Phenom 9600 Shows Up..

[JumpingJack]

For those waiting out on the 2.3 GHz model -- two sites are showing stock for the 9600:

http://www.compsource.com/ttechnote....X&vid=34&src=F

http://www.newegg.com/Product/Produc...82E16819103225

[informal]

Hi JumpingJack,thanks for those links.
Do you have any comment on this post written by Theo?

[Sparky]

lol @ newegg's misspelling quad "qaud"

[JumpingJack]

Hi JumpingJack,thanks for those links.
Do you have any comment on this post written by Theo?

Honestly, not much ... First, it is hard to understand what he is saying, is he saying that all the data we saw on 2.2 and 2.3 Ghz with a BIOS fix that depresses 10% or that the BIOS fix on a 2.4 GHz causes it to perform like the 2.2 and 2.3 without the fix....

Theo is one that I am skeptical when he posts anything.... he does not hold much credibility to be honest. The 30K 3DMark06 info is a good example. I can find others, I think he was the one who announced reverse-hyperthreading is real and it was benched.

Having made my comments about Theo, a few comments it is nice to see someone come out and say 'it is in all CPUs', this means the stepping will fix it and his post sounds upbeat that it is coming sooner rather than later (I will likely pick up a B3 Phenom, so long as I have confidence AM2 will accept it in my board).

It will be interesting to see if that stepping consumes less power overall, I would like to see power at the socket data from such sites as digit-life.com and lostcircuits because someone floated a theory that the problem was a hot spot (on a different forum). If so the fix may just lower power a bit to get rid of that hot spot, assuming that theory held.

Anyway, I am not surprised that they may have found a non-performance degrading way around the errata, if so that would be great for the early adopters as this would likely enable higher overclocks.

BTW -- Jumpingjack is an old childhood nickname, you can call me Jack.
Jack

[JumpingJack]

lol @ newegg's misspelling quad "qaud"

Misspelled or not, the price Newegg offers up has come in nicely ... though they also list the Q6600 at 279.99, it probably should drop another 20 bucks or so.... maybe not, MBs may make-up the difference.

Jack

[informal]

Thanks for your thoughts Jack ,Well i wanted to write a full username ,i will stick with Jack in future

Anyhow,Fuad(another inq/ex-inq "hack" ) reports some good news about "fixed" Phenom,presumably B3,that is already out in ES form (news if they are true)

AMD's Phenom 2.6GHz+ bug free samples are out

[JumpingJack]

Thanks for your thoughts Jack ,Well i wanted to write a full username ,i will stick with Jack in future

Anyhow,Fuad(another inq/ex-inq "hack" ) reports some good news about "fixed" Phenom,presumably B3,that is already out in ES form (news if they are true)

Yeah, if AMD intends to get this out in volume in Q1, they need a quick turn around. However, FUADs account of B3 silicon with the fix may or may not be accurate.... if this bug was discovered in the 11th hour, then to tape out and get first silicon would not be possible by now, however if this is the same TLB bug that is in the errata (#254), then likely it is fixed via that errata and they have had enough time to bet first silicon by now.

Personally, Fuad is probably right on this one -- if they wanted to get B3 in volume by Q1 (any time beginning or end), then they must have it taped out and running by now.

[Zytek_Fan]

Once the L3 errata is fixed I'm sure Phenom will have a nice performance gain. BRING NEXT BATCH BY MARCH PLZ!

[JumpingJack]

Once the L3 errata is fixed I'm sure Phenom will have a nice performance gain. BRING NEXT BATCH BY MARCH PLZ!

I am not so sure about that, if all they are doing is fixing a overheating issue (hot spot theory), this will not change the logic. The performance gain will be nothing more than being able to bin and release higher clock speed.

http://www.amd.com/us-en/assets/cont...docs/41322.pdf See errata 254.

However, since the L3 is indpendently clocked from the rest of the core, it could be a timing issue, in such a case the errata AMD lists in their revision spec is simply a "livelock" condition, as I understand the term this simply a conflict of shared resources... if all the fix does is resolve conflicts it may bump performance of some multithreaded code -- not sure myself.

Jack

[FireDemon666]

OK guys forgive my not being out of the woods on this one but I'm still living in the days of Barton core and plan on buying a Phenom in the next few weeks and am cramming all the AM2+ info into my head I can.

What exactly is "wrong" or "bugged" about the current batch of Phenom chips?

[informal]

The bug is in L3 and it's supposed to kick in at >2.4Ghz frequencies(OC or stock).So the deal with the microcode update for your board would be that it would circumvent the errata so that you could achieve higher clock with your chip.You shouldn't worry about this since the OC issue will be dealt with the mentioned update for BIOS.If you don't intend to OC your chip,than this "bug" has zero effect on you

[FireDemon666]

The bug is in L3 and it's supposed to kick in at >2.4Ghz frequencies(OC or stock).So the deal with the microcode update for your board would be that it would circumvent the errata so that you could achieve higher clock with your chip.You shouldn't worry about this since the OC issue will be dealt with the mentioned update for BIOS.If you don't intend to OC your chip,than this "bug" has zero effect on you

I own a Vapochill LS. To not overclock would be sacrilege lol.



Also, all the Phenom's are the same correct? Not counting the clock speeds of course. I know they all have the exact same cache and core so I'm assuming the only difference between the 9x00 models is the clock speeds. Is AMD still locking the multi's this time around?

[JumpingJack]

OK guys forgive my not being out of the woods on this one but I'm still living in the days of Barton core and plan on buying a Phenom in the next few weeks and am cramming all the AM2+ info into my head I can.

What exactly is "wrong" or "bugged" about the current batch of Phenom chips?

Informal did a good job explaining above.

It is not uncommon, especially for new chips, to come out and still have undiscovered bugs. Both Intel and AMD publish spec updates with errata listing the bugs, how they manifest themselves, and the workarounds or planned fixes, if any.

With respect to this chip, AMD discovered a bug in there TLB (translation look aside) buffer, this tracks activity and stores trace information about tag lines and page tables (memory blocks) in cache and where those page tables tranlate back into main memory.

What AMD found was that beyond a certain clockspeed, a processor would lock up under certain conditions (fully loaded on all cores, with certain benchmark software). It is important to note that it takes a very special set of circumstances to trigger the bug and will not always (and perhaps never) appear, but evidently it was showing up sufficiently often enough to cause AMD to postpone introduction of 2.4 GHz bin sizes until a fix could be implemented by issuing a new stepping (which is industry talk for new revision, sorta like a new model year of a car -- not sure how much you know about the jargon).

The initial news was a bit odd because typically errata are logical bugs in the logic of the processor, and speed bins of the processors all derive from exactly the same stepping or revision, so it confused people when the Inquirer said the "2.2 and 2.3 GHz chips don't have it and the 2.4 Ghz chip does" .. this makes no sense because the 2.2 and 2.3 Ghz chips are exactly the same as the 2.4 Ghz chip.... if the 2.4 has it all should have it... well, Theo of the inq comes back and says, yes, they all have it.... but the bug only triggers above 2.4 Ghz.

Winded, detailed explanation... sorry. Wanted to make sure it was clear.

[informal]

@FireDemon666
Yes they are the same and the multipliers are locked(unlocked downwards).
BUT in December there is 9600(2.3GHz) Black Edition coming with no price premium over the "regular" 9600s,so that one would be OCer dream come true.You could avoid the "HTT" issue by simply raising the multi(seems to be hard to pass low 240s for HTT on 9500 model for instance).

[FireDemon666]

@FireDemon666
Yes they are the same and the multipliers are locked(unlocked downwards).
BUT in December there is 9600(2.3GHz) Black Edition coming with no price premium over the "regular" 9600s,so that one would be OCer dream come true.You could avoid the "HTT" issue by simply raising the multi(seems to be hard to pass low 240s for HTT on 9500 model for instance).

You are kidding me? In December? Because the Phenom is gonna be my christmas present this year...

BTW has anyone discovered whether the Phenom's cold bug? I wanna push this guy as far as it can go on all frequencies (pun intended).

[JumpingJack]

You are kidding me? In December? Because the Phenom is gonna be my christmas present this year...

BTW has anyone discovered whether the Phenom's cold bug? I wanna push this guy as far as it can go on all frequencies (pun intended).

No data yet, or none that I have seen ... get one, cool it down, come back, tell us

[FireDemon666]

No data yet, or none that I have seen ... get one, cool it down, come back, tell us

Only thing that worries me is this L3 bug. If 2.4 starts the ruckus, I can only imagine what will happen past 3Ghz...

Unless these BE chips will have the revised stepping. This would only make sense since they are aimed at being overclocked...

[JumpingJack]

Only thing that worries me is this L3 bug. If 2.4 starts the ruckus, I can only imagine what will happen past 3Ghz...

Unless these BE chips will have the revised stepping. This would only make sense since they are aimed at being overclocked...

Yeah, that is why if someone should ask me I would advise waiting for B3. The bug should be hard fixed in logic with that stepping.

That is what is holding me back from picking one up.

[AMD666]

Interesting find.