backtop

Hardware AMD Preps 2x256KB L2 Athlon 64
Kristopher Kubicki (Blog) - June 29, 2006 1:18 PM
Print E-mail del.icio.us 70 comment(s) - last by fsardis.. on Jul 7 at 6:56 PM

CPU-Z shot of the Athlon 64 X2 3600+ engineering sample
AMD: Ditching more cache for less, since 2006

DailyTech has managed to get a hold of an AMD engineering chip labeled "Athlon 64 X2 3600+ for AM2."  According to CPU-Z, the processor is a 2.0GHz CPU, the same as the Athlon 64 Windsor 3800+.  However, half of the L2 cache has been disabled effectively making this AMD's first dual-core processor with a total of 512KB of L2 cache.

AMD recently discontinued all of its 2x1MB L2 cache processors less than two weeks after the processors were announced.  This may be an indicator that low-bin Athlon 64 X2 processors will instead transition to 2x256KB cache configurations, instead of the 2x512KB parts that were previously the low-bin.  However, since the CPU is only an engineering sample, it’s not easy to determine if AMD will even publically announce the CPU, or if it will simply become an OEM-only component for high quantity system builders like the original Socket 939 Semprons.

CPU-Z shows the Athlon 64 X2 3600+ to still sport the Windsor CPU core, the same CPU core that was used for the 2x1MB and is still used for the 2x512KB AM2 processors.

HKEPC has also come across the 3600+, claiming it will ship in Q4'06 for a price of $139 to $159.  Given that AMD is expected to announce major CPU price cuts near the end of July, the $139 price tag fits in with the rest of AMD's pricing.   Intel's lowest-bin Conroe, the 1.86GHz Core 2 Duo processor, is expected to ship with a price of around $183.  In contrast, the 3.0GHz Presler NetBurst processor is expected to hover around $133 at the same time HKEPC claims the X2 3600+ may ship.

Comments     Threshold


This article is over a month old, voting and posting comments is disabled
2x1kb?
By collegeguypat on 6/29/2006 2:05:40 PM , Rating: 2
"the 2x1KB and is still used" shouldn't it be 2x1MB? j/w




RE: 2x1kb?
By bobdelt on 6/29/2006 8:46:29 PM , Rating: 3
I'm sure they are doing this for heat reasons. Less cache=less heat. That way they can still say their chips produce less heat than intels.


RE: 2x1kb?
By Scrogneugneu on 6/29/2006 9:05:56 PM , Rating: 3
Yeah, right...

I have another idea : how about "cache takes a lot of room", and cutting it in half cuts the cache production costs by half, as well as reduce the size of the final chip, therefore increasing chips per wafer ratio as well as yields, which ultimately leads to lower production costs overall, enabling them to enter a price war with Intel without suffering too much?


RE: 2x1kb?
By bobdelt on 6/30/2006 7:01:25 AM , Rating: 2
Well I didnt say it was the only reason. But AMD did just come out and say that there chips will always have a lower TDP, including the quad cores. When AMD has 4 cores and 1mb of cache and Intel has 4 cores and 4-8mb of cache, of course AMD's chips will produce less and heat and will be cheaper.


RE: 2x1kb?
By GoatMonkey on 6/30/2006 10:33:14 AM , Rating: 2
The cache is not removed, just disabled. So it means they're just padding their corporate bank account on the other chips.


RE: 2x1kb?
By JackPack on 6/30/2006 1:25:34 AM , Rating: 2
LOL. If that were the case, then there wouldn't be any 1 MB Turions.


RE: 2x1kb?
By miahallen on 6/30/2006 6:58:54 AM , Rating: 4
Nobody seems to have noticed that in the screenie, this chip is running on only 1.1V...the X2 3800+ runs on 1.3/1.35v, this is awsome, and will allow this chip the headroom it needs for substainial OCing.

If the performance hit for 1/2 cache is as insignificant as the 2x512KB vs 2x1MB contest, I'd say this chip (at that price) would be a VERY hot seller! Go for it AMD!!!


RE: 2x1kb?
By plewis00 on 6/30/2006 7:52:17 AM , Rating: 2
I wouldn't take that for given. The Conroe ES chips which came out were showing 2.7V+ on some CPU-Z screenshots. You are right, it may well run at 1.1V which would make it very low power as the voltage increases, the wattage increases by powers of the voltage (I think if I remember right). It might just be a discrepancy of it not being a chip which is currently available. We will see...


RE: 2x1kb?
By nikbaj on 7/3/2006 11:13:23 AM , Rating: 2
Eh, the reason that Intel Chips run faster on certain applications with large matrix sizes involved is because of larger cache--for example in MATLAB I have a program that runs about 50% faster on a 2.8 GHz P4 HT than on my A64 X2 3800+ @ 2.5 GHz, despite A64's greater efficiency per clock. The P4 is Prescott core with 2 MB cache, and that's what makes the difference.


RE: 2x1kb?
By fsardis on 7/7/2006 6:56:24 PM , Rating: 1
sorry but you are wrong. if anything, more cache will decrease the total power requirements of the system and thus reduce the heat and power consumption. accessing ram takes more energy than accessing cache.


Athlon? No, Sempron
By sabrewulf on 6/29/2006 2:29:50 PM , Rating: 2
Based on past parts, naming, and marketing by AMD, I don't see how they could call this an Athlon if it makes it into retail. a single core K8 with 256k is and always has been a Sempron, why whould putting two of them together suddenly make it an Athlon?

Seems to me that AMD would be watering down their own model names if they did something like this.




RE: Athlon? No, Sempron
By cnimativ on 6/29/2006 2:39:21 PM , Rating: 1
It uses the same core as the rest of the AM2 Athlon X2s, so of course its an Athlon.

Sempron has different cores (Orleans)


RE: Athlon? No, Sempron
By plewis00 on 6/29/2006 2:47:57 PM , Rating: 2
Yes, well a Pentium M Dothan core (2Mb cache) with half its cache disabled is still a Dothan core but with 1Mb cache and goes with the Celeron M name instead. So no, it doesn't have to be an Athlon does it? Cores don't determine the processor name, the company who manufactures them does.


RE: Athlon? No, Sempron
By visaris on 6/29/2006 3:39:48 PM , Rating: 2
The sempron name implies low end tech. As this is a dual-core part, it really isn't at the low end of the product line, and I think the Athlon name is appropriate.


RE: Athlon? No, Sempron
By Scrogneugneu on 6/29/2006 9:09:17 PM , Rating: 2
You do know that Celeron were created basically to sell all those chips on which the cache had problems and couldn't be sold as fully functionnal Pentium, right?

Therefore, a Celeron was at the beginning a Pentium, without any (or with significantly smaller) cache. I don't know if over time Intel developped an architecture specifically for the Celeron but i doubt it.


RE: Athlon? No, Sempron
By coldpower27 on 6/29/2006 10:26:13 PM , Rating: 2

Well Intel is rich enough that they can just disabled cahce on Banias, Dothan, all iterations of Pentium 4 besides Prescott-2M, to just make the Celeron versions of it.

AMD did this as well with the Palermo core Sempron core, it's still a disabled Venice.


RE: Athlon? No, Sempron
By Viditor on 6/30/2006 1:54:25 PM , Rating: 2
quote:
Well Intel is rich enough that they can just disabled cahce on Banias, Dothan, all iterations of Pentium 4 besides Prescott-2M, to just make the Celeron versions of it

Actually, it makes a better profit by allowing them to use chips with cache errors from manufacturing and thus improving their yields...otherwise they would have had to throw the dice away.


RE: Athlon? No, Sempron
By coldpower27 on 7/2/2006 12:31:10 AM , Rating: 2

I guess I should have clarified, I mean that Intel has enough capacity that they can supply that channel solely on using disabled cores without the need to create a physical smaller die without the cache on it in the first place.


RE: Athlon? No, Sempron
By plewis00 on 6/30/2006 9:13:41 AM , Rating: 2
Yes, I am well aware of that, possibly and also to provide another breed of chip without producing another core. So some will be chips with parts of the cache bad and others will be chips which are binned to provide the demand for Celeron. I assure you Intel wouldn't be doing that if they were losing money, once the initial R&D has been invested the actual silicon production is cheap (relatively speaking) for them.

But, if you had read previous posts rather than just blurting out, you would have seen that I was simply proving a point responding that because the core was originally an Athlon, Sempron, Pentium or whatever, doesn't mean any future derivatives of it must be one.

And no, Intel never developed a Celeron Pentium M architecture, the chips were always Pentium Ms with half the cache switched off and Speedstep disabled.


RE: Athlon? No, Sempron
By Ypub on 7/1/2006 8:10:37 AM , Rating: 2
I am agree with you. Like all others processors sempron and celeron have evolution with time... This is a sempron.
We know the performance of the conroe. This performance are displayed all over the web and even, if tomshardware respect the nda, some other great sites break it cause conroe ES are available on all the web and all the things are going to be said before...
For exemple, those sites publish results of tests that are not the intel one :
http://www.hardware.fr/articles/623-1/dossier-inte...
http://www.matbe.com/articles/lire/306/merom-et-co...

and the E6300 (183$) is always better than x2 4000+, generally over 4400+, and a lot of time over 4800+

So How could this 3600+ (139$) something else than a New sempron? This just a commercial name for some afficiendos


By m666guy on 6/29/2006 3:53:05 PM , Rating: 2
First they cut the 1mb parts almost completely and now they are beginning to introduce lower-end parts. While i think that it is good they are finally targeting a more mainstream audience with their dual cores. It makes me wonder though. How is it Intel is able to pretty much give away their processors? are they not making any money on the low-end pds?