT speed, should take off

My CPU either likes high NB speeds or simply BIOS is not setting multiplier properly, but on the other hand I can't BOOT if I'm trying too high. The only thing I'm getting then is looong beeeep

Also higher HTT = lower NB speeds possible for stable operational...

Have fun

[Augustus]

Guys, anyone using a 6400+ on this board?

I've just swapped out my POS K9A Plat for this and no matter what bios i try the highest multiplier reported is 15.5x.

Anyone else had this?

bingo13, any chance of asking your contacts what the reason for this is?

Cheers.

[BadNizze]

Guys, anyone using a 6400+ on this board?

I've just swapped out my POS K9A Plat for this and no matter what bios i try the highest multiplier reported is 15.5x.

Anyone else had this?

bingo13, any chance of asking your contacts what the reason for this is?

Cheers.

Did the same with my 6400+, leave it @ Auto for *16