35 Watt X2 3800 on the way from AMD

afrost · Post by **afrost** » Mon Apr 10, 2006 8:23 am

http://www.anandtech.com/cpuchipsets/sh ... i=2738&p=2

This will be a nice candidate for silent systems....I wonder how much it will cost

Post by **MikeC** » Mon Apr 10, 2006 9:17 am

AMD processors are already very easy to run virtually silent. I don't think lower noise is a compelling reason to switch to AM2. There are other reasons, like lower power consumption, support for DDR2 or (hopefully, despite Anantech's doubts) better performance.... but lower noise? Not IMO, unless the new chips are so cool that they can be run fanlessly.... which I doubt.

afrost · Post by **afrost** » Mon Apr 10, 2006 10:18 am

huh....that's an interesting take.

I think I am often interested in even tiny incremental improvements because I am in the weird "quiet overclocker" niche. I want to squeeze every bit of performance out of my system for gaming....but keep it nice and quiet at the same time.

So I am kind of interested in the 35W processors to see if they can get to a higher stable clockspeed than other processors given a specific heatsink and fan.....like say a ninja with nexus at 7V.

maybe i'm only dreaming and it's really just marketing hype

Hifriday · Post by **Hifriday** » Mon Apr 10, 2006 10:29 am

There's no need to wait, it seems that dual core Opterons 165/170 are already shipping with 35.0W TDP rating. Of course there is no guarantee since the TDP is variable, but based on the postings at TCaseMax forums seems most of these chips are 35.0W.

Although AM2 may not offer much in terms of silencing, nor much in terms of performance (yet), maybe we'll be lucky and it'll help drive down prices of s939!

Aris · Post by **Aris** » Mon Apr 10, 2006 10:35 am

gamming performance of current athlon processors at a TDP of a mobil pentium M.

i cant believe you guys arnt going crazy over this thing. it seems to be the best of both worlds to me.

jaganath · Post by **jaganath** » Mon Apr 10, 2006 11:31 am

gamming performance of current athlon processors at a TDP of a mobil pentium M.

Also probably at the price of a top-of-the-line pentium M. This is not a part for anyone on a budget (ie me).

frostedflakes · Post by **frostedflakes** » Mon Apr 10, 2006 11:58 am

35w is a very impressive TDP for an X2 3800+ IMHO. Maybe my X2 3800+ is a bum undervolter, but I'm only able to get it down to 1.23V (which is ~49w TDP, based on the original 59w TDP @ 1.35V from AMD64 TCaseMax). I'd assume these 35w X2 3800+ will actually be dual-core Turions, as the only way a Manchester or Toledo core could have a 35w TDP is if Vcore was ~1.05V. Such a low voltage is probably not possible or very difficult to achieve on a Manchester or Toledo, it just makes more sense to me that they would use a lower-power Turion64 X2 core at around 1.1V (which I believe is what the 35w Turion64 X2 are rated for).

Also, if pricing is anything like the Turion64, lower wattage models will only cost a few percent (usually comes out to $5-10) more.

Then just imagine when AMD shrinks to 65nm... 20w dual cores? Exciting stuff.

Post by **MikeC** » Mon Apr 10, 2006 12:10 pm

Aris wrote:gamming performance of current athlon processors at a TDP of a mobil pentium M.

i cant believe you guys arnt going crazy over this thing. it seems to be the best of both worlds to me.

Here's why I'm not going crazy about it:

That X2-4800+ featured in my recent CPU power article... I've now got it running undervolted on another board w/ an 80+ PSU. At idle, system power is 36W. W/CPUBurn x2, it's at 107W. Avg power will be 43W if you go by the 90% idle/10% max formula I used. It can be cooled by any number of HSF at min noise.

stupid · Post by **stupid** » Mon Apr 10, 2006 12:23 pm

jaganath wrote:
gamming performance of current athlon processors at a TDP of a mobil pentium M.
Also probably at the price of a top-of-the-line pentium M. This is not a part for anyone on a budget (ie me).

Unlikely if the price difference between the Turion 64 MT & ML series are any indications. The lower TDP version will likely be only about $10 more than the 89w X2 3800+, or 110w X2 4200+ thru 4800+.

jaganath · Post by **jaganath** » Tue Apr 11, 2006 2:08 am

Then just imagine when AMD shrinks to 65nm... 20w dual cores?

Aren't the low voltage Core Duo already there? Duo LV-L2400 has a TDP of 15W.

afrost · Post by **afrost** » Tue Apr 11, 2006 6:30 am

jaganath wrote:
Then just imagine when AMD shrinks to 65nm... 20w dual cores?
Aren't the low voltage Core Duo already there? Duo LV-L2400 has a TDP of 15W.

X2 3800 would blow it away for gaming though

......yes i'm a fanboi

Aris · Post by **Aris** » Tue Apr 11, 2006 7:30 am

i see what your saying mike, but honestly i think your missing the "BIG PICTURE" here.

sure you can probably get current products, that are just as fast or very close to it, down to a virtually silent operation. but this is a momentus shift in types of processors being released. a modern processor, on the mainstream pinout, with good performance at a mobil processor TDP output rating. its an amazing step IMO. its definatlely a step in the right direction.

dont dog it just cause it may not be perfect for you. rather look at how things may look in a couple of years if things keep progressing in this manner. very exciting stuff IMO.

Post by **MikeC** » Tue Apr 11, 2006 8:03 am

Aris wrote:i see what your saying mike, but honestly i think your missing the "BIG PICTURE" here.

I think you missed the gist of what I wrote.

The original poster wrote...

This will be a nice candidate for silent systems

I was responding to that comment... and suggesting that there are many reasons for being interested in the new CPU, but making it run quieter is not really one of them. Surely you must know by now that I am very keen on power efficiency gains, and have been since my very first PSU review going back almost 4 years. What do you think much of my last couple articles have been about? ie -- The State of the Industry, Mar 2006; and the Desktop CPU Power Survey.

CPU power consumption is dropping -- for now. However, Intel & AMD will will continue competing for performance gains, and it's a no-brainer that the power envelope will start to rise once again as clock speeds rise, and as more cores are added to the mix. Hopefully, this will affect only the cutting edge and not all mainstream processors.

TomZ · Post by **TomZ** » Tue Apr 11, 2006 8:55 am

MikeC wrote:CPU power consumption is dropping -- for now. However, Intel & AMD will will continue competing for performance gains, and it's a no-brainer that the power envelope will start to rise once again as clock speeds rise, and as more cores are added to the mix. Hopefully, this will affect only the cutting edge and not all mainstream processors.

I couldn't agree more. For most people, this forum's readers notwithstanding, the amount of power dissipated by the CPU is not known and not interesting. It only becomes an issue in some circumstances, like building silent rigs, or when you have to pay the power bill for a large quantity of computers.

More power will always give more performance, either in terms of higher operating frequency, more cache, more cores, more parallel processing of instructions, etc. But CPUs these days are also fast enough that for most people, a 50W CPU is already more than enough speed.

jaganath · Post by **jaganath** » Tue Apr 11, 2006 9:11 am

More power will always give more performance

Hmm, I'm not sure I would phrase the power-performance equation in quite that way; for example, Prescott required/outputted more power, but the increase in performance over the Northwood core was not proportional to the power increase. Also with Core Duo and Conroe, these seem to be both more performing than their predecessors, while also requiring less power. So a more accurate phrasing would be that higher performing CPU's generally require more power, but not always (as there are exceptions to every rule).

TomZ · Post by **TomZ** » Tue Apr 11, 2006 9:16 am

jaganath wrote:Hmm, I'm not sure I would phrase the power-performance equation in quite that way

What I meant was, for a particular CPU design. Of course when you compare between different designs, or different technologies, that will not hold true. But if you start with a particular CPU that uses say 35W, that means you've left another ~100W or so of performance available.

I'm also not saying that there is any linear relationship between power and performance, either. I think in a lot of instances, there will be diminishing marginal returns. So in my example, maybe going from 35W to 135W, you might be able to only increase the performance by 40%, for example.

accord1999 · Post by **accord1999** » Tue Apr 11, 2006 9:34 am

afrost wrote:
jaganath wrote:
Then just imagine when AMD shrinks to 65nm... 20w dual cores?
Aren't the low voltage Core Duo already there? Duo LV-L2400 has a TDP of 15W.
X2 3800 would blow it away for gaming though

......yes i'm a fanboi

Who needs low-voltage, the standard voltage top-end 2.16 Core Duo is already capable of sub 20W full load performance:

http://www.silentpcreview.com/article313-page5.html

afrost · Post by **afrost** » Tue Apr 11, 2006 9:45 am

re: yonah

now all we need are some good core duo motherboards.....

the 1.8 GHz actually isn't that expensive at $300.....with a motherboard that can overclock I would consider it.

shadestalker · Post by **shadestalker** » Tue Apr 11, 2006 10:04 am

Details seem hard to come by still, but maybe AM2 will be the platform that their Pacifica virtualization finally rolls out on. I'm keen to install 2G of RAM and run linux and Windows side by side.

smilingcrow · Post by **smilingcrow** » Tue Apr 11, 2006 11:48 am

I find it pretty easy to cool my X2 3800+ (~48W @ 1.15V) silently, but making the whole system quiet is not so easy. So I agree with MikeC in that it’s not such a big deal from a quiet perspective.
I’m using a Midi Tower case though and with the seemingly increasing interest in SFF and beyond cases (Apple anybody), saving 13W (using my example) BEFORE under-volting does seem more significant. Does anyone know of a direct comparison test between Midi and SFF cases both using the same components that looks at noise and temps? They would probably need to use different power supplies for obvious reasons.
I’m assuming that people will be interested in buying SFF PCs ONCE they are marketed by the large manufactures and retailers. After all, people have lapped up TFTs even though they are more expensive and generally give a poorer performance.

It’s good to see 35W X2s on the horizon, but is this another example of AMD not being able or willing to deliver low power parts in large quantities a la Turion MTs!
I really appreciate AMD but ironically I am feeling frustrated by the way that they keep on improving the power efficiency of their parts OVERALL, but not in such a way that it is easy for the end user to get their hands on them. i.e. there does seem a particularly large variance in power efficiency of CPUs for a given model number and no way to differentiate between them.
With AMD being the only game in town for low power desktop chips for ages now, I don’t think that this has really hurt them. But with Intel back in the hunt soon, this might well become an issue, at least for me. I don’t want to buy a particular CPU in the hope that it will come in at the lower end of a large wattage range. I’d rather buy a known quantity.
Unfortunately, with Intel and AMD both only publishing maximum TDPs for a range of CPUs, CPU buying will probably continue as a crap shoot, unless you buy a laptop chip where the variance becomes much less significant, due to much lower max TDPs involved.

35W at 90nm does hopefully bode very well for the 65nm shrink. Bill Gates envisioned a PC in every home (just one Bill! or maybe just one running Windows

); I have a dream, a PC in every home that is environmentally friendly and ultra quiet. So maybe every last watt does count?

jaganath · Post by **jaganath** » Tue Apr 11, 2006 12:11 pm

After all, people have lapped up TFTs even though they are more expensive and generally give a poorer performance.

Poorer performance for high-fidelity gaming, perhaps. For ordinary office and home use the difference is not so critical. Also, TFT's are more energy-efficient and space-saving than CRT's, which in Europe with its cramped housing and high energy prices, are big selling points.

stupid · Post by **stupid** » Tue Apr 11, 2006 12:23 pm

smilingcrow wrote:It’s good to see 35W X2s on the horizon, but is this another example of AMD not being able or willing to deliver low power parts in large quantities a la Turion MTs!

Let's be fair to AMD. They never had any intentions of building up a desktop platform around the Turion 64. It is strictly a laptop component as far as they are concerned. That's why only OEMs were initially offered; I think there are retail versions on sale now, but they are most likely for upgrading laptops. Of course it can also be argue the AMD may be testing the general market to find out if there is actually any interest in using it for a desktop.

I am assuming that AMD is releasing the 35w Athlon 64 X2 directly at the desktop market so these CPU should show up at online stores somewhat soon. This may well be the last harrah for S939 since the entire line is to be phazed uot this year. As I stated in a prior post, I guess these CPUs will have around a $10 premium over "normal" Athlon 64 X2.

At little off topic, but i would like to see GPU power consumption to decrease, I mean come on 135w for a Radeon X1900XTX? Yeah, it's a powerhouse of a GPU, but 135w? I think nVidia has been making headway, their 7900GT 256MB is more powerful than the 7800GTX 256MB, but consumes 9w less. It's not a huge drop, but it's a step in the right direction.

smilingcrow · Post by **smilingcrow** » Tue Apr 11, 2006 1:26 pm

jaganath wrote:
After all, people have lapped up TFTs even though they are more expensive and generally give a poorer performance.
Poorer performance for high-fidelity gaming, perhaps. For ordinary office and home use the difference is not so critical. Also, TFT's are more energy-efficient and space-saving than CRT's, which in Europe with its cramped housing and high energy prices, are big selling points.

I left out the most important part of my argument, which is that I think that most people buy TFTs PURELY because they have a SFF; hence people will likely be interested in SFF PCs also. Admittedly, not to the same degree, as most people probably have their PC on the floor rather than on their desk, but I still think it is a compelling reason. There is a general trend in life towards miniaturisation I would suggest, apart from in certain areas such as pornography and 4x4 off roaders/trucks.

As for the image quality of TFTs, CRTs generally are MUCH better for video usage as well as having massively better viewing angles and better colour representation. I use a Hyundai (L72D) 17” TFT with a DVI connection, which is probably better than the average TFT that ships with most systems. I recently sold a Philips 15” CRT that was 5 years old for only £10, and it still had better image quality than my Hyundai in most cases. The market for CRTs is pretty dead and I think that is purely down to their size.
I personally like the fact that TFTs tend to be silent (many CRTs are not), have a SFF, produce little heat and are environmentally friendly. But, from experience, I don’t delude myself that other people buy them for the same reasons. I’d liked to be proved wrong though.

stupid wrote:Let's be fair to AMD. They never had any intentions of building up a desktop platform around the Turion 64. It is strictly a laptop component as far as they are concerned. I am assuming that AMD is releasing the 35w Athlon 64 X2 directly at the desktop market so these CPU should show up at online stores somewhat soon. This may well be the last harrah for S939 since the entire line is to be phazed uot this year.

I wasn’t intending to refer to the usage of Turions on the desktop, but the fact that AMD released 2 different mobile CPU ranges and that the MTs didn’t seem to be that available within laptops or in the retail channel as discrete chips. It does make me slightly cynical about their abilities to deliver low power parts in quantity. Again, I REALLY would like to be proved wrong. I do find it curious though; they release desktop parts whose power ranges vary very considerably for the same part number, but, they go out of their way to differentiate between 2 mobile CPU ranges with a fairly small power difference in absolute terms, but admittedly rather large in relative terms. If Intel’s Centrino brand hadn’t been so spectacularly successful, I wonder whether AMD would have bothered to release MT and ML. It does look a bit desperate to me.
I would rather think that this is purely due to marketing bullshit, which although I hate this with a vengeance, it is still preferable for me than the thought that AMD can’t deliver. For the record, I haven’t bought Intel on the desktop since purchasing a Northwood 1.6A in 2002; I’m no Intel fanboy.

smilingcrow · Post by **smilingcrow** » Tue Apr 11, 2006 2:03 pm

shadestalker wrote:Details seem hard to come by still, but maybe AM2 will be the platform that their Pacifica virtualization finally rolls out on. I'm keen to install 2G of RAM and run linux and Windows side by side.

AM2 will include virtualisation and it’s a shame that AMD’s and Intel’s virtualisations are not compatible. Due to market share and also Apple’s use of Intel’s chips, I hope that the difference between the two is trivial; otherwise AMD may find itself at a serious disadvantage in this area. I’m a big fan of VMWare’s current products and look forward to full(er) hardware virtualisation that allows 2 OSs to run independently rather than one being dependent on the other. I’ve been VERY impressed with running Windows under Windows using VMWare, have you tried running Windows as a VM under Linux?

~El~Jefe~ · Post by **~El~Jefe~** » Tue Apr 11, 2006 10:05 pm

will 1% of 1% of the population really care about VMware?
besides hackers looking for something cool and new to infect, I can't thinkof a use for it.

most people are slap happy fine with integrated graphics, 60 gb hd, and a celeron D chip

nutball · Post by **nutball** » Tue Apr 11, 2006 10:36 pm

smilingcrow wrote:I really appreciate AMD but ironically I am feeling frustrated by the way that they keep on improving the power efficiency of their parts OVERALL, but not in such a way that it is easy for the end user to get their hands on them. i.e. there does seem a particularly large variance in power efficiency of CPUs for a given model number and no way to differentiate between them.

Well said! To be honest though I kinda can see their point-of-view -- there's enough complexity in the product lines of both companies without introducing a third dimension (true power consumption). I doubt most retailers would appreciate a factor 3 or 4 increase in the number of SKUs.

Maybe there's enough of a niche market (from overclockers and the silence obssessed) for a third-party reseller to make some cash testing each processor individually and listing the exact specs -- but then they'd formally be selling "used" kit, and it's a hell of a lot of work. I guess the question is how much extra cash would you pay to know precisely what you're getting?

jaganath · Post by **jaganath** » Wed Apr 12, 2006 12:29 am

Maybe there's enough of a niche market (from overclockers and the silence obssessed) for a third-party reseller to make some cash testing each processor individually and listing the exact specs -- but then they'd formally be selling "used" kit, and it's a hell of a lot of work. I guess the question is how much extra cash would you pay to know precisely what you're getting?

This is sort of what happens on Ebay already; most/many chips which overclockers like (ie Opterons) have something in the listing saying "will do 2.7Ghz on air" or whatever, and some screenshots of benchmarks; however I don't expect the same to happen for TDP's, we will still have to rely on SPCR for that.

The TDP's for the X2's do seem to vary a lot, but it seems clear that there is a pattern where older steppings of the Athlon 64 (CG, D0, E3) produce more heat than the newest ones (E6), and you would expect this as AMD learn to optimise the chip better over the production lifetime.

smilingcrow · Post by **smilingcrow** » Wed Apr 12, 2006 1:36 am

jaganath wrote:The TDP's for the X2's do seem to vary a lot, but it seems clear that there is a pattern where older steppings of the Athlon 64 (CG, D0, E3) produce more heat than the newest ones (E6), and you would expect this as AMD learn to optimise the chip better over the production lifetime.

I was intending to refer to the variance in TDPs for a particular chip revision and the newer revisions at that. The program AMD64 TCaseMax shows the individual TDP and TCaseMax for the newer revisions and reportedly they can vary quite a lot between chips with the same part number.
It’s good to see AMD will be releasing a part marked as being 35W for those who desire them. Hopefully, the price premium will not be remotely as large as it is for the low power Opteron HE series.

Hifriday · Post by **Hifriday** » Wed Apr 12, 2006 7:52 am

I do believe AMD is making low TDP dual-core chips more and more widely available... in the form of the s939 Opterons.

So far all posters (except one) on the TCaseMax forum post 35.0W, including Halycon's Opteron 180 on the X2 TDP thread. Although certainly this is not any kind of guarantee, I suspect that lower TDP chips with lower heat/power consumption make for more reliable and longer lasting chips, which is a requirement for the Opteron.
In the past the s939 Opterons were hard to find and even out-of-stock at many online retailers, but today it's quite a different story. Even in my local market where s939 Opterons were unheard of just a few months back, now has them stocked at plently of local retailers. The Opteron 165 is selling for only $30 more than the X2 3800+ which has a 200mhz higher clock but half the L2 cache. With so many shops displaying the Opteron 165 in their window, I had to struggle to keep myself from buying this chip (unfortunately my PC use doesn't justify dual-core).

Mats · Post by **Mats** » Tue Apr 18, 2006 3:02 pm

The good thing about this new CPU is that we can use a more widely available socket. We don't need T64 anymore for desktop, and we don't have to look so hard for a working mobo. The situation is even better for Intel users, if they can wait for Conroe. They don't have to buy strange, overpriced mobos that no one would buy for, say, a Pentium D. Buyers have been forced to sacrifice a lot to get a low power CPU for the desktop, and I really hope those days are over.

MikeC said that the curent X2 is running cool enough, and I think he's right. And no, AM2 won't bring you much more performance today, but now you have a lot of CPU's to choose from. Sempron, this 35 W with either single or dual core instead of a T64, A64 X2 and FX. Don't know if we'll see Opteron 100 (or was it called 2100, don't remember) for this socket.

This new CPU will be especially good for newbies who're looking at the TDP only when buying. But then again we have those few people who really must have the lowest possible TDP, they have to look for socket S1. And so the quest starts all over again.

AMD have moved the release date for AM2 to may 23, so it's only one month left. BTW, I've already posted this in another thread. I think the new T64 X2 will show up on may 11.