SPCR

AMD Athlon 64 3000+ and 3200+ for socket 939 will soon be available, using 90 nm technology and the new D0 revision. They cost 173 and 208 USD respectively at monarchcomputer.com . This should give us cheaper 939 systems with lower power consumption (~50 W?).
I don't care much about dual channel memory, but with these ones I don't have to feel like I pay any extra for it. The price will go down (only 512 kB and smaller die overall = cheaper production) and socket 754 ones will become unefficient for us (need more power), as long as they don't start using 90 nm on those too.

I wonder what's going to happen with the mobile platform, will they start using socket 939 there too or what? If not, then mobile A64 might be the only way to get newer D0 ones for socket 754.

Related post by me ->
http://forums.silentpcreview.com/viewtopic.php?t=15760

Should mean higher overclocks as well, according to Ed. So, FX-55 130nm, or a 3200 @ 90nm... *thinks for eternity*

Of course I wasn't first with such big news!

Since athlon64's are supposedly just athlonxps with 64bit extensions and a memory controller, I wonder if there's a chance that this will lead to 90nm socketA chips. Sure would clobber celerons, at least in performance and value, unless AMD pulls a sempron again.

http://www.amd.com/us-en/Corporate/Virt ... 00,00.html

Here's a new mobile chip. No word on voltage or wattage, but I wonder if it's 35W or less...

Chris

About time too!
That is good news. I've been waiting for either this to happen, or prices of S939 A64s to drop. With any luck by xmas/new year, when I build my new system, both will have happened because I definately want S939 for future upgrade paths.

mathias wrote:Since athlon64's are supposedly just athlonxps with 64bit extensions and a memory controller, I wonder if there's a chance that this will lead to 90nm socketA chips. Sure would clobber celerons, at least in performance and value, unless AMD pulls a sempron again.

I was more under the impression they're derrived from the Opteron aren't they?
As I understand it, A64 = scaled down A64FX, and A64FX = Opteron (with a few changes, but overall it's the same processor I thought).

If i'm right I doubt 90nm Socket A will make it to market.
Another reason to doubt it is the rebranding of AlthonXP to Sempron, and Sempron is being advertised as the budget part. I just can't see AMD putting money into developing a more sophisticated fab setup for a chip which is, for all intents and purposes, on the way out.

Mats wrote:AMD Athlon 64 3000+ and 3200+ for socket 939 will soon be available, using 90 nm technology and the new D0 revision. They cost 173 and 208 USD respectively at monarchcomputer.com . This should give us cheaper 939 systems with lower power consumption (~50 W?).
I don't care much about dual channel memory, but with these ones I don't have to feel like I pay any extra for it. The price will go down (only 512 kB and smaller die overall = cheaper production) and socket 754 ones will become unefficient for us (need more power), as long as they don't start using 90 nm on those too.

I wonder what's going to happen with the mobile platform, will they start using socket 939 there too or what? If not, then mobile A64 might be the only way to get newer D0 ones for socket 754.

This is excellent news. Cheap socket-939 processors are just the thing that was needed. Next year there will be dual core processors that run on socket-939, so there is a very nice upgrade path. Smaller transistors means more leakage current, so the wattage drop might not be as big as it has been in earlier process transitions. But, we'll see soon.

AMD's processor roadmap tells that they are going to start producing all their processors with 90nm technology eventually. I'm pretty sure it includes all socket-754/939/940 processors. Who knows, maybe they will make some 90nm socket-A Semprons too.

Looks like the socket-939 3000+ runs at 1.8GHz and 3200+ at 2.0GHz.

meglamaniac wrote:About time too!
I was more under the impression they're derrived from the Opteron aren't they?
As I understand it, A64 = scaled down A64FX, and A64FX = Opteron (with a few changes, but overall it's the same processor I thought)

If i'm right I doubt 90nm Socket A will make it to market.
Another reason to doubt it is the rebranding of AlthonXP to Sempron, and Sempron is being advertised as the budget part. I just can't see AMD putting money into developing a more sophisticated fab setup for a chip which is, for all intents and purposes, on the way out.

AFAIK, athlon64fx is just the top of the line athlon64, and athlonMP's were practically the same as XP's. The reason they'd move thoroughbred/palomino to 90nm is it'll make them cheaper to produce. A64's are big and expensive to make, kind of like bartons, only moreso.

mki wrote: AMD's processor roadmap tells that they are going to start producing all their processors with 90nm technology eventually. I'm pretty sure it includes all socket-754/939/940 processors. Who knows, maybe they will make some 90nm socket-A Semprons too.

Usefull chart; it says they'll continue making mobile XP's until at least halfway through 2005, would be nice if they moved those to 90nm(even if they bring the cache back down to 256).

Usefull chart; it says they'll continue making mobile XP's until at least halfway through 2005, would be nice if they moved those to 90nm(even if they bring the cache back down to 256).

The time and cost it would take to move the XP line to 90 nm may not be recouped before they discontinue them in 8+ months.

Of course the transition might be quick and relatively cheap, and with the volumes they would deal with in the value market, it might be worth while

Moving bartons to 90nm probably wouldn't be worth it, but since semprons are also thouroughbreds, I'd assume it would make sense for them to move athlons back to thouroughbred, they clock the semprons low, so it would be real easy for them to pick out higher grade chips for high end XP's and mobiles.

mki wrote: Smaller transistors means more leakage current, so the wattage drop might not be as big as it has been in earlier process transitions. But, we'll see soon.

In this thread you can read about the power consumption (and my optimistic dreams about my next system..). So far I've seen that the new D0 ones will be 6 % faster and dropping from 62 to 35 W in power consumption in one gase (must be the 1.4 V mobile). While I do believe in the increased speed, I don't know what to think about those 35 W...

One guy in that thread thought that they just made some wrong conclusions with the 62 W mobile A64 and the one running at 35 W. I don't believe in that either, it would be just too simple. A lowered power consumption of that kind would be like when P3 went from 180 to 130 nm I think, don't know any numbers.

Mats wrote: In this thread you can read about the power consumption (and my optimistic dreams about my next system..).

If my "34% reduction" calculation in that thread is correct, the thermal design power of a 90nm Athlon 64 could be something like:
89W * 0.66 = 58.74W
That ~59W is the maximum, not necessarily the real heat dissipation.

The current 130nm Athlon 64 3000+ (ADA3000AEP4AX) has a 1.8GHz/1.4V Cool'n'Quiet P-state. In that P-state the TDP is 67W. If we use that to calculate the heat dissipation of the new 1.8GHz socket-939 3000+, it would be:
67W * 0.66 = 44.22W

I've also read that the TDP of the new 90nm socket-939 processors will be 105W. That includes the upgrade path to dual core processors. If the 90nm dual core processors put out 105W of heat, a similar 90nm single core processor would have TDP of:
105W / 2 = 52.5W

44-59W... It's very interesting to see the real thermal specifications for the new processors after this. I'd say that your earlier guess of 50W might not be far off.

Mki:
If the 90nm dual core processors put out 105W of heat, a similar 90nm single core processor would have TDP of:
105W / 2 = 52.5W

I strongly doubt that the dual cores will put out 105W. AMD seems to be quite determined to stay below 100W of heat. I think this has a lot to do with the reactions to consumer reactions to Intels very hot CPU versions.

And dual cores does not produce even close to twice the heat that a single core does, so your calculation with division by 2 will not be correct.
BTW, have you checked out Sun's new CPU (finished, but not in full production yet). Each core runs 4 threads at a time (kind of more advanced hyperthreading) and it has 8 cores, so it can run 4*8=32 threads simultaneously.
Read more at Jonathan Schwartz's blog:
http://blogs.sun.com/roller/page/jonath ... humans_and

Well, this is at least not what I wanted. The new one runs 4-5 degrees hotter than the old one...
I really hope that either me or the reviewer have missed something.

Here's another review, but it's really not interesting at all.

Mats wrote:Well, this is at least not what I wanted. The new one runs 4-5 degrees hotter than the old one...
I really hope that either me or the reviewer have missed something.

It might still be better, Thouroughbreds were harder to cool (at least thr. br. A's), but they were easier on the PSU, and they produced less heat that has to be pulled out of the case.

Mats wrote:Well, this is at least not what I wanted. The new one runs 4-5 degrees hotter than the old one...
I really hope that either me or the reviewer have missed something.

It gets even worse: Seems that the new 90 nm CPUs (at least 3000+ according to the AMDZone review) have a default voltage of 1,4V. This makes the temps lower - on my 3000+ (S754) going from 1,5V to 1,3V drops the load temps by ~10°C (from 57°C to 47°C, respectively).

Cheers,

Jan

ouch, this all sounds terrible, combined with the fact that they're more or less the same as the s754 CPUs (dollar for dollar at least, since they're all 200MHz slower than equivalent numbered s754 CPUs).
wonder if 130nm a64 prices will drop?

Straker wrote:wonder if 130nm a64 prices will drop?

Why should they? They got a larger die size >> higher production cost. And if there really is no big advantage with the new ones then the older ones won't get impopular. If they will, then it's another issue, but on the other hand there will be less 130 nm ones made.

Right now I'm really not into 90 nm because of the very few bad numbers I've seen (yeah, it could change, hope so), and I don't care about dual channel memory either. For these reasons I could go for a socket 754 instead. The "socket 754 becoming out of date"-thing is no problem either, since I believe that I will need a new motherboard anyway the next time I will upgrade. The only problem with Athlon 64 is that there's not enough good motherboards. amdboard.com have 138 socket 754 boards listed and only 31 socket 939 boards. What does this mean? Well, you can probably forget about 100 of those socket 754 mobos (old chipsets like nForce3 150 and K8T400 for instance, plus it seems to be more common with VGA on those 754 mobos compared to 939 mobos - I want a separate graphics card), although they're only one year old or less. Since 939 is so new, many of those mobos are more up to date.

In the end I'll probably buy 939 anyway, just because there are more newer motherboards for it. But there's still no rock solid chipset that has been around for a long time (yeah, it's unfair since A64 only been out for a year) like Intels i865PE

Too bad that AMD doesn't make their own chipsets (except for the Opteron ones). I'd say thats the advantage with Intel, worse CPUs - better chipsets (ok, I think they're pretty good CPUs, but not my favorites, PM excluded!). That's not anything new though.

Keep in mind that the 90nm parts have about a 40% smaller die. Therefore to keep the temps the same the amount of heat produced will have to be less. Give the much smaller die a 5C temp increase is very small and I would have expected much higher temps if the power consuption was the same. Given this it looks like AMD is not having the same problems with poer consumption/heat that Intel did with their 90nm parts. I think as we learn more about the 90nm parts that we will learn that power consumption is lower than same speed 130nm parts.

well, there's no reason the prices "should" drop, but if you want to be that way, you could also say the CPUs are made out of a few cents' worth of sand and should be sold in gumball machines.
thing is just that it doesn't make a whole lot of sense for socket 754 and 939 CPUs that are otherwise nearly identical to coexist... price aside i'd prefer s939 as well for the motherboard selection alone.

[H]ardOCP wrote:AMD Price Cuts, New Processors?
TechPowerUp and TheInq are both reporting on AMD price cuts coming in the near future. The dates on the two price cuts differ slightly, but they are close enough together to make it a viable thought on the rumor mill. TheInq also states that AMD will introduce the Athlon 4000+ and the FX55 processors as well.

The next price cut is not far away, it will be on October 18th, affected are:

AMD Athlon64 3400+, 3500+ and 3700+
Mobile Athlon64 (Low Power) 2800+, 3000+, 3200+ and 3400+
Mobile Athlon64 (Desktop Replacement) 2800+, 3000+, 3200+, 3400+ and 3700+
Mobile Sempron 2600+ and 2800+

This info is from TechPowerUp's note.

... FX-55... Athlon 4000... 90 nm... 939 pin... *drools over keyboard*... *hisses "MY PRECIOUSSSS...." *

acaurora wrote: ... FX-55... Athlon 4000... 90 nm... 939 pin... *drools over keyboard*... *hisses "MY PRECIOUSSSS...." *

Sorry, Gollum, but the latest spy reports from within Baradur indicate that the highest end AMD chips will be fabricated at 130nm. It appears that the orcs at Mount Doom cannot yet forge chips to rule over all chips at 90nm.

http://www.xbitlabs.com/news/cpu/displa ... 72101.html

Ah... my PRECIOUS... NOOOOOOOO *leaps into the fiery depths of Mount Doom (3) *