Phenom Exposed! Shipping with flaky 3rd cores.

Page 6 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar


Not necessarily...

look at the 45nm quads from intel -- one of the reasons they are holding back due to issues with their thermal sensor or something. It's not something which will plague ALL processors, only SOME. Most likely that had to do with process variability, too.

granted, you'd think that AMD's QC would catch this... and maybe they are in binning.
 


No way it's a motherboard problem. Here is a (partial) list of the motherboards which have this problem...

MSI K9A2 Platinum
Gigabyte GA-MA790FX-DQ6
ASUS M2N-sli Delux (this last one is confirmed to have the latest TLB patch bios)

As MU said, still could be a software (vista) issue. Or a CPU issue. And since somebody reported that people can swap CPUs and get rid of the problem, that pretty much rules out software issue. What else is left?
 


It could be- did anybody happen to catch the date code and batch of the Phenoms they used and the ones they had trouble with before seating the heatsink on the chip? It could be a bad batch, run, or simply a bad CPU here and there. Nobody has said what they had to do to get the bug to occur, so perhaps it is something that would not be gathered in the quick testing that IC makers give their gear before separating out by bin.
 


Thats the problem. When KTE was doing his OCing thing with the Phenom BE's he was keeping track of the batch and week numbers on them. One processor out of the same batch and week would OC better than another one of the same instance. Or one of the same week/batch wouldn't oc at all, while another would easily hit 2.9g. It's almost like going luck of the draw with these things. Some people get lucky, others get a giant shaft.
 


AMD 9600 Blue (screen) Edition mate, its all the rage 😀 got one yet?

Dont people remember, bigger numbers are better - higher ghz = better, higher cache = better, higher power (consumption) = better, higher blue screen is also better and the main feature, its all the rage.
 


That's good data.

This is yet another (big) sign that AMDs 65nm manufacturing solution (in its entirety) is just not up to prime time. Sounds like they are battling all sort of variability problems that they are literally on a knife edge. Combine this with (most likely) agressive design choices for Barcy, and you get this.
 
After doing a rather extensive web search, this seems to not just be related to the phenoms, but there are reports of 4000+ and 6000+ athlons as well that exhibit this 'phenomenon'. (pun intended).

So could this indicate that a minor process failure has finally reared up and has caused major issues with the shrink from 90 mm to 65 mm? I'm no expert here, but from the anecdotal evidence it does seem to be a bit more pervasive than just the phenom line.
 


Heh i actually ordered in 2 x 6000+'s for a client and built two matching machines, one would fail orthos - tried everything and eventually narrowed it down to the cpu, so would i be calling that a 50% sucess rate? Thats the first brand new faulty cpu i have come across, and now it makes me wonder wether this has anything to do with it, it did fail on the second core (Core2 lol yes), even when the memory was limited to 533, two motherboards, psu's everything...
 
the problem can be combination of few factors
like people wrote befor, it isd not only deign, not only manufacturing and not only process.
but it certainly be combination of few factors: marginal design that would be OK with perfect manufactoring and/or testing. the design may be "on the egde" and then with too wide manufactoring variability, there is high percentage of phenoms that have this flow. most are still functioning om stock speed, but fail with small OC. some are failing even on stock speed, and some would fail only with much higher frequencies. in such cases, especially when trying to save the yiled numbers, testing is complicated - if they put the limit too low they will throw away too many good units, if it is too high they will miss some units which crush even on stock frequency. it seems that AMD blew up with this one as well, and put the limit too low.
maybe in later stage at least the testing was improved, and then later phenoms are OK




 
Hmm... Pretty interesting. If comfirmed, I wonder how this will affect the majority of PC users. At my local best buy is seems like almost every other desktop is running a Phenom. Also Gateways FX Desktops will be exclusively Phenom (as I've heard).

As much as I want AMD to rise up, its hard to support them when you hear about stuff like this.
 
I have zero problems with my 9600BE so far. No BSOD. No bugs. No stability problems. OCed at 2.7 it runs like a champ. I have built 6 Phenom media PCs for folks so far and none of them have any problems. 3 of them are OCed to 2.7 for gaming as well.

The problem with the 2nd core you are seeing is not nearly as bad as people here are making it out to be. The phenom is a new chip and the only ones buying it are not main stream. Most people never even hear about it and nobody is realy building Phenom systems in mass. Its like the release of any CPU. There will be a defect in a few. I still see tons of Intel buyers complaining every day about bad cores ect.. Any vender will tell you the amount of defecto returns for the Phenom is well within the norm.

I am sorry but I am rofl at this coolaid party here. Call any major vender around. There is nothing out of the ordinary with Phenoms. They are not showing up at peoples doors in mass DOA. No AMDs are. Gossip is not healthy. I expect better from a forum full of guys that probably spend more time on the PC then most.
 


Not the Brisbane's too! That thought makes me want to stick with Windsor until 45nm arrives and shows itself to be top notch. For me, it doesn't have to be as fast as Intel, but I want good quality control. Right now, I'd be more inclined to buy a 65nm triple core Phenom than a quad core. I can't see them accidentally disabling a working core and leaving the bad core enabled.

As for names making fun of AMD's Phenom's, that reminds me of the names we called the Pentium 60:

Spentium

Repentium

They're the only ones I can recall offhand. Two things that got people miffed at Intel back then; the errata that wrecked havoc with using the calculator in Windows, and Intel's claim that you'd better buy one anyways because we aren't coming out with anything faster for quite awhile; then they came out with Pentium 75's and Pentium 90's.

AMD and Intel really copy each other's playbooks, both the touchdowns and the fumbles. Imagine what would happen if they both fumbled in the same generation?

Let's hope the B3 Phenom's not only correct the errata, but don't have bad core 2's (unless they're triple core Phenom's with core 2 disabled). Xbit labs reports AMD will be releasing B3's in April:

http://www.xbitlabs.com/news/cpu/display/20080208183158_AMD_Plans_to_Introduce_New_AMD_Phenom_Chips_in_April_Source.html

They should do so on April 1st, but deliver a winning product. Then the joke would be on everyone who'd thought they'd lost all hope of recovering from a year of fumbling (except for ATI, they've had a good year).


 


Oh, I don't think it's a wide spread issue.
Lots of bashing.

First all, it's only an issue if happens at stock.
OC's are not promised.

It could also be an important point when testing OCs.
If you hit a wall OC'ing the Phenom, try bringing up particular cores.

For non-quad optimized stuff, if you can get 2 cores higher than you can get four, you could set the affinity for your games to the two fast cores.

 


I think people are simplifying things way too much. It could easily be a combination of any number of things ie process, mask, mb.....

What if it only effects Chips that were shot using this Rev of this mask for this layer which are edge die produced between these dates that ran on this tool and were delayed XX minutes or more between these steps and run on 4 layer PCB motherboards with this trace exceeding this length.

Lets be realistic, we are not talking about making a cheeseburger.
 
fyi when i got mine i had the same problem but it was expected. ive had it happen a couple times before during different upgrades. i always had to reinstall, so i expected to this time as well, and sure enough i did. go figure? you make a drastic change to your system and you have to reinstall. thats fairly typical.
 

It's targeted to gamblers. Let's say only 1% of all the Phenom BE can get to 3GHz. If you're lucky, you get a 3GHz Phenom!
 
fyi when i got mine i had the same problem but it was expected. ive had it happen a couple times before during different upgrades. i always had to reinstall, so i expected to this time as well, and sure enough i did. go figure? you make a drastic change to your system and you have to reinstall. thats fairly typical.
 

Those are all real world scenarios, that cause real problems for all chipmakers. That's why chips spend 4 weeks in testing.
AMD's QC department is made up of mainly German workers. They take a great deal of pride in what they do.
 


Yep, and if you can't overclock a "black edition", then all that particular marketing term really means is an attempt to put AMD back in the black. This is worse than Intel's whole Netburst mishegoss, which Intel fans seem to think is just ancient history. What's shown with both situations is how far they go to sell parts that don't cut it and can't be sold on merit.

I'm an AMD fan, but not a fanboy. It's fine if they sell triple cores that have one bad core disabled. That's honesty and meeting a market that might exist, especially as Microsoft has expressed interest in optimizing for odd numbers of cores.

It's not fine if a design issue at 65nm, or a QA issue gets enabled bad cores out into the marketplace, especially if those bad cores aren't on Gateway's but are in "enthusiast" marketed parts. IMHO, this just sounds like a QA issue and I doubt that many black editions have this problem, but the fact that any do is really sad.

I want AMD to succeed, and not just to drive down prices on the Evil Netburst Empire's finally good processors. With a viable AMD, we have competition, innovation and reasonable prices that allow profits to be made by both companies. Without AMD, we have Intel, and I don't want to return to circa 1995 pricing.

Like I said, I'll get two B3 triple cores when they arrive. That's the only 65nm Phenom part I'll trust right now. What I really want are the 45nm parts, but they're some distance off. Still thinking of a Q6600 for my wife, as her graphics and modding would benefit, but both my PC and our kid's would do perfectly well for games with just AMD.

Now all I need is to get that Antec TPQ 850 or 1000 with a 6+2 so I can finally install that MSI factory overclocked 3870x2. ATI's success makes me think that they should have bought AMD instead of the other way around. AMD with ATI suits in charge might just be a viable company.
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom