New CPU OPNs (incl. C3 Stepping)

[Lightman]

The latest BKDG has some info about the C3 stepping. The chips will have two p-states for the nb and there will be a new C-state with the following characteristics. The L1/L2 caches will be flused to L3 or memory the cores will be in C1E state and running at ALTVID. Also there will be support for low voltage (1.35V) DDR3 1066.

http://support.amd.com/de/Processor_TechDocs/31116.pdf

p32 for the features
p48 2.4.2.2 NB p-states
p64 2.4.4.3 C5

That sounds interesting ...
This mean that finally NB will save the power as well. P-State 1 must be half the freq. of P-State 0 but any set vNB. This is good improvement.
Also they mention registers responsible for NB multis, but it is too early in the morning for me to figure out if reboot is still required to apply changes.

[Milans]

4ghz at standard voltage.. I dont say no thanks for that.

no, CPUz shows Core VID = 1.35 V. Real core voltage can be different.

@charged3800z24
Please ask him for a online valid.

[FlanK3r]

no, CPUz shows Core VID = 1.35 V. Real core voltage can be different.

@charged3800z24
Please ask him for a online valid.

charged said: He said that was actually bootable. It is Windows Vista 32bit. Voltage was correct and set in bios. I asked him for some benches at that speed, but he won't share them with me. Is there even a NDA on the new revion chips? I haven't really heard anything major about them. There has to be someon on here that has one or knows more.
-->really only 1.35V boot from BIOS

[justapost]

That sounds interesting ...
This mean that finally NB will save the power as well. P-State 1 must be half the freq. of P-State 0 but any set vNB. This is good improvement.
Also they mention registers responsible for NB multis, but it is too early in the morning for me to figure out if reboot is still required to apply changes.

That sounds like the FID can not change without an reboot and is picked up from an pci register. If the GfxNbPstateDis bit is not set and all cores have an NBDID of 1 in the current active p-state the NB runs at half the default speed. NBVID changes always worked so the voltage can be lowered when running at half the speed.

[DedEmbryonicCe1]

The latest BKDG has some info about the C3 stepping. The chips will have two p-states for the nb and there will be a new C-state with the following characteristics. The L1/L2 caches will be flused to L3 or memory the cores will be in C1E state and running at ALTVID. Also there will be support for low voltage (1.35V) DDR3 1066.

http://support.amd.com/de/Processor_TechDocs/31116.pdf

p32 for the features
p48 2.4.2.2 NB p-states
p64 2.4.4.3 C5

NB P-state 1 is always half the frequency of NB P-state 0 and is specified
by MSRC001_00[68:64][NbDid]=1 for any enabled P-state.

I wish the NB FID were controllable instead but I'll take this as progress.

All cores must drop down to a P-state that specifies the NB DID set at 1 for the NB to drop into P-state 1 (NB of 2000mhz stock becomes 1000mhz at P-state 1).

The frequency relationship of (core COF / NB COF) <= 2 must be maintained for all supported P-state combinations.

P0 CPU 3000mhz / NB 2000mhz VALID
P3 CPU 800mhz / NB 1000mhz VALID
P2 CPU 2200mhz / NB 1000mhz INVALID

All core P-states are required to be defined such that (NB COF/core COF) <= 32, for all NB/core P-state
combinations. E.g., if the NB COF is 4.8 GHz then the core COF must be no less than 150 MHz.

This won't be an issue...

All core P-states must be defined such that:
• CPU COF >= 400Mhz.
• NB COF >= 2 * MEMCLK frequency. E.g., for DDR1333 support, NB COF must be 1.4 GHz or higher.
(MEMCLK would be 667 MHz).Note if NB P-states are enabled, an NB P-state 0 of 2.8 GHz or higher is
required to provide an NB P-state 1 of 1.4 GHz.
• NB COF >= 600MHz.

For DDR3 1600mhz (MEMCLK 800mhz) you would need to have a NB P-state 0 of >= 3200mhz with NB P-states enabled to get P-state 1 >= 1600mhz!

[edit]
Too slow.

[justapost]

I wish the NB FID were controllable instead but I'll take this as progress.

All cores must drop down to a P-state that specifies the NB DID set at 1 for the NB to drop into P-state 1 (NB of 2000mhz stock becomes 1000mhz at P-state 1).


P0 CPU 3000mhz / NB 2000mhz VALID
P3 CPU 800mhz / NB 1000mhz VALID
P2 CPU 2200mhz / NB 1000mhz INVALID


This won't be an issue...


For DDR3 1600mhz (MEMCLK 800mhz) you would need to have a NB P-state 0 of >= 3200mhz with NB P-states enabled to get P-state 1 >= 1600mhz!

[edit]
Too slow.

Yep it would be cool to keep the NB speed always one multi below the corespeed. I wonder if we can expect higher nb speeds now,I mean they mention 2.8GHz. If not these two p-states will be usefull only for the powersaving oriented people. With low voltage DDR3 mem at 1066MHz 2.2GHz NB speed would be enough. And with 1.1GHz for p-state-1 and VID of 0.8-0.9V should be good to go and save a few watts in idle. Together with the new C5 state the gap to i5 in idle consumption will be much lower.
On my 250 it saved ~3W going from 2GHz/1.1V to 1.6GHz/0.9V and that without an L3 cache.

[charged3800z24]

no, CPUz shows Core VID = 1.35 V. Real core voltage can be different.

@charged3800z24
Please ask him for a online valid.

sorry been gone all day. He won't validate or show me any benches yet. He won't even let me tell you where he works. He annoys me.

[chew*]

sorry been gone all day. He won't validate or show me any benches yet. He won't even let me tell you where he works. He annoys me.

No offense charged but if it is real and he is legit the shoe could be on the other foot.........He could be under NDA and you could get him in trouble and you asking him for valids could be annoying him not to mention getting him in trouble..

ON a side note an OEM box PC with adjustable timings is laughable. Timings are way to tight for an OEM HP.

[SocketMan]

rocessor AMD Phenom™ II X2
Model B53
OPN Tray HDXB53WFK2DGM
OPN PIB
Operating Mode 32 Bit Yes
Operating Mode 64 Bit Yes
Revision C3
Core Speed (MHz) 2800
Voltages 0.80-1.425V
Max Temps (C) 70
Wattage 80 W
Virtualization Yes
L1 Cache Size (KB) 128
L1 Cache Count 2
L2 Cache Size (KB) 512
L2 Cache Count 2
L3 Cache Size (KB) 6144
CMOS 45nm SOI
Socket AM3
AMD Business Class Yes
Black Edition No

http://products.amd.com/en-us/Deskto...False&f12=True

AMD says it is a C3,why would it not be

[DedEmbryonicCe1]

http://xbitlabs.com/news/cpu/display...rocessors.html

The exact launch date of the new commercial platform from AMD is not known at the moment. What is known is that the company plans to launch the new desktop processors in the third quarter and that in Q1 2010 it intends to introduce higher-performance business class models.

I hope it is true that these C3's are so close to hitting the shelves.

[Daveburt714]

I hope it is true that these C3's are so close to hitting the shelves.

There are still 10 of us waiting for Simon to send us some Stepping 3 Phenoms.. (I hope, I hope... )

http://www.xtremesystems.org/forums/...1&postcount=91

Whole Thread:
http://www.xtremesystems.org/forums/...d.php?t=233236

[demonkevy666]

No offense charged but if it is real and he is legit the shoe could be on the other foot.........He could be under NDA and you could get him in trouble and you asking him for valids could be annoying him not to mention getting him in trouble..

ON a side note an OEM box PC with adjustable timings is laughable. Timings are way to tight for an OEM HP.

who puts NDA on steppings

this just sounds silly, but anyways he said it's an HP machine. He can tell us cpu, but not much else cause HP NDA...?
it is a new machine to replace old ones in there product line.

[chew*]

who puts NDA on steppings

this just sounds silly, but anyways he said it's an HP machine. He can tell us cpu, but not much else cause HP NDA...?
it is a new machine to replace old ones in there product line.

Heh you don't find it funny that the guy will not offer much info.........Also HP machine with 290 HT bus clocking.....seriously...

[demonkevy666]

Heh you don't find it funny that the guy will not offer much info.........Also HP machine with 290 HT bus clocking.....seriously...

we don't know what boards in it or what memory.
what we know is it's just a C3 phenom II chip.
I think the stepping under NDA is just silly, like it's a holy grail for them. lol

[EnJoY]

I think the stepping under NDA is just silly, like it's a holy grail for them. lol

Maybe it is? Maybe it's significant enough to warrant keeping under wraps for the time being? Maybe it's not? When you don't know, you don't know.

[ridney]

There are still 10 of us waiting for Simon to send us some Stepping 3 Phenoms.. (I hope, I hope... )

[FlanK3r]

...I know Phenoms III will revision D

[charged3800z24]

No offense charged but if it is real and he is legit the shoe could be on the other foot.........He could be under NDA and you could get him in trouble and you asking him for valids could be annoying him not to mention getting him in trouble..

ON a side note an OEM box PC with adjustable timings is laughable. Timings are way to tight for an OEM HP.

I perfectly understand, I just mean he is anoying to me in general. He just acts all weird an all, and not just with this info. He gave me my 940 back before launch and I was hoping he'd give me a test drive with this one. Back then I had to keep the system at our shop though, and he had to be there. After launch he let me keep it. He did say it was not in the HP box. But won't say what board.

[DedEmbryonicCe1]

The rest of the new Business Class processors have appeared at AMD.com and only the previously mentioned B53 is listed as a C3. However, the B55 OPN ends with the same 'GM' like the B53 and the other new processors in my OP. The XBit article I linked above mentions the B5x/B7x/B9x as having "enhanced idle power", which I thought could be the new NB P-state. Looking at the specs though it may be a reference to the lower idle voltage, which is <=0.80v.
95W: B95 0.80-1.425V / 945 0.85-1.425V
80W: B55 0.775-1.425V / 545 0.875-1.425V

[DedEmbryonicCe1]

http://h71016.www7.hp.com/dstore/Mid...05_Pro_MT_CFG9

The B53 and B55 are selectable and this new model lists the ship date as 9-28. Hopefully someone with access will post some information next week to confirm/deny these are C3's.

[demonkevy666]

control on the Nb for low power usage has been around for while it just hasn't been implemented.
in fact an old post by KTE mentions this very thing.

[uncle john]

Shouldn't we assume that Shanghai would be first with new stepping?

[Gurr]

Does anyone know when c3 will be making a show? I think that is when I will be upgrading from phenom 1, been waiting for a low power NB p state to show up. [I do like to conserve power when I'm just searchin the web]

[uncle john]

http://www.fudzilla.com/content/view/15557/35/

[demonkevy666]

C3 is suppose to last threw 2010 must be some pretty good performance pre-watt.