Phenom Exposed! Shipping with flaky 3rd cores.

[bombasschicken]

Via still makes processors, although they specialize in the Small Form Factor and ultra low power chips these days.

http://www.via.com.tw/en/products/processors/isaiah-arch/

Wow.. VIA... I knew they did chipsets but I had no clue they still did processors.. I remember my old cyrix C3 800mhz chip.. That thing was a work horse..

Ok so I take it back. Perhaps there is someone out there that can fill the viod..

(Although I'm perty sure that IBM won't being that they focus mostly on business customers...)

 

[ragemonkey]

@jimmysmitty

It will definately be a confirmation of this issue should the Tri-Cores show Core #s of 0,1, and 3... with Core 2 (logical #3) turned off. Considering the mounting number of people having this issue, AMD will probably have a decent launch of the Tri-Cores... but by that time the actual reason why will probably have hit the masses.

 

[ragemonkey]

I can't resist anymore.... In a Phenom, Core2 is defective, is this AMDs idea of a sick joke??

 

[endyen]

Well, all I can say is I had the same problem to begin with. Ended up having to reset the bios, since I had originally had an Athlon x2 4200+ installed. And I had to reinstall windows, before that AOD kept causing those errors, and ended up corrupting the boot files.

Unless your using v2.0.10 or 2.0.14 of AOD don't even bother using it, it will cause that every time you mess with a processor setting, and even with those versions its likely to happen. The ONLY way to stably oc is through the bios, if you have a board that can support the thing properly.

Other than finding OC limits, and a weak rail on my PSU I have had zero problems. And yes, part of the time I had instability was at stock settings.

Good post, too bad the Intel Inside croud have selective reading skills.
To the OP. Do as Mathos suggested.
Reset the bios, then do a clean OS install.
You should also try changing your v-dimm.

 

[systemlord]

I would love to hear from thunderman about this issue, respecting a company that comes out with crap like this loses respect!

 

[Guest]

The defective core is all part of Phenom's superior design! 😛 lol

 

[randomizer]

I have a question. Why is it always core 2? There must really be something wrong if they managed to always stuff up the same core in every die.

 

[ragemonkey]

I have a question. Why is it always core 2? There must really be something wrong if they managed to always stuff up the same core in every die.

Hmmm... must be a consistent problem in the manufacturing process. Either that or AMD is purposely sabotaging Core2 due to all of the financial problems Intel has caused...lol I think the better solution would be to number the cores 0...1...3...4 , just like a hotel skips the number 13! lmao... i couldn't resist!

 

[croc]

I have a question. Why is it always core 2? There must really be something wrong if they managed to always stuff up the same core in every die.

Best guess, they have a flaw in either masks, or their litho process. I have yet to see any hard evidence that this is a 3rd core only, but anecdotally it appears to be so.

I'd wonder at AMD's motivation to release these chips to the general public if they knew about this glitch. And I'd also wonder about why the announcement of a tri-core so far ahead of phenom's release if they didn't know about this glitch.

But then again, I wonder about four undersea fiber cables failing in the Mideast in so few days....

 

[mrmez]

Its (obviously?) a manufacturing error, as opposed to a design error, i would think.
I believe its a result of AMD not having enough options and control of who and how their chips are made, and shooting their marketing mouths off and backing themselvs into a corner.

As someone whos been involved with manufacturing and quality control, u sometimes have a situation where u, like AMD have just produced tens of thousands of chips only to find a problem in ALL of them.

Option 1: Scrap them all. Throw them away or turn them into Tri cores, sacrifise some profits and in the process maintain a quality brand image.

Option 2: Release the chips to market knowing u will earn more for them, and the cost of replacing a few of them will leave u financially ahead of option 1... in theory. At the same time further destroy your already shoddy brand image.

Making a decent quad core cpu MAY be rocket science and i wont hate anyone for getting it wrong. I will hate amd for intentionally releasing faulty chips.

How can faulty chips get mass produced in this day and age?
A quick story...


The hubble space telescope. A massive engineering feat, even by todays standards. After sparing no expense to design, build and deploy the telescope, they turn it on and BAM, its all out of focus. ARRRGGG!
Long story short, when making/calibrating the massive mirror, a tech dropped a floor standing calibrating calibrator. The metal coating on the bottom of the leg chipped off. Not realising this and not being bothered with checking or recalibrating the device, the tech continued as usual.

The 10th of a mm that was chipped off turned a billion dollar project into garbage. Lucky they found a 'cheap' way of re-aligning the mirror in space.

If NASA can fk up so badly, imaging what AMD can do.

 

[Hellboy]

Will the last engineer switch the light off when leaving AMD lab...

 

[Reynod]

It's a hot spot ... over the corner where the OMC feeds the core ... my prediciton.

Core #2 (3)

But we will never really know eh ??

I doubt AMD tried to cover anything up ... otherwise why would they release the AOD and allow individual core overclocking?

Might explain the tri-core offerings to follow.

I'd imagine they will likely release with higher speeds.

This reminds me of the Palomino days ... three steps till they managed to get the core speed up ... tried to do one or two fewer layers to speed / reduce the costs.

One of the threads has a few good points ... if you can't push the fsb past 215 (even with lowering the mult) take it back.

 

[Reynod]

The problem is they are not listening to the Engineers.

Someone listen to the Engineer and not the Marketing guy ???

 

[Ycon]

Well, I guess you have bought one of the first Taliban processors; quite an honor, isnt it?

Too bad it cost you quite some cash.

 

[spoonboy]

^ I agree with you here. I was very suspicious when AMD kept making a big deal about the TLB errata because in all seriousness, there is nothing "big deal" about it at all... at least in a consumer setting. I always thought there was more to it than that. Sadly enough, AMD has been all about misdirection and misinformation over the past year plus. I am just very surprised that AMD allowed the BE to be launched... did they not think the enthusiast would figure out what is really going on?

What was even more suspicious was the recent pushback for B3. I am guessing at this point, B3 cures the errata but not the defect in the manufacturing process that is causing the Core 3 defects hence still no steady crop of Phenoms above 2.3 ghz.

If the SAME core is problematic on all Phenoms, that surely points to a design fault; the coordination of that cores activity with the others. It could be a manufacturing fault, but if its the exact same core then something is up in the design/architecture not the process. No one should buy a B2 Phenom, you know it will just annoy you.

 

[chunkymonster]

Wow! Interesting reads. Thanks for the info.

That's a damned shame. It's one thing after another.

The B3 better be all-that-and-a-bag-of-potato-chips.

I'm on schedule for an upgrade the end of this year as I'm feeling the limits of my dual opty workstation. If AMD doesn't deliver B3 or 45nm it will be a Nehalem for me.

 

[buzzlightbeer]

Even tho my phenom does not overclock, im still happy with it i play every game on high settings and the cpu is still at about 25%

im not saying its a bad CPU i just thought a BE would actually over clock what else would you buy it for, and just looking at those new intel 8 core systems my mind has finaly turned ,i was stubbon about it but i do realise now that AMD is getting owned majorly

and i just wanted to add in this edit


i think the reason my BE wont overclock is the PSU im getting a new one . i tried many things , when i first got it with my old PSU i was running one 3870 it ran MUCH better, with this newish one and 2 video cards its a dog.
im gonna test and see if it works out ill let you guys know

 

[ragemonkey]

Even tho my phenom does not overclock, im still happy with it i play every game on high settings and the cpu is still at about 25%

im not saying its a bad CPU i just thought a BE would actually over clock what else would you buy it for, and just looking at those new intel 8 core systems my mind has finaly turned ,i was stubbon about it but i do realise now that AMD is getting owned majorly

and i just wanted to add in this edit


i think the reason my BE wont overclock is the PSU im getting a new one . i tried many things , when i first got it with my old PSU i was running one 3870 it ran MUCH better, with this newish one and 2 video cards its a dog.
im gonna test and see if it works out ill let you guys know

Use this to calculate the aproximate PSU you will need. It helped me out a lot when I picked up my Neopower 550.

http://www.extreme.outervision.com/psucalculatorlite.jsp

 

[Mathos]

Or you could just talk to someone who's familiar with OCing the processor, or getting it to run stable? A lot of the board manufacturers are having issues with getting bios's that will run stable with a phenom installed. 1.1b3 for the k9a2 plat seems to be the one for OCing, 1.2 for OCing but with TLB fix enabled and no way to turn it off.

 

[ryman554]

If the SAME core is problematic on all Phenoms, that surely points to a design fault; the coordination of that cores activity with the others. It could be a manufacturing fault, but if its the exact same core then something is up in the design/architecture not the process. No one should buy a B2 Phenom, you know it will just annoy you.

No, the SAME core points to a weird manufacturing flaw, since hte deisgn of all four cores are identical. Or at least they should be if they did it right. If the design indeed differs, I know of a root cause, and the fix isn't pretty.

That being said, it really does smack of a reticle defect, but this should be really easy to spot/fix. Unless it's deterministic, in which case all bets are off.

Actually, now that I think of it, I think I know now *where* the issue lies. Sadly, due to the nature of my current job, I can't comment any more.

 

[randomizer]

Considering tri-cores were announced so long ago, it looks like AMD knew about this for a very long time. Why would you release a tri-core which has one fully functioning core disabled if it would just add to loss of profits that could have been made by selling it as a quad? From the looks of it, they won't even need to test the cores, they need simply always disable core 2 in a batch of Phenoms that they know will be faulty and they have a fully functioning tri-core (well, they'll probably disable some other things too, but thats beside the point). Although as croc said, there isn't any hard evidence to suggest it is always core 2.

@ryman: Just cry "freedom of speech" or join a union or something :lol:

 

[bmadd]

well i can comment that it isn't ALWAYS core2 because my core3 overclocks worse then my core 2 does. i have an asus m3a32mvp deluxe and in bios version 0801 you have downcore. if i downcore core 2 and 3 i can clock my 9500 as a dual core to 2.9 and everest registers it as a KUMA not an agena

 

[Scarchunk]

if i downcore core 2 and 3 i can clock my 9500 as a dual core to 2.9 and everest registers it as a KUMA not an agena

Now that sounds like a pretty sweet chip with some decent OC headroom. I'd like to see it benched against a similar core2duo. Again, why the hell hasn't AMD released a Phenom dual core?

 

[bmadd]

well phenom isn't not an over clocking persons friend. I just fried my one trying to get my core2/3 to go further bad idea. i doesn't like 1.68V. dont know y. so its back to a 5000+ BE till they get there act together. just thought i would add this in

 

[croc]

Try for an RMA... Sorry.