Sign in with
Sign up | Sign in
Your question

AMD has an ace up its sleeve - magic 6 pins

Last response: in CPUs
Share
June 6, 2008 4:17:28 AM

http://www.tgdaily.com/content/view/37816/135/

AMD smoke amd mirrors last year it was the "true" quad this year its a magic 6 pins that mysteriously over clock the cpu! Is it a new method of sb clocking? of coarse not!

Well is there anything to the 6 unknown pins?

sounds like some kind of noobie oc for masses - buy a black get a blacker?
June 6, 2008 5:37:31 AM

Well now, instead of running hotter and slower than a Q6600 AMD owners can spend some more cash and then run at a similar speed and much high heat disipation....

Wait they'll need to upgrade their cooling as well... He he... Starting to make the Q6600 look like a much cheaper/cooler/faster solution for decent quad core performance.

At least with this AMD's tri-cores can now hope to out run Intel's E7200...
June 6, 2008 6:53:05 AM

depends on the overclocking, you have to remember there all the same chip, if they overclock with the multiplyer there not gonna increase heat dissipation by much. what worrys me is all the people that have already overclocked there systems.. wont it screw them up?
Related resources
Can't find your answer ? Ask !
June 6, 2008 7:15:31 AM

Flakes, actually a Q6600 at say 400x8 does definately generate less heat than the same chip at 400x9 and yes it probably will their screw them over or just override their settings.
June 6, 2008 9:54:32 AM

I don't believe a word of it.


If I recall correctly, it was Theo Valich that said AMD could double their hypertransport speeds on the K8 or something, effectively doubling their performance overnight to compete with Conroe.
June 6, 2008 10:17:30 AM

JDocs said:
Flakes, actually a Q6600 at say 400x8 does definately generate less heat than the same chip at 400x9 and yes it probably will their screw them over or just override their settings.

You are comparing the heat dissipation of a chip at 3.2ghz and a chip at 3.6ghz...
What he's saying is compare how hot a chip @ 400x9 is compared to 450x8.
June 6, 2008 10:22:57 AM

450*8 should be warmer (as a whole through out the computer) as it places more stress on the motherboard and depending on the the RAM speed is set (in most cases with a devider) the RAM as well generating more heat on those components. Additionally the higher FSBs with lower multipliers generally generate more heat from than the same speed acheived via a lower FSB and higher multi, well from what I understand.
a b à CPUs
June 6, 2008 12:26:19 PM

Magic 6 Pins... Magic 8 Ball...

<Signs Point To No>
a c 127 à CPUs
a b À AMD
June 6, 2008 2:02:11 PM

JDocs said:
Well now, instead of running hotter and slower than a Q6600 AMD owners can spend some more cash and then run at a similar speed and much high heat disipation....

Wait they'll need to upgrade their cooling as well... He he... Starting to make the Q6600 look like a much cheaper/cooler/faster solution for decent quad core performance.

At least with this AMD's tri-cores can now hope to out run Intel's E7200...


Dude the Q6600 always has. But I remember reading a few reviews of the Phenom BEs that were being OC'ed and they would hit a certain limit depended on the processor) where their power consumption would go up by quite a bit.

Personally I think its crap. It may be true but I think its not. Although there are a few pins that you can do something with on a Q6600 to have it run on a 1333FSB stock.
a b à CPUs
a b À AMD
June 6, 2008 3:31:24 PM

A boost in performance simply by pressing a button [:mousemonkey:2] , now where have I seen that before?
June 6, 2008 9:03:00 PM

maybe i missunderstood this 6pin thing

i thought this was just fast oc - so its actually moving part of the hypertrasport to the sb?

all my sarcasim aside - i really do not understand what the 6 pins do - hence my thread magic 6 pins.

but know i think they are passing part of the data through the sb to kinda like hyper threading?
June 7, 2008 1:01:24 AM

dragonsprayer said:

but know i think they are passing part of the data through the sb to kinda like hyper threading?

Huh? What has Hyperthreading to do with the Southbridge?
June 8, 2008 1:46:43 AM

sorry

i think the amd systems works by splitting the data flow and running part through the nb and part through the sb

the parts are reassembled in the cpu (hun?)

this way if the memory controller of nb are a problem they can by pass it hence the extra pins
June 8, 2008 2:34:06 AM

dragonsprayer said:
sorry

i think the amd systems works by splitting the data flow and running part through the nb and part through the sb

the parts are reassembled in the cpu (hun?)

this way if the memory controller of nb are a problem they can by pass it hence the extra pins

Umm, no.
First off, all instructions to the south bridge, are routed through the north bridge.
In case you haven't heard, AMD has an On Die Memory Controller. This ODMC has nothing to do with the nb.
Oh, and by the way, if you are going to read "articles" by Theo, I recommend you invest in a very large bag of salt.
June 8, 2008 3:11:51 AM

I don't think this is honestly true, but it might be only time can tell.
a c 113 à CPUs
a b À AMD
June 8, 2008 2:17:55 PM

bump
June 9, 2008 3:22:37 PM

Mousemonkey said:
A boost in performance simply by pressing a button [:mousemonkey:2] , now where have I seen that before?


had not thought of it like that, turbo button here we come :) 
a c 127 à CPUs
a b À AMD
June 9, 2008 3:37:40 PM

endyen said:
Umm, no.
First off, all instructions to the south bridge, are routed through the north bridge.
In case you haven't heard, AMD has an On Die Memory Controller. This ODMC has nothing to do with the nb.
Oh, and by the way, if you are going to read "articles" by Theo, I recommend you invest in a very large bag of salt.


You must forgive him. He is now an AMD fan more than Intel and is pretty much a noob. So either way ne is new to the AMD arena. I think it was b/c he thinks Intel wont allow OCing for Nehalem but whatever.

Either way I still find this hard to believe.
June 9, 2008 5:38:39 PM

If AMD had an ace, they would have played it already.


Yay! Tom's is no longer going to base their articles off of those stupid things called "facts".

Welcome to the new Inquirer my friends.



REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1! REVERSE HYPERTHREADING!1!
June 9, 2008 5:41:49 PM

This shouldn't be that hard to implement, as long as you have the instruction written in the chipset, or simply supported by the BIOS.

EDIT: I'm not sure what Theo meant by "overclocking through the south bridge". AFAIK, south bridge operates independently to processor. Sounds to me more like a platform rewarding system: if you purchase AMD's latest SB, they reward you by giving you a significant automatic overclock.

EDIT2: Actually if you continue to investigate Theo's report, the likelihood of such implementation is getting smaller. Agena 9850 will likely dissipate over 200W of heat at 3.2Ghz. Under that circumstance, can the stock cooler hold it back? On the other hand, if AMD can easily overclock even the Tolliman to 3.2Ghz, why haven't they done so, and selling them for higher ASP?
a c 127 à CPUs
a b À AMD
June 9, 2008 5:54:45 PM

Um its AMD. They have a lot of great ideas but the worst implimentation and execution I have ever seen.
June 9, 2008 6:20:36 PM

jimmysmitty said:
Um its AMD. They have a lot of great ideas but the worst implimentation and execution I have ever seen.



Geesh

If the Phenom didnt have such a crappy start and now they say its got special talents then its like saying the 939 has got six extra pins for overclocking...

Who cares...

AMD lost this race with a blind donkey with one leg and a crusted saddle...

Its like saying we gave this donkey an injection that will make it last a few more days.....

I wish that AMD would stop this bs poop and bring us a chip we all deserve from someone who innovated the pc...

I was gonna get a 4870x2 but not im gonna get a new Nvidia card when it comes out...

Ive given up now.. really given up

Im gonna go for the black helmet number, become a sith lord and join the dark side...

Im gonna jump ship so far that there is no comming back as the rubber dingy came with it...

9450 Quad, 8GB and quad hard disk boot here i come....
a c 113 à CPUs
a b À AMD
June 9, 2008 10:54:38 PM

It ain't magic.

Microprocessor pin assignments include bidirectional data bus lines, IRQ requests, power, etc.

Most importantly in this case microprocessor pin assignments include clock signal input/output.

The higher the clock rate the more difficult it becomes to keep the clock signal in phase. Of course the south bridge is important to overclocking. South bridge phase lock loop timing factors are most likely the reason the Phenoms are having OC problems.

The best way to keep in phase with an OC is to bump the volts when you hit the wall.

If the SB750 fixes the PLL issues it is not beyond reason that a core clock of 200MHz with a 16x multi will do 3.2GHz at stock volts.
June 10, 2008 2:06:03 AM

Wisecracker said:

If the SB750 fixes the PLL issues it is not beyond reason that a core clock of 200MHz with a 16x multi will do 3.2GHz at stock volts.


The problem is, according to Theo, all 8000 series Tolliman and 9000 series Phenom can benefit from this. Given that most of them are multiplier locked, the only way is to increase core clock.
June 10, 2008 6:44:04 AM

TechnologyCoordinator said:
YES! GREAT FIND!!

REVERSE HYPERTHREADING!!! WOOT! AMD4LIFE!


http://www.warrenuncensored.com/images/THG/Theo.png



All Women Treat you self to AMD... ( All Mothers and Daughters )

Because you deserve it.............


Remember .. Women.... Know your limits - AMD4WIFE



June 10, 2008 7:04:10 AM

jimmysmitty said:
You must forgive him. He is now an AMD fan more than Intel and is pretty much a noob. So either way ne is new to the AMD arena. I think it was b/c he thinks Intel wont allow OCing for Nehalem but whatever.

Either way I still find this hard to believe.

I doubt the AMD croud would want him on thier side.
June 10, 2008 10:22:02 PM

joefriday said:
Here's a look at anther one of Theo's memorable FUD articles:
http://www.theinquirer.net/en/inquirer/news/2006/06/23/...

Everytime i read this i am more and more shocked. Even though i know it wasn't written on the first of April i keep checking the date on the article. This is so bad.
June 11, 2008 7:09:58 AM

TechnologyCoordinator said:
YES! GREAT FIND!!

REVERSE HYPERTHREADING!!! WOOT! AMD4LIFE!


http://www.warrenuncensored.com/images/THG/Theo.png



Could someone remind Theo. Theodore - wasnt that a name of a Chipmunk... Boom Chicca Wow wow


Anyway - Im still waiting for my AMD 2 940 3800 - or was it + to beat the 6300 Core Duo ( which wasnt intels top chip and was never claimed to be ).

Im sitting here waiting for it to work faster, but it never has..

Theo you are worthless.. How much did AMD pay you for that one...

Hang on Thunderman - Theo are they the same person...


Anyway Reverse Hyperthreading works, AMD has been going backwards since....


June 11, 2008 10:31:16 AM

Well, I'll wait and see if it works. 200 megahertz isn't unreasonable for a 9850, let alone an 8750. I don't think it will work on existing quads but rather the upcoming 95 watt BE versions of the 9850. I'd be happy if it worked with the existing 8750.

a b à CPUs
June 11, 2008 1:05:57 PM

If you think Theo is thunderman then your a n00b and haven't been here long enough to warrant a nappy change yet ... sorry ... I meant "muck out the pen".

The hacks don't post in this area ... your unworthy.
a c 113 à CPUs
a b À AMD
June 11, 2008 3:26:40 PM

yomamafor1 said:
The problem is, according to Theo, all 8000 series Tolliman and 9000 series Phenom can benefit from this. Given that most of them are multiplier locked, the only way is to increase core clock.


:) 

Sorry for not making that more clear. Musta been da voodoo in da magic 6 pins.

Fixing the PLL and the SB is the big AMD dawg bone going forward - any backward compatibility (like with the BEs?) is just gravy ...

I don't know if a multi can be unlocked thru a combination of model-specific cpu registers, pin assignments & BIOS. Sounds reasonable but it's way above my pay grade ...
June 11, 2008 10:25:05 PM

Reynod said:
If you think Theo is thunderman then your a n00b and haven't been here long enough to warrant a nappy change yet ... sorry ... I meant "muck out the pen".

The hacks don't post in this area ... your unworthy.



What you on about fanboi

I was making an assumption.. Thundermans rants are simular to Theo.. Unfounded and down right stupid.


My assumption now is your backing a loosing horse..

Carry on sunshine - let the horse loose..

The company is doing this on its own accord anyway, Whats the matter wanting the dual core they just wasted developement time on...
a c 100 à CPUs
June 12, 2008 3:15:40 AM

Wisecracker said:
:) 

Sorry for not making that more clear. Musta been da voodoo in da magic 6 pins.

Fixing the PLL and the SB is the big AMD dawg bone going forward - any backward compatibility (like with the BEs?) is just gravy ...

I don't know if a multi can be unlocked thru a combination of model-specific cpu registers, pin assignments & BIOS. Sounds reasonable but it's way above my pay grade ...


There was a thread somewhere (EOC?) that tried to figure this out. The consensus was that the multiplier is set in microcode/internal MSRs and you need a JTAG and AMD's CPU microcode-setting tool to change it. The CPU just ignores things put in visible MSRs that do not mesh with microcode- try to undervolt a desktop K8 with RMclock or cpupw below 1.100 V and see the CPU ignore it. There are also no pencil tricks or "golden fingers" either.
June 12, 2008 5:42:43 AM

MU_Engineer said:
There was a thread somewhere (EOC?) that tried to figure this out. The consensus was that the multiplier is set in microcode/internal MSRs and you need a JTAG and AMD's CPU microcode-setting tool to change it..


I'm just wondering if it needs a new board, like the upcoming 880G's; or if it will work with an existing 8750 in a GIGABYTE GA-MA78GM-S2H. I'd be happy to have an 8750 @ 2.6 instead of 2.4. Dratted vacation, can't buy the CPU and new memory until afterwards!

At any rate, it's impact on Deneb will be interesting; but Nehalem will win hands down -- especially if Intel actually comes out with $200 or so parts to compete against AMD in the mainstream.
a b à CPUs
June 12, 2008 8:58:47 AM

Harry-Plopper said:
What you on about fanboi

I was making an assumption.. Thundermans rants are simular to Theo.. Unfounded and down right stupid.


My assumption now is your backing a loosing horse..

Carry on sunshine - let the horse loose..

The company is doing this on its own accord anyway, Whats the matter wanting the dual core they just wasted developement time on...


Hmmm ...

1. Who are you to compare Theo ? He wouldn't waste his time wiping his a$$ on your face. He has a job as a hack ... your just a "dumb reader".
2. What horse did I back?? Meathead ... I don't recall making a bet on any race?
3. Sunshine ... so you were once an army recruit were you?? Get tortured a bit eh? Fail the physical and got booted for being too fat?

4. This is old news. We knew they wouldn't produce it ... or they would have done a 2 X 1Mb mask for the AM2's ... and they didn't then. They don't have the resources ... unlike Intel which have a dozen projects going at once.

If your going to make nasty comments about others expect to cop some flack yourself buddy ... especially me as I take no crap from n00bs who make sweeping comments without either something funny or silly ... or heaven forbid "enlightening" to add !!

I honestly don't think they finished the dual core ... architecturally It could not have been a great proposition anyway, has less IPC than the Core2, and no distinct advantage over the low power K8's - Turiun X2 Ultramobile.

Plus their process can't push the 65nm past 3.2 on the 90nm node ... the 65nm process is worse.

I imagine the 45nm process isn't going well either.

So that makes me a real fanboi.



a c 113 à CPUs
a b À AMD
April 23, 2009 12:15:01 AM

I just bumped this for doomsdaydave11.

Theo did in fact get it wrong. The sb700 was only a minor improvement over sb600 - sb750 was where AMD started the long crawl back ....
!