When it Rains, it Pours - More Phenom Delays

J

JackPack

Limp Gawd
Joined
Aug 1, 2005
Messages
276
According to HKEPC, B3 stepping has been pushed from Q1 to Q2. To reflect the change in stepping, AMD will name these "new" processors as 9x50 (i.e. 9550 and 9650). Anyone else find this funny? Fixing a performance-robbing bug to restore the CPU's original performance deserves a higher model number? Too bad AMD wasn't this honest when they sent Phenoms with overclocked north bridges to reviewers for benchmarking when Phenom was released.

http://www.hkepc.com/?id=550
These look like official slides, but slightly altered by HKEPC to avoid getting nasty emails from AMD's legal affairs team.
 
god damn dude I've been to two chinese servers today and both are slow as f00k. Can you cheap bastards not afford bandwidth? Comon your sites are like ad-farms... jeez. The java on the site's being a bitch too. Im running firefox btw :eek:

anyways, to the point at hand:

Discribing the L3 bug as "performance robbing" is a little much. It caused a BSOD so so rarly and only when really pushing the virtualization technology (something and end user is likly never to do). I think every heres being a little harsh on amd.

and I dont trust this "semi-altered" slide-id-ness. Sounds shifty to me

As allways, we'll have to wait and see.
 
Someone start a poll asking people on here if they have had a problem with these chips in their normal use, and put this shit to bed.
 
MrWizard6600 said:
Discribing the L3 bug as "performance robbing" is a little much.
Click to expand...
It's a TLB/L3 timing bug where incorrect values are available for a short period of time that should have been refreshed under certain conditions. The BIOS fix to avoid the bug is "performance robbing," and as recommended by AMD, is not available for the user to turn off in CMOS set up. While the 1337ies will be able to turn it off in a future version of the overclocking utility, or patch his or her Red Hat kernel with an unofficial/unsupported patch, everyone else will see a significant performance reduction on B2 Phenoms.
-----

The TLB bug is a big deal, and no major OEM which usually account for 80% of Opteron sales, are touching these chips with a 10 foot pole. You'll also notice an almost complete lack of desktop Phenom systems for sale by any major OEM. This is not a minor problem, Dethred. About the only good to come out of this is the relatively small numbers affected after 3 months since AMD still can't make the chips in quantity.
----

Anyways, this is HKEPC. They can make a picture but this is just another rumor. If AMD is getting samples of B3 in January, mid- to late Q1 still looks like it's possible for the fixed chips releases.
 
MrWizard6600 said:
Discribing the L3 bug as "performance robbing" is a little much. It caused a BSOD so so rarly and only when really pushing the virtualization technology (something and end user is likly never to do). I think every heres being a little harsh on amd.
Click to expand...
Machines from OEMs are likely going to ship with the BIOS fix enabled. The fix is what robs performance.
 
http://www.digitimes.com/mobos/a20071224PD200.html is reporting that the 9700 and 9900 are delayed to Q2:
AMD has recently notified its partners that the launch of higher-end quad-core Phenom processors, including the 9700 and 9900, will be postponed to the second quarter of 2008 from the original schedule of early 2008, according to sources at motherboard makers.
Click to expand...
Maybe that's where HKEPC is confused.
 
At least one mobo maker has a bios to turn it off IIRC, and the newest AOD allows you to toggle it off as well.

We'll see about those extended delays....
 
Actually Rev B3 is ready - well sort of. They have it done and perfected... and I was not aware of any delay. Might be something happening now that I don't know of since I am on vacay :D
 
JackPack said:
According to HKEPC, B3 stepping has been pushed from Q1 to Q2. To reflect the change in stepping, AMD will name these "new" processors as 9x50 (i.e. 9550 and 9650). Anyone else find this funny? Fixing a performance-robbing bug to restore the CPU's original performance deserves a higher model number? Too bad AMD wasn't this honest when they sent Phenoms with overclocked north bridges to reviewers for benchmarking when Phenom was released.

http://www.hkepc.com/?id=550
These look like official slides, but slightly altered by HKEPC to avoid getting nasty emails from AMD's legal affairs team.
Click to expand...

I don't find that funny at all actually. It makes a heck of a lot more sense than keeping the same name and leaving it up to the consumer to figure out what stepping they're going to get, especailly since most retailers don't advertise the stepping at all, only the model number.
 
The article is full of crap, unless something happened between Friday and Monday (I'm on vacation) :rolleyes:
 
pxc said:
It's a TLB/L3 timing bug where incorrect values are available for a short period of time that should have been refreshed under certain conditions. The BIOS fix to avoid the bug is "performance robbing," and as recommended by AMD, is not available for the user to turn off in CMOS set up. While the 1337ies will be able to turn it off in a future version of the overclocking utility, or patch his or her Red Hat kernel with an unofficial/unsupported patch, everyone else will see a significant performance reduction on B2 Phenoms.
-----

The TLB bug is a big deal, and no major OEM which usually account for 80% of Opteron sales, are touching these chips with a 10 foot pole. You'll also notice an almost complete lack of desktop Phenom systems for sale by any major OEM. This is not a minor problem, Dethred. About the only good to come out of this is the relatively small numbers affected after 3 months since AMD still can't make the chips in quantity.
----

Anyways, this is HKEPC. They can make a picture but this is just another rumor. If AMD is getting samples of B3 in January, mid- to late Q1 still looks like it's possible for the fixed chips releases.
Click to expand...

LstOfTheBrunnenG said:
Machines from OEMs are likely going to ship with the BIOS fix enabled. The fix is what robs performance.
Click to expand...

I'm simply forwarding the view of our head [H]s, Kyle and steve. They both said so in a news item which looked very similar to this one. I have no evidence to back up their claims (which I posted), but I'm sure they do. Hope one of em tumbles across this thread.
 
morfinx said:
The article is full of crap, unless something happened between Friday and Monday (I'm on vacation) :rolleyes:
Click to expand...
I heard from a halfway reliable source last week that there might be a delay. This was before the rumors starting popping up at places like DigiTimes.. Unfortunately, everyone I know from AMD is off until the 7th - or later.. :(
 
MrWizard6600 said:
god damn dude I've been to two chinese servers today and both are slow as f00k. Can you cheap bastards not afford bandwidth? Comon your sites are like ad-farms... jeez. The java on the site's being a bitch too. Im running firefox btw :eek:

anyways, to the point at hand:

Discribing the L3 bug as "performance robbing" is a little much. It caused a BSOD so so rarly and only when really pushing the virtualization technology (something and end user is likly never to do). I think every heres being a little harsh on amd.

and I dont trust this "semi-altered" slide-id-ness. Sounds shifty to me

As allways, we'll have to wait and see.
Click to expand...

Rarely for you, not so rarely for others. There were major delays of some products because of that bug.
 
Pepsiennis said:
AMD is doing a lot of seeding if they aren't planning to put the 9900 out sometime soon. ;)
Click to expand...

I do believe they've learned a thing or two from fellow Texans and taken a page out of the Karl Rove PR book - set em up and knock em down. Easy as pie.
 
Pepsiennis said:
As for Q2'08 delay... then why is there a rush of Phenom 9900 previews yesterday and today?

Sharky extreme, sysopt.com, and Neoseeker. AMD is doing a lot of seeding if they aren't planning to put the 9900 out sometime soon. ;)
Click to expand...

If Intel can preview their QX9770 a quarter in advance, why can't AMD do the same? ;)

Not that the Phenom 9900 reviews do much to help their situation, but it's better than nothing.
 
You must log in or register to reply here.
Back
Top