curious on binning

[bob4933]

I know amd has done this before with their "tri cores" simply being quad cores that failed...

Any thoughts on 6300's being 8300's that have "failed"?

Likewise, are 8320's just simply lower clocked, or are they binned out from 8350's?

Not expecting "for sure" answers, but knowledgable speculation is welcome

 

[tungureanu]

This is how the bin them best to worse 9590-9370-8350-8320. This is what they do with the 8 cores. As for the 6 and 4 core FX`s idk for sure but from what i heard somewhere they begin life as a 8 core and they chop off the modules that don`t work. Therefore no unlocking chips.

 

[ShrimpBrime]

Here's a picture of a FX-6100 next to a FX-8320.

I'd say the 6100 had a couple of cores shut down. Meaning the cores are there, but cannot be accessed by Bios software anymore. AMD put the fire out on that. Too bad, cause that was a good selling feature.

 

[Mandrake4565]

Here's a picture of a FX-6100 next to a FX-8320.

I'd say the 6100 had a couple of cores shut down. Meaning the cores are there, but cannot be accessed by Bios software anymore. AMD put the fire out on that. Too bad, cause that was a good selling feature.

Out of curiosity Shrimp, what makes you say that. Is it because they both look the same or am I missing something?

 

[tungureanu]

Here's a picture of a FX-6100 next to a FX-8320.

I'd say the 6100 had a couple of cores shut down. Meaning the cores are there, but cannot be accessed by Bios software anymore. AMD put the fire out on that. Too bad, cause that was a good selling feature.

we have to talk to the Stilt. maybe he can hack a bios for the chv so we can atleast try to unlock them.

 

[caddi daddi]

sheesh, manny, have you ever seen a low leakage 6 core fx?

and take a look at the quad cores, now were talking my kind of vcore!!!!!

 

[Mandrake4565]

Actually CD I've never seen a Fx 6 up close.

 

[ShrimpBrime]

Out of curiosity Shrimp, what makes you say that. Is it because they both look the same or am I missing something?

Ya because they are the same size.

This can also be seen with Phenom II quads vs Hexa T series quads. edit (have also de-lidded)

The T series quads are bigger and sport 6 cores and are "locked" and are able to be "unlocked" via bios firmware and probably firmware on the cpu as well. The regular x2 Phenom II's are also quad cores, we know this from unlocking them.... So no need to pull IHS plates there..... But ya, more cores = bigger die sizes.

Remember there is two sides to running a cpu. The bios for board settings and Cpu firmware for Pstates and locked multi and so forth. If I weren't mistaken, the firmware on the actual processor is what will need the modification, not the bios and chipset relation so much.

EDIT:

In otherwords, I do not believe AMD spent the time and money to actually "cut" cores off physically.

 

[ShrimpBrime]

Any thoughts on 6300's being 8300's that have "failed"?

Wanted to comment on this really quick.

Answer, Yes and No!

Some wafers probably had failures past 6 cores and some did not. It's so easy to control the Cpu for them, the cores don't need to "fail" tests to be a hexa core. If they where to run out of bad 8 core wafers, they'd use good ones and just simply turn off the last module or w/e it's called now a days lol.

 

[RGone]

Well I don't want to get just too freeken crazy since the ones that know cannot say, but the one thing I know about any cpu is that within the 'same' series, no cpu manufacturer sets out to make a "lesser" cpu. I am sure AMD wishes they could be productive enough to make all 9590s.

There is one thing that I have heard and again AMD is not going to say yes or no, but they know which wafers will make 9xxx cpus, if all goes as expected.

Just remember I said within a series of processors. Just like S_B said the Thuban Ts are bigger than the Deneb 4 core cpus. Have to be since there were no Deneb 6 core processors and they had to increase the size to slide two more cores into a Thuban. They probably kept the basic IMC from the Deneb when they added the extra 2 cores and that maybe the reason quite a few Thubans were speculated to have weaker IMCs.
RGone...

 

[caddi daddi]

manny, it looks just like your 8350, just takes more vcore and is easyer to cool, thet even scale the same so I'm with shrimp-brime and Rgone, failed 8 core becomes a 6 core, faied to bin as a 6 core becomes a 4 core, any other way would be wasteful.

 

[ShrimpBrime]

Well I don't want to get just too freeken crazy since the ones that know cannot say, but the one thing I know about any cpu is that within the 'same' series, no cpu manufacturer sets out to make a "lesser" cpu. I am sure AMD wishes they could be productive enough to make all 9590s.

There is one thing that I have heard and again AMD is not going to say yes or no, but they know which wafers will make 9xxx cpus, if all goes as expected.

Just remember I said within a series of processors. Just like S_B said the Thuban Ts are bigger than the Deneb 4 core cpus. Have to be since there were no Deneb 6 core processors and they had to increase the size to slide two more cores into a Thuban. They probably kept the basic IMC from the Deneb when they added the extra 2 cores and that maybe the reason quite a few Thubans were speculated to have weaker IMCs.
RGone...

I'm in shock that AMD hasn't applied efforts into better steppings for possibly lower v-core or higher clocks or both. It was nice to have choices C2 or C3 B2 B3 steppings like of the Phenom I & II's. The challenge of a 4ghz C2 was always fun!. They say couldn't be stable at 4ghz, my de-lidded 965BE proved people wrong there lol.

I thought the Thuban IMC was quite a bit better than that on Deneb. I was slamming 1100mhz effective ram clocks on 1.68v or less. The Deneb would need over 1.75v in most cases closer to 1.8v. Asus boards have a habbit of pumping 1.9v while set to auto. But the stability was there!

If I buy into AMD again (for personal gaming ext) It would have to be a FX-9590. That way I can watch every one with 8 series chips try and keep up. 5ghz turbo from the box is nice. I recon in the right hands these chips "could/should" be doing 5.4ghz daily despite the high wattage which we'd expect anyways.

 

[Mandrake4565]

I thought the Thuban IMC was quite a bit better than that on Deneb. I was slamming 1100mhz effective ram clocks on 1.68v or less. The Deneb would need over 1.75v in most cases closer to 1.8v. Asus boards have a habbit of pumping 1.9v while set to auto. But the stability was there!

I agree with you here Shrimp, my 960T though I haven't put a lot of work into as far as the ram, I have been able to run it tighter and faster then my Phenom. I was able to push the NB Freq on the 960T above 3k for some benches.

 

[ShrimpBrime]

I agree with you here Shrimp, my 960T though I haven't put a lot of work into as far as the ram, I have been able to run it tighter and faster then my Phenom. I was able to push the NB Freq on the 960T above 3k for some benches.

The NB speed should be able to hit 3ghz on just about any chip that says Phenom on it. And that does include Agena based cpu as well. My 9950/9850BE ran a 3ghz NB daily.


Not sure what happened to FX though. I got just over 2.9ghz max and it wasn't stable for pooop. 2.7ghz was ok to bench on, couldn't pass my stability testing which usually consits of either days/weeks on F@H or hours days with Prime95 and OCCT running together. Tell you what, those two programs together will get a mouse to lag.

I don't suppose any one has gotten 3ghz NB and beyond here? ( FX chips) If so I haven't taken notice in it.

 

[Mandrake4565]

I think Johan has had his there or at least pretty damn close for benching. So far I haven't figured how to get my 8350 to like anything above 2800.

 

[ShrimpBrime]

I think Johan has had his there or at least pretty damn close for benching. So far I haven't figured how to get my 8350 to like anything above 2800.

Hmm. I did want to go super cold DICE/LN2 or such to see about 3ghz NB speeds. Did try a TEC as well.

Not sure if maybe some chilling on the chipsets would help? I thought about the TEC on the NB, but chickened out about it. Didn't want to break the board either, it was doing fine with a fan, good temps.

Used voltages....

CPU/NB 1.55v NB-Core-V @ 1.2v. Was thinking maybe I should have tried more NB core voltage, but again got the chicken out thought and figured, meh why bother. It may be epic, but no doubt some one has beat me to it with LN2 ya?

 

[Mandrake4565]

I've seen 3k+ on LN2 for Sp1m on the bot, Johan has one at 2935mhz on cold water with his 9370.