Lets talk breaking the HTT 230 limit on k10 here

[Tony]

Tictac may have our answer:

NB Register = Bus#0 Device#24 Function#3 Offset D4h (32bit access mode) , Bit 0,1,2,3,4 (5bit)

NB clock = 200 x (NbFid + 4).
*NBFid = in decimal

Bit = Hex = Decimal = NB Speed calculation
00001b = 1h = 1 = NB clock 200x (1+4) = 1000MHz
00010b = 2h = 2 = NB clock 200x (2+4) = 1200MHz
00011b = 3h = 3 = NB clock 200x (3+4) = 1400MHz
00100b = 4h = 4 = NB clock 200x (4+4) = 1600MHz
00101b = 5h = 5 = NB clock 200x (5+4) = 1800MHz
00111b = 6h = 6 = NB clock 200x (6+4) = 2000MHz
....
11011 = 1Bh = 27 = NB clock 200x (27+4) = 6200MHz

So i may be out by 1 in my original tables but thats easy to sort out

[Brother Esau]

Sorry wasn't aware of time out by you Tony NP just funnin anyways and tomorrow is fine of course! Email.......datauplink@npgcable.com

[Tony]

i will link it up here for all m8....no special treatment

[Brother Esau]

Yah ....keep up the great work as usual Tony I look forward to whats going on with this board for sure

[Unoid]

So Ideally we want to crank NB clock high as possible.. at least always matching Core clock (for full L3 speed)

Why not benchmark a phenom at 2.5ghz for example with a 2.5ghz NB clock (x) with the y value being HTT Speeds? say vary from the base 200 and go up by 10 cycles or so.

Find out if theres a bottleneck at all. If not no sense trying to max htt to a certain speed.

I'd still like to see someone run some benchies showing the difference in these settings

[Brother Esau]

Here ya go and more to come Not Tony's settings as I don't have those options in Bios yet and need bios update first. Wondering if what you mention in this thread will have any impact on Dual Core at all Tony?

[macci]

Brother Esau,

You dont see the AM2+ PState options there becuase you are running with K8 CPU.

K8 has the "NPT multiplier" controls (posted on your picture). Phenom has totally different options for PStates. Thats why your BIOS menus look different.

[Brother Esau]

Macci's results, let break them down.

NB core speed 2500MHZ, NB Multi 10X, HT speed 250MHZ

Bios values.

NB FID HEX 06 =10
NB DID 01

So HTT speed =250x10 for NB core clock of 2500MHZ

Moving to CPU core.

CPU core speed 3000MHZ, multi 12, HTT 250MHZ

CPU FID HEX 08 = 12
CPU DID 01 = 1 so NO change to CPU FID as 12/1=12

CPU core speed 250HTT x 12 for 3000MHZ

Moving to HT bus speed.Approx 2500MHZ

HT multi X HTT= HT speed
10 x 250HTT=2500MHZ

Options for HT multi:

200=1
400=2
600=3
800=4
1000=5
1200=6
1400=7
1600=8
1800=9
2000=10
2200=11
2400=12
2600=13



This is how you would have set bios up to hit the speeds shown in Macci's bench

Tony does AMD Overdrive work for you on the Bios on you're Board? As it stands on 11/22/07 Bios its a worthless pile of Random Lock Up even just opening it at most times let alone even using it!

[Brother Esau]

Brother Esau,

You dont see the AM2+ PState options there becuase you are running with K8 CPU.

K8 has the "NPT multiplier" controls (posted on your picture). Phenom has totally different options for PStates. Thats why your BIOS menus look different.

Really Bios options are different with Phenom Don't even see it greyed out even.

[macci]

it detects your CPU and gives the options that are available for it

[Brother Esau]

Well thats pretty trick now aint it

[STEvil]

pretty standard actually.

[kyosen]

My results...just with air or water cooling.

CineBench10 @ 2.9G=253x11, with NB x8:


also SP1M @ 2.9G=253x11, with NB x8:


SP1M @ 2.9G=265x11, with NB x7...yeah, it's slower than above:


I couldn't reach 3G...SP1M @ 2952MHz:


As macci already mentioned in another thread,
"HT Link multiplier = x6" looks suitable for higher HT base clock overclocking.

In my experience, Windows XP 64bit can't boot up at higher HT base clock than 225MHz.
WinXP 32bit is robust about this issue...

Also I tried preliminary dryice test, but Phenom looked heavily cold buggy...

[KTE]

HT/NB link speed make no difference to me, they are not the bottelnecks. Downclocking them makes no difference to my overclock potential. I've ran them at very low and high speeds to test. Overclocking the NB speed makes a v.good difference in my L2/L3 latencies. Most people are getting 250-270MHz on HT ref max.

I can boot 255 HT ref. pretty easily at stock volts on all things. Have used OD to 267HT max. 268HT froze the system.

If I overclock using P-States, then CPU-Z fails validation.

OD is too buggy for me. Sometimes just upon starting it just freezes my system.

BIOS 1.0, 1.1, P0C and P0D are quite bad. Theyt restrict my CPU HT ref. to 260HT.

BIOS 113 BETA allows 268HT max.

My CPU is slow above ~2.5GHz. Slower than at 2.46GHz. Going from 2.42GHz to 2.7GHz makes hardly any performance difference on BIOS 1.0/1.1/P0C. Not tried the rest with this yet.

[lukija]

Tictac may have our answer:

So i may be out by 1 in my original tables but thats easy to sort out

What about NbDid?

NB Speed = 200 MHz * [ (NbFid + 4) / ( 2 ^ NbDid)]

When NbDid is 0 Nb Speed = 200 MHz * (NbFid + 4)
When NbDid is 1 Nb Speed = 200 MHz * (NbFid + 4) / 2 <- Half multis?

This whole thing reminds me of good all days when K8 appeared. Only that time it was MemClk and famous Oskar Wu's memory ratio table

[KTE]

What about NbDid?

NB Speed = 200 MHz * [ (NbFid + 4) / ( 2 ^ NbDid)]

When NbDid is 0 Nb Speed = 200 MHz * (NbFid + 4)
When NbDid is 1 Nb Speed = 200 MHz * (NbFid + 4) / 2 <- Half multis?

Which BIOS/MB are you talking about here? This isn't how it works on mine (K9A2). Plus those values only work if you set the same or lower CPU multi than stock and leave NB multi to below 9x or it downclocks everything to adjust NB speed to around 2GHz.

Like mine here now (CPU-Z reads what it should be, but reports wrong values):




Also macci: setting 28 for CPU VID gives 1.2000V rather than 1.250V.

[lukija]

Which BIOS/MB are you talking about here? This isn't how it works on mine (K9A2). Plus those values only work if you set the same or lower CPU multi than stock and leave NB multi to below 9x or it downclocks everything to adjust NB speed to around 2GHz.

I guess this is how mobo makers implement multipliers. Actually final multiplier that you see in BIOS setup, AOD ..... is calculated using CpuFid and CpuDid (NbFid and NbDid). AMD BKDG explains some math how to calculate COF (calculated operating frequency). In short:

CPU Speed = 100 MHz * (CpuFid + 16) / (2 ^ CpuDid) or
200 Mhz * (CpuFid + 16) / [2*(2 ^ CpuDid)]

NB Speed = 200 MHz * (NbFid + 4) / (2 ^ NbDid)

^ - power functions: 2 ^ 0 = 1, 2 ^ 1 = 2, 2 ^ 2 = 4, 2 ^ 3 = 8 .....
200 MHz should be default reference clock
When you increase CpuDid and NbDid , final multiplier is lower 2,4,8,16 times ...

Please try this

http://www.xtremesystems.org/forums/...4&postcount=29

[Tony]

What about NbDid?

NB Speed = 200 MHz * [ (NbFid + 4) / ( 2 ^ NbDid)]

When NbDid is 0 Nb Speed = 200 MHz * (NbFid + 4)
When NbDid is 1 Nb Speed = 200 MHz * (NbFid + 4) / 2 <- Half multis?

This whole thing reminds me of good all days when K8 appeared. Only that time it was MemClk and famous Oskar Wu's memory ratio table

I mentioned there are 2 DID's 1 and 2, thing is i did not fully mention what they do...sorry.

Its now very easy to see these boards work in a whole different way with k10 , with K8 they are awesome clockers but you can not compare a K8 OC to a K10 OC on the same board

[Tony]

I guess this is how mobo makers implement multipliers. Actually final multiplier that you see in BIOS setup, AOD ..... is calculated using CpuFid and CpuDid (NbFid and NbDid). AMD BKDG explains some math how to calculate COF (calculated operating frequency). In short:

CPU Speed = 100 MHz * (CpuFid + 16) / (2 ^ CpuDid) or
200 Mhz * (CpuFid + 16) / [2*(2 ^ CpuDid)]

NB Speed = 200 MHz * (NbFid + 4) / (2 ^ NbDid)

^ - power functions: 2 ^ 0 = 1, 2 ^ 1 = 2, 2 ^ 2 = 4, 2 ^ 3 = 8 .....
200 MHz should be default reference clock
When you increase CpuDid and NbDid , final multiplier is lower 2,4,8,16 times ...

Please try this

http://www.xtremesystems.org/forums/...4&postcount=29

I changed the NB multi's...you were right, sorry i did not get back and thank you for an awesome tool that makes life that little easier.

[mAJORD]

Wow super pi takes a huge hit with the low NB clock

[Tony]

new beta bios from DFI, its passed internal testing but at time of posting I have not flashed it.

As usual test at your own risk...its beta

Db30 bios for DFI 790fx

[KTE]

I guess this is how mobo makers implement multipliers. Actually final multiplier that you see in BIOS setup, AOD ..... is calculated using CpuFid and CpuDid (NbFid and NbDid). AMD BKDG explains some math how to calculate COF (calculated operating frequency). In short:

CPU Speed = 100 MHz * (CpuFid + 16) / (2 ^ CpuDid) or
200 Mhz * (CpuFid + 16) / [2*(2 ^ CpuDid)]

NB Speed = 200 MHz * (NbFid + 4) / (2 ^ NbDid)

^ - power functions: 2 ^ 0 = 1, 2 ^ 1 = 2, 2 ^ 2 = 4, 2 ^ 3 = 8 .....
200 MHz should be default reference clock
When you increase CpuDid and NbDid , final multiplier is lower 2,4,8,16 times ...

Please try this

http://www.xtremesystems.org/forums/...4&postcount=29

lukija, thanks. You have the values mapped out incorrectly for my board according to my system. My board is MSI K9A2.

These are the settings I chose in AM2+ P-States:

I chose BIOS NB multi 12x (07), CPU multi 11x (06), HT multi 9x, HT ref. 218.
I chose NB DID 1 and CPU DID 1.
I chose NB VID 36 and CPU VID 28.

This is what your tool calculates and CPU-Z shows what it should have calculated:



You have NB DID and CPU DID values wrongly. There is no 0 value there, there is only 1 and 2 a maccci said. 1 stands for 1 and 2 for, you guessed it, 2. That's the end of the equation.

HT ref x (NB FID / NB DID) = NB speed

HT ref x (CPU FID / CPU DID) = CPU speed

NB DID options = 1 and 2
CPU DID options = 1, 2, 4, 8 and 16 (IIRC)

That's how my system is calculating speeds. NB speed is worked out exactly as your formula described but CPU speed is worked out exactly like that too.

Hope it helps.

[Lightman]

In my experience, Windows XP 64bit can't boot up at higher HT base clock than 225MHz.
WinXP 32bit is robust about this issue...

Also I tried preliminary dryice test, but Phenom looked heavily cold buggy...

Thats is in line with my findings!
But I managed to work it out up to 240HTT with lowered multi to 10x...

I think problem appears above 2.5GHz when logging to XP x64 desktop. Can you confirm that for me on your CPU?? I can boot to desktop even at 2550MHz but that is like lottery though!

[lukija]

KTE, i found these formulas in AMD documentation. I didn't invent them I just put them in program One thing also: CpuFid and NbFid are not the same as Cpu multi and Nb multi. Combination of CpuFid and CpuDid gives you Cpu multiplier. Same goes for NB.

Currently I'm at work, after I return home we will check it out using WPCREDIT

[knightwolf654]

you cant run it on dryice but will it run under chilled liquid? around -15 to -20?