Undisclosed AMD Processor - Some results that AREN'T with extreme cooling...

[justapost]

Watch the movie again and pay attention to CPU clock speed close to the end of it

@Linus@ncix:

@justapost:

Did you see the processor clock, changing from 3612.3 to .2 to .1 (as far as I remember there was a 3611.7 as well ) ?

Thanks I missed that. But I can not see if prime95 is running or a background image. Nothing personal linux, just questioning the proove.

[chew*]

Thanks I missed that. But I can not see if prime95 is running or a background image. Nothing personal linux, just questioning the proove.

It's running, trust me, the reason its placed where it is is because it also identifies cpu strings. He went above and beyond what was requested IMHO, I would have been happy with just the screenie.

[LinusTech]

The truth will be apparent in due time. I just throught it would be a great first topic for my new video blog

Thanks for all the postitive feedback everyone.

[PetNorth]

What happened with vid?

"We're sorry, this video is no longer available."



EDIT: nah, now works Thanks Linus!

[informal]

The truth will be apparent in due time. I just throught it would be a great first topic for my new video blog

Thanks for all the postitive feedback everyone.

Linus,what do you think about OCing headroom on your particular chip,how much from it could you possibly get with more tweaking?

[Glow9]

Great video, besides getting it to 3.6Ghz any benches been made available?

[LinusTech]

No benches unfortunately, but that shouldn't be too much of a mystery. It's a Phenom without the TLB bug. It has some more cache, and some tweaks. I'd say probably within 2-10% of original Phenom clock for clock.

As for more headroom, I don't know how good the compatibility of this board is with the mystery chip. It recognizes it correctly, but it takes a long time to POST. I think with a better board/better cooling there has got to be more headroom in here. Like I said before though, this was a VERY quick & dirty overclock and I don't even have the chip/board anymore.

I know you guys are foaming at the mouth for this stuff, but I'm lucky I was able to post what I did.

It may still come back and bite me in the butt!

[Particle]

People are still trying to question your trustworthiness...let 'em wiggle. If I were on your side of it, I'd be pretty offended by this point, Linus.

[informal]

I don't know why people try to question the validity of Linus' post or video.Maybe with the last fake "blog" results some are being just careful.

Anyhow,big thanks to Linus who made the video !

[George_o/c]

No benches unfortunately, but that shouldn't be too much of a mystery. It's a Phenom without the TLB bug. It has some more cache, and some tweaks. I'd say probably within 2-10% of original Phenom clock for clock.

As for more headroom, I don't know how good the compatibility of this board is with the mystery chip. It recognizes it correctly, but it takes a long time to POST. I think with a better board/better cooling there has got to be more headroom in here. Like I said before though, this was a VERY quick & dirty overclock and I don't even have the chip/board anymore.

I know you guys are foaming at the mouth for this stuff, but I'm lucky I was able to post what I did.

It may still come back and bite me in the butt!

Now what's that TLB bug ?

[EvilBlitz]

Particle as to your response about the TDP ratings. You will notice we said AM2+ and AM3, they are both desktop chips so it is nothing to do with that.
The server socket is different, though I prolly expect some 1P AM3 chips eventually.
There is still a difference between the AM2+ and the AM3 desktop chips rating ACP/TDP at the same speed. The only things I can think of is 1) better stepping for AM3s or 2) In DDR3 mode the IMC uses less power? or 3) Or higher power on the AM2+ chips because they have to support a single power plane and work in AM2 mobos. All seem a little odd.

[LinusTech]

TLB bug was a problem with the very first release of Phenoms. It was fixed, but it dropped performance by about 5-15%.

The current Phenoms and (as reason would dictate) any future AMD quad cores wouldn't be affected by this problem.

[prostoman]

Not trying to be negative here or upset anyone, but in that vid, where he blacked out voltage, it looks like 1,5, u can see a little bit of 5 sticks out of the tape, no other number besides 7 fits the bill. I am just guessing here, dont want to get anyone in trouble.

[informal]

Not trying to be negative here or upset anyone, but in that vid, where he blacked out voltage, it looks like 1,5, u can see a little bit of 5 sticks out of the tape, no other number besides 7 fits the bill. I am just guessing here, dont want to get anyone in trouble.

Yeah it's 5 it seems.But hey,since def. vcore for Deneb/Shanghai is ~1.35V ,even 1.55V is ~15% above stock ,so no biggie if you have decent air cooling .

[justapost]

Since you all trust linus, I think i have no point in doubting him. Have CPU temps during the prime95 run been mentioned?

[Particle]

Particle as to your response about the TDP ratings. You will notice we said AM2+ and AM3, they are both desktop chips so it is nothing to do with that.
The server socket is different, though I prolly expect some 1P AM3 chips eventually.
There is still a difference between the AM2+ and the AM3 desktop chips rating ACP/TDP at the same speed. The only things I can think of is 1) better stepping for AM3s or 2) In DDR3 mode the IMC uses less power? or 3) Or higher power on the AM2+ chips because they have to support a single power plane and work in AM2 mobos. All seem a little odd.

I think you're still missing the point of my post there--no desktop chip is rated as ACP. Only their server chips are ACP, so there's no reason to compare TDP and ACP on desktop chips since all ratings given are TDP.

[charged3800z24]

I'm personally going for the GA-MA790GP-DS4H..

It's got AMD 790GX + SB750 Chipset and plenty of MOSFET-cooling. Consider this; It has worlds fastest onboard graphic's, the HD3300, and this is cooled by a heatsink connected to the MOSFETS. Now disable the onboard HD3300, and plug in whatever card you like in the x16 slot, and voila! One extra massive heatsink for you'r MOSFET's only

Besides.. this was actually the same board (and theres only 1 revision) that was used to clock the AM2+ PII to the amazing 6+ GHz

The board is quite cheap as well.

EDIT: Looks like this http://www.gigabyte.com.tw/Products/...ProductID=2887

I use that board.. been good..had some ACC problems, but nice cooling on the fets.. they never fried while running 1.65v's when trying to get 3.6ghz running. I ran 3.4ghz 1.47v on the 9950 with no problems.

[charged3800z24]

CPU-Z says "Agena"....? is CPU-z messedup......
did that say CPU-z 1.41...? either way it say stepping 2...I never saw a B2 Phenom clock that high ...lol

[LinusTech]

CPU-Z is an old version and did not detect a lot of things correctly about the CPU.

The frequency is no problem though

[justapost]

Linus have you had ACC enabled for that oc? And if it does not break NDA did it run at 1.35V at stock clocks?

[RiverRicer]

CPU-Z is an old version and did not detect a lot of things correctly about the CPU.

Must be a really old version as Deneb/Shanghai support was added 5/08 in v1.45.
The lastest version (1.48) has added dual core (Kuma) support, so maybe we'll actually see a K10(K10.5) dual!

[Particle]

I wouldn't worry about the Agena string. No Agena has a 6MB L3 cache, and that's what CPU-Z is showing.

[ZL1Killa]

just waiting here for this thing to come out, I did 3.54ghz on my 6400+be stable on my WC kit, I would hope we are going to see 4ghz+ as these other threads have shown, that would be awesome! and apparently this ASUS m3A79-T board is a really nice board, i see a lot of people using it. I'm waiting for am3, deneb, and mobos with ddr3

[charged3800z24]

I wouldn't worry about the Agena string. No Agena has a 6MB L3 cache, and that's what CPU-Z is showing.

YA, that and I don't think a B2 Phenom would hit that frequency, even on LN2..