![]()
| #45. Posted at 03:58 PM on Jun 15th 2006 | Edit Reply |
|
Vaughn |
hey as long as can still find a S939 170 Opteron 6 months from now it matters not to me. that is gonna be my next chip to replace my 146.
|
![]()
| Edit Reply |
|
just brew it! |
I don't see this as a big deal. Anyone who really cares about 1MB cache can still opt for 1xx series Opterons.
|
![]()
| Edit Reply |
|
axeman |
Conroe has more execution resources than K8. It also suffers from higher memory latency because of the off-die memory controller. So it likes big caches. It has been clear from the inception of the K8 that large amounts of L2 make little to no difference outside of marketing. Heck, even the 128k Semprons perform admirably for such a tiny amount of L2.
|
![]()
| Edit Reply |
|
AKing |
I'd better get a 2mb X2 CPU before they stop producing them then. I dont like the sound of this, as i like more cache (who doesnt?).
|
![]()
![]()
| Edit Reply |
|
Chrispy_ |
PC makers who live for strong feature checklists
These will be the sort of vendors I love to hate anyway, selling PC's by bigger numbers as if the e-penis is all that matters. ... 9600XT > 7900GS because the number's higher and there's an X. 256MB graphics!! (never you mind that it eats half your system memory and doesn't do squat....) 27.1 digital stereo surround sound (but we're giving you a pair of cheap plastic speakers not even fit for windows chimes and error beeps) 130-key keyboard with at least 25 keys that get in the way and you will never use. (9 button mice suffer the same problem) I liked it when Cyrix came along and pretended their 110MHz socket 7 chips were pentium MMX 233 beaters - After all, 233+ is sureley better than Intel's plain-old 233, right? |
![]()
| Edit Reply |
|
wierdo |
ok so in other words, X2 will become a consistent term referring to 512kb cache variants on AM2 for now, and the Opteron 1XX models that will fit on AM2 sockets and have 1MB cache will absorb the redundant models...
I think they should have done this from the start instead of creating confusing ratings based on combinations of cache vs clockspeed improvements. |
![]()
| Edit Reply |
|
green |
AMD takes the "less is more" attitude
Intel takes the "more is more" attitude i see this as a cost cutting measure more chips per wafer for the price war expect x2's to come back with 1mb cache in a year or so when they're shunk to 65nm.... actually by then it's likely to be x4 |
![]()
| Edit Reply |
|
Proesterchen |
Great move, should earn me a couple of extra Euros when I'll be selling my Opteron in the near future. ("Full 1MB L2 cache per core, don't take anything less!", "Say no to AMD's castrated X2s!") :-)
|
![]()
| Edit Reply |
|
LiamC |
AFAIK, the 1MB cache chips have three working HT links, they just aren't used. You wouldn't want people to plug just any old chip into their 4x4 platform now would you ;)
The 512KB cache chips use a different mask and have only one HT link. |
![]()
| Edit Reply |
|
BobbinThreadbare |
Fewer transistors should mean cheaper chips, and higher clocks, right?
Sounds like a good thing to me. |
![]()
| Edit Reply |
|
blastdoor |
I think some clarification regarding total cache and cache-per-core is in order here. The 2MB and 4MB figures for Core2 refer to total, while I'm assuming that when you say 512k for the X2s, you mean 512k per core, not total, right?
|
![]()
| Edit Reply |
|
Dissonance |
Updated to add clarification about Socket 939 4800+.
|
![]()
| Edit Reply |
|
Krogoth |
AMD is merely shifting their 1M L2 dual-core K8s to be converted into Opterons and a couple cherry-picked samples to be used as FX chips.
|
|
Jazztags: (they MUST be closed) r{ red }r g{ green }g /[ italic ]/ *[ bold ]* _[ underline ]_ -[ |