Sign in with
Sign up | Sign in
Your question

AMD announces 9W Sempron

Last response: in CPUs
Share
May 31, 2007 3:45:04 PM

Today AMD furthered their commitment to low power chips with the announcement of a Socket S1 Sempron for embedded and single board cmputers that reportedly disspates only 9W.
It should do well. It's clocked at 1 GHz.

Linkage!

More about : amd announces sempron

May 31, 2007 3:57:00 PM

S1 is for mobile right?
May 31, 2007 4:36:02 PM

That rocks. I'm sure it is not the fastest, but for ultra-mobile non CPU intensive applications this will rock.
Related resources
May 31, 2007 5:22:48 PM

Quote:
S1 is for mobile right?


Yea I think it is the new mobile socket that replaced the socket 754.

If I'm not mistaken
the S1=current turion platform
May 31, 2007 5:26:03 PM

yeah but only 9w, think about battery life! it may only be 1ghz but it will be able to run all day with out being plugged in. 8)
May 31, 2007 5:58:25 PM

gotta say it: but can I overclock it?
May 31, 2007 6:07:52 PM

Remember Johnny Mnemonic?

I think this and a processor would be really weird

Memory in the brain
May 31, 2007 6:22:42 PM

Couple this with a flashdrive and reduced power consumption OLED screens, that would be a huge increase of battery life not to mention the passive cooling possibility. It should be enough performance to get things done. When they do the CPGPU merge is when you should really get the total consumption down though but there is quite a bit of time until those get released.
May 31, 2007 6:37:22 PM

While it is good to see AMD at least trying to compete in low power, they have a long way to go. The Pentium M is still a better product at 90nm. What is AMD going to do to compete with Intel's 45nm product that is coming out next year?
May 31, 2007 6:50:53 PM

Quote:
It should do well. It's clocked at 1 GHz.


While I wish them all the best, I'm not sure that it will set the world alight. After all, VIA have been selling things like the C3 for quite some time now without attracting a great amount of interest: Via C3. OK, this is a 1.4 GHz, 20 Watt part but they have slower/cooler parts available.

My guess is that it will get some wins in embedded (against ARMs, MIPS and proprietary designs from Freescale & Hitachi, etc) and they may well even cannibalise the top end of the Geode's market, too. So I'm guessing some networking kit, some tablets. But, I can't see them being everywhere in the way that, say, ARMs are.

But then, I don't know the price.
May 31, 2007 7:08:28 PM

Quote:
gotta say it: but can I overclock it?

good one :lol: 
a b à CPUs
a b À AMD
May 31, 2007 7:11:57 PM

Wasn't VIA's "Eden" CPU a 9W "monster"? I think it also ran at 1GHz. I wonder if we can get a low power review of these two chips, not that it would mean much...
May 31, 2007 7:16:05 PM

Quote:
yeah but only 9w, think about battery life! it may only be 1ghz but it will be able to run all day with out being plugged in. 8)


Thought CPU's eat a lot of power, there are many, many other things that make up power drain. The biggest two are hard drives and LCD screens. When those become ultra low voltage/low wattage then you'll be able to run ur laptop all day. Until then, you're still stuck w/ 5hrs MAX.
May 31, 2007 7:26:19 PM

Quote:
It should do well. It's clocked at 1 GHz.


While I wish them all the best, I'm not sure that it will set the world alight. After all, VIA have been selling things like the C3 for quite some time now without attracting a great amount of interest: Via C3. OK, this is a 1.4 GHz, 20 Watt part but they have slower/cooler parts available.

My guess is that it will get some wins in embedded (against ARMs, MIPS and proprietary designs from Freescale & Hitachi, etc) and they may well even cannibalise the top end of the Geode's market, too. So I'm guessing some networking kit, some tablets. But, I can't see them being everywhere in the way that, say, ARMs are.

But then, I don't know the price.


Well, when you consider that a 1.9GHz dual core is under $75 you gotta figure these will be no more than $40. With their own embedded chipset, I can see thispicking up a UMPC win or two.

Also, these would work well in Smart Phones. They could also use them for small home appliances.

Combining them with the new ATi mobile graphics chip, this will make sme noise in several markets. I can also see handheld game boxes using someting like this.

9W is incredible.
May 31, 2007 7:50:24 PM

Why is this important?

Intel has a 5.5W Core Solo part that will wipe the floor with this Sempr0n.

Link!
May 31, 2007 8:07:18 PM

Quote:
Why is this important?

Intel has a 5.5W Core Solo part that will wipe the floor with this Sempr0n.

Link!


Because this is a CPU forum. And also everyone knows that at the lower clocks it's not the same. I don't think I've seen a single core review in ages, but either way AMD is usally much better on idle power than Intel, which is where a chip like this would spend a majority of it's time.
May 31, 2007 8:12:02 PM

Quote:
Wasn't VIA's "Eden" CPU a 9W "monster"? I think it also ran at 1GHz. I wonder if we can get a low power review of these two chips, not that it would mean much...



VIA is not the CPU design company that AMD is and Sempron was ruling the retail market for awhile. It woul dbe interesting to test all of the ULV mobile chips. Though those tests would more so concern power and not perf as something like this with a flash drive would be sweet.
May 31, 2007 8:24:53 PM

This is a rgeat product. I use a Pentium M 1.0 ghz tablet PC with Win XP Tablet edition, 512mb RAM and a GF4 MX graphics card, and the battery life is decent but not great, especially when using Windows Journal a lot.

A CPU like this could make a big difference.
May 31, 2007 8:48:48 PM

Sorry but.. Intel has an 8W 1.2GHz Core (1) Duo ULV... (we are using it)
It's a dual core, and it's smoking fast.
The best you can get, for embedded applications.
May 31, 2007 9:03:15 PM

Quote:
Sorry but.. Intel has an 8W 1.2GHz Core (1) ULV... (we are using it)
It's a dual core, and it's smoking fast.
The best you can get, for embedded applications.


Well, what it sounds like you're saying is that Intel should just forget about CSI since AMD already has it and it will more than likely be faster in its HT3 incarnation.
May 31, 2007 9:06:10 PM

Quote:
Sorry but.. Intel has an 8W 1.2GHz Core (1) ULV... (we are using it)
It's a dual core, and it's smoking fast.
The best you can get, for embedded applications.

I wonder if this sempron is going to be recognized by all S1 laptops.
May 31, 2007 9:53:05 PM

Well, that looks interesting. I wonder if I could run an ultra low-power desktop PC by stacking together like 4 of these babies. ;) 
June 1, 2007 12:49:10 AM

the 5.5 watt core solo sounds sweet, the sempron has a f**king pr rating, i want amd to stop that crap now, thats an archaic cyrix con, and we all no what happened to them. Just give the thing a model number, jeez
June 1, 2007 9:32:22 PM

Current dual core models aren’t ideal for very low power solutions as they consume too much power at idle. This will change when each core has its own clock and voltage. Here’s a comparison of three Intel mobile CPUs:
http://www.silentpcreview.com/forums/viewtopic.php?t=35...

I didn’t notice anyone mentioning the U7600 which is an ULV Core 2 Duo at 1.2GHz with a TDP of 10W.
June 1, 2007 9:39:41 PM

Quote:
I am just wondering if maybe this is the chip that will be used in the laptop for the One Laptop Per Child organization?
Nah...Intels going to ruin that, remember? :wink: :roll:
June 2, 2007 1:49:28 AM

Hi Jack, I was hoping to make two points:

1. Show the highest performing Intel part with a 10W TDP – the U7600.
2. Show that current dual core CPUs consume significantly more power at idle and when only using 1 core than single core CPUs. This is a big issue for this market segment.

I measured power consumption for a very low power desktop system using a Core Solo and Core Duo. A Core Solo T1400 and Core Duo T2600 were used and I threw in a Celeron M 420 for good measure. I restricted the load CPU speed to 1.66GHz for the Core chips and 1.6GHz for the Celeron as that was the closest speed match due to the differing FSB speeds. The two Core chips were under-volted which brought their power consumption closer to that of an LV or ULV part. The wattage figures below are for the whole system believe it or not.

Celeron M 420
Idle - 1.6GHz, 1.216V – 24W
Load - 1.6GHz, 1.216V – 28.5W

Core Solo T1400
Idle – 1GHz, 0.95V – 20W
Load – 1.66GHz, 1.1V – 26W

Core Duo T2600
Idle – 1GHz, 0.95V – 22W
Load (1 core) – 1.66GHz, 1.05V – 32W
Load (2 cores) – 1.66GHz, 1.05V – 36.5W

The first thing that stands out is that the Core Solo system consumes 2W less than the Core Duo at idle. It doesn’t sound much but it equates to a 10% higher power usage for the Core Duo system.
The second difference is than when fully loading one core using Prime95 the Core Duo system consumed an extra 6W even though its voltage was .05V less than Core Solo. This equates to an extra 23% power consumption for the whole system.
Even the Celeron M running at 0.166V higher VCore consumed 3.5W less power than the Core Duo.

This is why AMD and no doubt Intel are moving towards separate clocks and voltages for each core in mobile CPUs; the current method is way to inefficient.
It’s also shows why it’s not always meaningful to compare apples and oranges and say that a 9W Sempron is necessarily a weaker chip than the 10W core 2 Duo U7600 even if they were priced the same. If you don’t need the 2nd core why pay the power price of having it there drawing juice.
June 2, 2007 11:35:29 AM

Quote:
Compliments for a fine compilation of data -- however, to the subject point at hand -- do you think a 1 GHz sempron will take a 1.2 Ghz Yonah? In either performance or performance/watt?

No, the U1500 (1.33GHz Yonah 5.5W) will easily beat it.
The Sempron may beat or match the ULV Core 2 Duos for power consumption in a system that is typically running at idle or at low load though. The same goes when comparing the Sempron against a mobile Turion X2 I imagine as I think both AMD and Intel are handicapped in the same way in this area for now. I wonder who will rectify this shortcoming first.

I think people are often too easily impressed by the 9W figure and go Wow that’s impressive. I’m not impressed by a 9W TDP for a single core K8 at 1GHz. AMD have had a dual core X2 on 90nm at 2GHz with a TDP of 35W out for quite a while now, which offers roughly the same performance per watt:
2 x 2GHz = 4GHz for 35W = 35 / 4 = 8.75W per GHz.
Admittedly, these X2 3800+ EE SFF chips seem to be about as easy to locate as a certain Mr Bin Laden.

Quote:
I am not disagreeing with your analysis, however, what I am pointing out is that you are not measuring with the lowest wattage parts built on the lowest power/lowest leakage process at clocks comprobable to what the Sempron of the subject thread is touting.

Do you think that Intel’s LV and ULV parts are anything more than under-volted and under-clocked Core (2) Duos? I don’t get the impression that they are. I say this after looking at the power consumption data for under-volted Yonahs & Meroms and extrapolating. Not that it matters as they are still very impressive but it will be interesting to see what Intel will achieve now that they are focussing more in this area.

The Intel LV and ULV parts in question tend to be only available to OEMs and are soldered to the motherboard so I can’t test them.

Quote:
I am willing to bet you much lower idle power than what would be meausred out since you are measuring against an older 90 nm process.

I’m not so sure that idle power draw will be much lower when comparing Yonah (90nm) with Merom (65nm). Intel’s 90nm process was fine and at idle there isn’t much power draw to reduce on the mobile parts. I was expecting more from the 65nm process in terms of power reduction but it didn’t seem to deliver. Merom seems to beat Yonah more down to architectural changes than process improvements.
The disparity between the single and dual cores with regard power consumption will still exist at 65nm but should be slightly less.
BTW, is this 9W Sempron a 65 or 90nm part?

Quote:
When you add up the major consumers in the ultra-low power space, even at 10 Watts, this is not the major power draw... it is the display, and drives if present, contribute a significant portion.

Any device with a large screen will consumer a fair amount of power, agreed. The screen less system that I mentioned previously consumed 26W at full load at 1.66GHz and that included a mobile hard drive. In that scenario a 9W 1GHz CPU of modest IPC is underwhelming.
The only way that this 9W Sempron can compete is on price, an all too familiar story for AMD at the moment.
June 2, 2007 2:08:40 PM

Quote:
Without question, Intel's leakage on their ULV parts is much much lower than that of AMD's 9W part.

COOL :lol:  Baron gets owned again! :trophy:
June 2, 2007 2:29:03 PM

Quote:

Do you think that Intel’s LV and ULV parts are anything more than under-volted and under-clocked Core (2) Duos? I don’t get the impression that they are.


No, they are definitely different chips. I linked above references to the Intel P1265 process. Once intel develops the latest technology revision, they also produce a different process using the newest tech. as a base and optimize it for power. This is much different than simple undervolting.
Thanks for clearing that up Jack. I made the mistake of assuming that because the 65nm process doesn’t seem to offer much power saving at load compared to the 90nm process that it hadn’t been optimised to reduce consumption at idle.
SilentPCReview measured Yonah at idle as consuming 1.5W at 1GHz 0.95V; so there isn’t much room for improvement there. Every little helps when it comes to mobile parts though as Intel will mention sub 1W savings in presentations.
http://www.silentpcreview.com/article313-page5.html

Quote:
One of my very favorite refrences as it contains all the most interesting summary data:
http://www.realworldtech.com/page.cfm?ArticleID=RWT1230...
Notice on the right half of the table, a section labeled low power -- Intel is the only MPU company that derives a completely different process for this purpose, other companies do similar work, but for different applications. Finally, notice the NMOS and PMOS Ioff values compared to the high power counterparts.

A very interesting technical link but mainly way over my head. I’d be grateful if you would explain the relevance of the parameters (NMOS & PMOS) that you highlighted and how they relate to the discussion in hand. I see that half of the values for the low power process are missing for some reason!

Quote:
Without question, Intel's leakage on their ULV parts is much much lower than that of AMD's 9W part.

The SPCR article that I linked above shows the following data for power consumption at idle:

Dothan 800MHz 0.734V 1W
Yonah 1GHz 0.95V 1.5W
Turion (Lancaster) 800MHz 0.92V 2.2W

They are all 90nm parts so the Turion doesn’t fare well at all as it’s beaten by Yonah which is clocked faster and is dual core. These aren't even LV or ULV parts either.
June 2, 2007 3:40:36 PM

Wow. Too much data. I am overwhelmed.

Anyway, I'm glad to see amd and intel are focusing some energy on efficiency. When will ati and nvidia jump on the bandwagon?
June 2, 2007 5:04:08 PM

Quote:
I think ati already did, them jumping to 65nm that quickly means they're getting forced to lower power consumption (possibly by amd)


man, i hope so. i just saw the video review on the his hd 2900xt consuming 215 watts at full load. ouch.
June 2, 2007 6:43:12 PM

Quote:
When will ati and nvidia jump on the bandwagon?


I wish. I bought an 8800 recently and needed a new PSU, my poor old Thermaltake 430W just couldn't handle the thing. :cry: 
June 2, 2007 9:42:49 PM

Quote:
Sorry but.. Intel has an 8W 1.2GHz Core (1) ULV... (we are using it)
It's a dual core, and it's smoking fast.
The best you can get, for embedded applications.


Well, what it sounds like you're saying is that Intel should just forget about CSI since AMD already has it and it will more than likely be faster in its HT3 incarnation.
....What?!?!? :?:
He never said anything about cs3 compared to ht. He only mentioned that intel already has a dual core with lower power rating. I'm all for amd in all, but they have to do something if intel can get out a faster dual core for lower power rating. Of course my hunch says from what I've seen that chip's actual consumption is quite a bit more than 8 watt (the silenX of cpus lol), but there's also the chance that chip will be k8 based considering they're launching a 2008 k8 laptop cpu :?

That's how it sounded. Griffin is not K8. Intel is using mainly single core also as the speed isn't the greatest factor. Most of these will do simple calculations and communication control.
June 2, 2007 9:45:21 PM

Quote:
Without question, Intel's leakage on their ULV parts is much much lower than that of AMD's 9W part.

COOL :lol:  Baron gets owned again! :trophy:


WTF is wrong with you? I made no mention of whose chip would be lower power.

Get a life. And you can't have mine.

BTW, I noticed your girlfriend is making a comeback as your avatar. Way to go.
June 2, 2007 10:31:20 PM

Where are the Barcelona First ?
June 3, 2007 12:47:45 AM

Quote:
I am just wondering if maybe this is the chip that will be used in the laptop for the One Laptop Per Child organization?


No, I doubt it. That will probably stay Geode.
June 4, 2007 5:15:11 PM

*yawn*

My undervolted/underclocked Pentium M consumes 3 watts. 9 Watts is a waste of battery.
!