New AMD Dual Cores New News

[geotech]

http://www.dvhardware.net/article24134.html

Eh?

 

[jimmysmitty]

I just hope that the L3 cache for these don't have the TLB bug. Or maybe that just affects the Quad cores.

 

[Evilonigiri]

About time AMD started to phase out the 90nm cpus. It has served it's time real well already.

 

[OlSkoolChopper]

Unless the Athlon 6250 and 6050 are going to be 3.2 and 3.0 GHz parts and sell for under a hundred bucks, they're only going to sell to the Ruiz family.

 

[Reynod]

So they are going to stop selling their best (90nm 2Mb A64 X2) chips?

Jeez boys, better snap them up now.

 

[tlmck]

This can be read as "We do not know how to make a new processor, so we will just slap some more cache on the old ones".

Wake me up when they get to 45nm.

 

[Kaaliya]

wont the L3 bug affect the dual cores as well ?

 

[spuddyt]

I can't see why it wouldn't

 

[justinmcg67]

I think the TLB bug specifically addresses four cores rather than two; even tri-cores should be immune to the errata. If the bug targets quad-cores than disabling specifics parts of that processor, although on the same die, should in theory disable the bug. I expect the dual core and tri-core K10's to be a huge hit, but not so much with the quad cores.

I want a tri-core or dual-core K10, no doubt about that. If priced reasonable and has competitive results towards the mid-range Core 2 Duos than AMD has a sale from me. Lets just hope that pricing is good and performance is on-par with mid-range. It doesn't have to be the fastest, it just has to be marketed in my segment and that'll guarantee them a sale.

 

[Grimmy]

<--sings

"And all that I want from tomorrow.. Is to get it better today..
Step By Step..
One By One..
Higher and Higher..
Step By Step..
One By One..
Climbing Heccc... *cough**cough*Jacob's Ladder!!"

 

[gpippas]

AMD will launch a new series of dual-core products based on its next-generation architecture, Athlon 6000, in the later half of second quarter next year. First to launch will be the Athlon 6250 and 6050, the sources revealed. Although the core frequencies of the two CPUs have not been set yet, they will adopt 1MB L2 cache and 2MB shared L3 cache, the sources added.

I don't think people will be very impressed if in 6 months time AMD can't get rid of the TLB errata. These can also only be considered budget cpu's. Seeing that the Stars arch is so far off Core clock for clock I don't see it being very competitive. The most interesting prospect is how much faster will it be than K8. For AM2 it owners it might be a nice upgrade path.

I'm also slightly worried about this timeframe. Because due at the same time as these 65nm dual cores are the 45nm quads from AMD. If it is going to take that long to release dual core K10's which are just failed Phenoms I don't see them getting a die shrink to 45nm done on time.

 

[Reynod]

Hope they checked the FPU ... don't want that maths problem again eh Paul ??

4195835.0/3145727.0 = 1.333 820 449 136 241 000 (Correct value)
4195835.0/3145727.0 = 1.333 739 068 902 037 589 (Flawed Pentium)

 

[systemlord]

The reason Intel put two Core 2 Duo's together to form a Core 2 Quad is becuase it was far easier to manufacture them. AMD has found this out the hard way for the release of their Quads cores which is why there not clocked anywhere near what AMD wanted, now AMD and Intel are having to deal with these small holdbacks. It will take them both time to try and figure out how to get the highest quality out of them.

 

[Reynod]

The core 2 chips also are a generation ahead with the cache and prefetchers ... that's why the fsb isn't an issue with single socket systems.

AMD's cache structure with the L3 only pays off for multi-socket systems ... large servers.

The K10 from a single user perspective is really a K8 with a bit of fine tuning ... the L3 is really a waste and frankly slowes the system up.

They would have been better off expanding the L2 cache ... however they are not as good at making L2 cache arrays as Intel is I guess.

I could be drunk on air though ... don't mind me dribbling here in the corner.

Check the latency figures Scott Wasson has put out ...

I can't see the dual core K10 being anything spectacular then ... unless the speed ramps quickly.

With all of the extra transistors on the K10 maybe they have some "additional" features ... who knows.

 

[jonisginger]

All they are trying to do is to sell the last of the A64 cores off cheaply and attractively...

 

[spuddyt]

All they are trying to do is to sell the last of the A64 cores off cheaply and attractively...

did you read the article? they have "stopped taking orders" for said athlon 64, these will be new ones, its not like they are simply pulling out ones that have been in stock for ages.....

 

[LukeBird]

I think the TLB bug specifically addresses four cores rather than two; even tri-cores should be immune to the errata. If the bug targets quad-cores than disabling specifics parts of that processor, although on the same die, should in theory disable the bug.

I'm sure you're right.
I'm sure the TLD bug is probably only going to affect the quad-cores.
Be interesting to see how dual & tri K10's perform

 

[pathan7]

I am wondering where and at what price segments will AMD place dual and tri-cores? because seeing that AMd will also release low end quad-cores 9100e (9150e with B3 stepping) will be priced rather low. Any one any ideas?

 

[pathan7]

http://www.xbitlabs.com/news/cpu/display/20071226214456_Advanced_Micro_Devices_Faces_Further_Delays_of_Phenom_Processors.html

Seems like from this article that the tri-core (AMD Phenom 8600 and 8400) MAY have TLB errata as they are based on B2 stepping, also the highest clock speed Tri-core will have is 2.40GHz

 

[Slobogob]

Of course the tricores will have the TLB bug. The only thing the core count affects is the probability of the bug actually occuring. Unless they deactivated a portion of the processor which in turn would have severe performance penalties.

 

[sailer]

This is really nice. I mean, AMD is dumping some of its best selling, highest performing processors so that in 6 months to a year they can sell unproven chips based on a design that has so far been a failure. Is AMD that embarrassed by the fact that these older design chips outperform their new Phenom? And that 6 months to a year timeframe, like when has AMD been able to release anything on time lately? They could end up with no chips available to sell if the Phenom doesn't get fixed in a hurry.

It looks to me that if someone wants an AM2 setup, better buy now. And as for the idea of buying a 790FX board and sticking an AM2 chip into it until the Phenom gets fixed (which I've had), I've read that there are some big problems with the 600 south bridge. Until the new 700 south bridge comes out, it might be best to avoid those boards.

 

[spuddyt]

Unless the Athlon 6250 and 6050 are going to be 3.2 and 3.0 GHz parts and sell for under a hundred bucks, they're only going to sell to the Ruiz family.

you'd think they could afford the quad cores... with his pay rise

 

[OlSkoolChopper]

you'd think they could afford the quad cores... with his pay rise

Yeah, but even his family isn't dumb enough to buy a Phenom. :lol:

 

[I800C0LLECT]

I've read that there are some big problems with the 600 south bridge. Until the new 700 south bridge comes out, it might be best to avoid those boards.

Originally(2 years ago) I heard that the SB600 had issues with USB transfer rates. That's why they were attempting to use a ULi southbridge for a short term solution. However, that was sometime ago and they were supposed to have fixed that issue.

The SB600 is getting long in the tooth but I believe it's operating just fine now days. Correct me if I'm wrong, it happens

 

[sailer]

The SB600 has worked fine for quite a while. The problems seem to be more related to loading Vista on the 790FX chipsets. There were a couple workarounds, but they weren't very good. I'll try to look up the article and post it. The proposed solution was the SB700, but its possible that there is some other design problem as well. The tested motherboards were from ASUS, Gigabyte, and MSI, so it seems to be more than a single company making an error, pointing to a systemwide fault.

I found the review that details the SB600 problems:

http://techreport.com/articles.x/13832

From the article, though the whole aticle should be read for clarification:

AHCI Issues

A documented compatibility problem between Windows Vista and the 790FX chipset's SB600 south bridge component currently complicates running drives in AHCI mode.

That's if it works at all. One can install Vista by putting the SATA controller into IDE mode, but AHCI mode is necessary to take advantage of Native Command Queing and SATA device hot swapping, so its not a trivial capability to loose.