Socket AM3+ is DEAD

Sheppard

[H]ard|Gawd
Joined
Jan 3, 2011
Messages
1,199
http://www.digitimes.com/news/a20130826PD216.html

"AMD has recently updated its product roadmap and is set to release its Hawaii-based GPUs at the end of September, Kaveri-based APUs for the high-end segment and Kabini-based APUs for the entry-level segment in the first quarter of 2014, according to sources from the upstream supply chain.

AMD has declined to comment on unannounced products.

The Kabini-based APUs will adopt the Socket ST3 for notebook applications and the Socket FS1B for desktop applications. The APUs will enter mass production in February 2014 and be announced in March 2014.

The Kabini-based APUs will have power consumption of 25W and AMD will release two quad-core models, A4-5350 and A4-5150, as well as a dual-core model E1-2650.

The Kabini-based APUs for desktop applications were originally scheduled to release in the second half of 2013, but have now been rescheduled for March 2014. Kabini's successor, Beema, will also be rescheduled for launch in the second half of 2014 or the first half of 2015, adopting the FS1B socket and Heterogeneous Systems Architecture (HSA).

AMD's FM1 and AM3 sockets will start phasing out in mid-2013 and the end of 2013, respectively. By the end of 2013, Socket AM3+ processors will account for 30% of AMD's total processor shipments, while Socket FM2-based processors will account for the remaining 70%.

In 2015, for the desktop market, AMD will release Carrizo-based APUs, featuring Excavator architecture with two power consumption specifications: 45W and 65W. The company will also release Nolan to replace Beema."

EDIT/UPDATE"
Did some digging, according to this Techspot Article, me intial Impression may have been inaccurate
http://www.techspot.com/news/53799-a...-for-q114.html
"By the end of 2013, AMD will have begun phasing out their FM1 and AM3 sockets, with AM3+, FM2 and new sockets being utilized going forward. In 2015, 'Carrizo' APUs will be launched to succeed Kaveri in the desktop market, featuring the Excavator architecture at two TDPs: 45W and 65W. 'Nolan' chips will also be unveiled to replace Beema, if the reports are correct."


or not....

http://www.maximumpc.com/secret_amd_roadmap_suggests_end_line_fx_processor_lineup2013
We're a little hesitant to fan the flames on this one, but according to a series of documents AMD supposedly handed out to OEMs "behind closed doors," the end of the road for the Sunnyvale chip designer's FX Series might be fast approaching, if not already here. Should AMD ultimately decide to retire its FX Series sooner than later, the FX 9590 could end up being the brand's swan song.

Don't shoot the messenger -- this information comes courtesy of VR-Zone, which claims to have obtained and combed through the secret documents. You can't see them, and neither can we because VR-Zone is opting not to post the images publicly due to heavy watermarking.

It would be a curious move by AMD to retire the FX Series after having broached 5GHz, earning the company a bit of bragging rights. The 5GHz feat did not, however, earn AMD the performance crown, which is held firmly by Intel and its Haswell architecture at this point.

Another alternative is that AMD could skip an FX refresh for all or most of 2014 and launch new chips either at the end of the year or sometime in 2015.
 
Last edited:
I'll still support them no matter what they do with their current socket they seem to have more reliable motherboards then Intel anyday.
 
I don't know about being dead, but I am definitely interested in an FM2+ chip, but nobody knows what it will be capable of video wise. GPU wise I dont need much, but my 540m is hurting me BAD. Really low FPS and shuddering can cause me to pass out so I have had to give up most gaming until I build my new rig.
 
This is weird if anything this tells you nothing more then we already know, the speculation is on if AM3+ will see Steamroller or not.

AM3+ already was dead long before you posted this if and when (or not) the Steamroller would come out would be the last AM3+ cpu for that platform.
 
Last edited:
AMD's FM1 and AM3 sockets will start phasing out in mid-2013 and the end of 2013, respectively.

Does this mean that there will be no more new AM3+ chips at all and that after 2013 production of all AM3+ processors will end?
 
If AM3+ is dead & FM* doesn't support ECC, I probably won't be buying AMD anymore. I don't really want G34 or C32 for ECC support.

The AM3+ Opertons are basically non-existent for the most part. Anyone ever seen one for sale at any major retailer (eg: not ebay).

That $199 1230v2 (microcenter) & a $150-$250 Intel workstation board is looking better by the day.
 
http://www.digitimes.com/news/a20130826PD216.html

"AMD has recently updated its product roadmap and is set to release its Hawaii-based GPUs at the end of September, Kaveri-based APUs for the high-end segment and Kabini-based APUs for the entry-level segment in the first quarter of 2014, according to sources from the upstream supply chain.

AMD has declined to comment on unannounced products.

The Kabini-based APUs will adopt the Socket ST3 for notebook applications and the Socket FS1B for desktop applications. The APUs will enter mass production in February 2014 and be announced in March 2014.

The Kabini-based APUs will have power consumption of 25W and AMD will release two quad-core models, A4-5350 and A4-5150, as well as a dual-core model E1-2650.

The Kabini-based APUs for desktop applications were originally scheduled to release in the second half of 2013, but have now been rescheduled for March 2014. Kabini's successor, Beema, will also be rescheduled for launch in the second half of 2014 or the first half of 2015, adopting the FS1B socket and Heterogeneous Systems Architecture (HSA).

AMD's FM1 and AM3 sockets will start phasing out in mid-2013 and the end of 2013, respectively. By the end of 2013, Socket AM3+ processors will account for 30% of AMD's total processor shipments, while Socket FM2-based processors will account for the remaining 70%.

In 2015, for the desktop market, AMD will release Carrizo-based APUs, featuring Excavator architecture with two power consumption specifications: 45W and 65W. The company will also release Nolan to replace Beema."

Show me where socket AM3+ is dead? It is not there at all and AM3+ accounts for 30%of their total shipments. I read AM3 is dead but that is NOT AM3+.

Once again: To make things clear, AM3 is NOT AM3+ at all. However, I do believe that steamroller would be the last CPU on the AM3+ anyways. Not really a big deal since it had a great run and Intel typically supports only one processor type on one chipset each year. (No thanks, I prefer not to replace my board each time I would like a new processor.) Having ECC support, 6 native Sata III ports, USB3 natively and a drop in cpu is great.
 
Well if that really is AM3 and not AM3+ this is nothing at all to worry about.
 
If AM3+ is dead & FM* doesn't support ECC, I probably won't be buying AMD anymore. I don't really want G34 or C32 for ECC support.

The AM3+ Opertons are basically non-existent for the most part. Anyone ever seen one for sale at any major retailer (eg: not ebay).

That $199 1230v2 (microcenter) & a $150-$250 Intel workstation board is looking better by the day.

If true this would be a problem, I don't have the money for ECC in all my builds but need it for some of them. I already encountered this when building my latest NAS: no mATX boards for 970/990 and no ECC for FM2 meant an AM3 & 880G was as modern I could get in the form factor I wanted.

This might not be the case in the future though. Since AMD puts the memory controller on the CPU couldn't FM2 support ECC if they just put it on the chip?
 
Show me where socket AM3+ is dead? It is not there at all and AM3+ accounts for 30%of their total shipments. I read AM3 is dead but that is NOT AM3+.

Once again: To make things clear, AM3 is NOT AM3+ at all. However, I do believe that steamroller would be the last CPU on the AM3+ anyways. Not really a big deal since it had a great run and Intel typically supports only one processor type on one chipset each year. (No thanks, I prefer not to replace my board each time I would like a new processor.) Having ECC support, 6 native Sata III ports, USB3 natively and a drop in cpu is great.

Come on really? Socket "Am3 non+" has already been phased out....
 
Just a note - AMD still has AM3 non AM3+ out there in the form of a few low end Sempron and Athlon skus.
 
Come on really? Socket "Am3 non+" has already been phased out....

Read the article again then please. It specifically states AM3, not AM3+ is being phased out at the end of 2013. Anything else to add?

Edit: What a minute, you are the original poster and yet you did not seemingly read the article?!?!?!?!? I tend to look at what is actually written, not what someone wants it to say.
 
I read the entire article...
AMD-Roadmap-2014-2015.png
 
I'll still support them no matter what they do with their current socket they seem to have more reliable motherboards then Intel anyday.

I can confirm that my Richland 6800K and newer socket FM2 motherboard have been rock solid. Pleasant surprise!
 
I can confirm that my Richland 6800K and newer socket FM2 motherboard have been rock solid. Pleasant surprise!

Ditto. I can also confirm that my Asrock 990FX Extreme 4 and 990FX Extreme 9 boards have been rock solid as well. :D I know I prefer at least an 8 core processor for what I do on a daily basis.

If they switch the CPU's all to FM2+, I would need an 8 core on that to make it worth my while. (Already have my FX 8350 and FX 8320 so I do not want to go back to 4 cores.) My Extreme 4 is 2 years old and going strong, my Extreme 9 is one month old but have had no issues with it so far.
 
Look I don't want to be right, I would rather am3+ keep maturing as I feel there is more that can be done. Do some digging and see if you can find anything.
 
Did some digging, according to this Techspot Article, me intial Impression may have been inaccurate
http://www.techspot.com/news/53799-...admap-for-2014-15-kaveri-slated-for-q114.html
"By the end of 2013, AMD will have begun phasing out their FM1 and AM3 sockets, with AM3+, FM2 and new sockets being utilized going forward. In 2015, 'Carrizo' APUs will be launched to succeed Kaveri in the desktop market, featuring the Excavator architecture at two TDPs: 45W and 65W. 'Nolan' chips will also be unveiled to replace Beema, if the reports are correct."

Or not...

http://www.maximumpc.com/secret_amd_roadmap_suggests_end_line_fx_processor_lineup2013
We're a little hesitant to fan the flames on this one, but according to a series of documents AMD supposedly handed out to OEMs "behind closed doors," the end of the road for the Sunnyvale chip designer's FX Series might be fast approaching, if not already here. Should AMD ultimately decide to retire its FX Series sooner than later, the FX 9590 could end up being the brand's swan song.

Don't shoot the messenger -- this information comes courtesy of VR-Zone, which claims to have obtained and combed through the secret documents. You can't see them, and neither can we because VR-Zone is opting not to post the images publicly due to heavy watermarking.

It would be a curious move by AMD to retire the FX Series after having broached 5GHz, earning the company a bit of bragging rights. The 5GHz feat did not, however, earn AMD the performance crown, which is held firmly by Intel and its Haswell architecture at this point.

Another alternative is that AMD could skip an FX refresh for all or most of 2014 and launch new chips either at the end of the year or sometime in 2015.
 
the last I heard is after FM2/AM3+ AMD wanted to go with a unified socket so there is less RnD and much better choice for the consumer.

Also AMD does not need to do platform overhauls all the time like Intel needs to do, they have much different approach on their architecture(not talking performance just the design side)

AMD to ramp performance just needs up overall IPC, reduce latency and up the HT to the fastest level all of which can be done in the cpu itself, why they do not is well beyond me as they are only using at beast 1/2 of the available speed HT that they can be using, if they can give it the most bandwidth possible THEN tinker with getting the other things up to par it might matter more when cpu is clocked stupid high?

When things start shifting multi-threaded and relying more on cpu/gpu offloading I think AMD current design which is lackluster is going to see a major performance gain, and Intel will need to catch up, hopefully AMD did the right thing in this approach as historically they are not the first to the win, but are the ones that put up the best fight and force the industry to new standards.
 
Well at the rate I upgrade now, I won't need to be vaguely interested for at least another 2-3 years.
 
Also AMD does not need to do platform overhauls all the time like Intel needs to do, they have much different approach on their architecture(not talking performance just the design side)
I actually think this is probably holding back AMD to some extent. Having DDR2 & DDR3 memory controllers on die must cost die size and maintaining backwards compatibility to the level they do probably has some design/layout implications. Now, I think Intel could be making more of their CPUs backwards compatible, but it does mean they've been able to make clean breaks and move forward with optimal designs rather than compromises.

AMD to ramp performance just needs up overall IPC, reduce latency and up the HT to the fastest level all of which can be done in the cpu itself, why they do not is well beyond me as they are only using at beast 1/2 of the available speed HT that they can be using, if they can give it the most bandwidth possible THEN tinker with getting the other things up to par it might matter more when cpu is clocked stupid high?
Yes... they "just" need to increase IPC. You make it sound so easy. IPC isn't just something that can be added like some kind of ingredient to the processor... it has to be designed in and takes time, money, and skill. As far as using 1/2 the HT speed... whos to say it needs more? Maybe theres a power cost associated with going higher and the performance gain is negligible?

When things start shifting multi-threaded and relying more on cpu/gpu offloading I think AMD current design which is lackluster is going to see a major performance gain, and Intel will need to catch up, hopefully AMD did the right thing in this approach as historically they are not the first to the win, but are the ones that put up the best fight and force the industry to new standards.
"when" things start shifting to multi-threaded, AMD will still be seriously behind because their individual cores are so far behind. Intel could drop 6-8 core CPU's in the consumer segment if they wanted to, but they don't need to. If they ever needed to, AMD would be even further behind and putting 16 cores in a CPU and waving their arms saying "just wait, we'll need 16 cores soon and Intel only has 8!" I'm also not as optimistic about their GPU chances. Heterogeneous computing might help AMD right now, but Intel is already at near parity with their GPUs
 
I actually think this is probably holding back AMD to some extent. Having DDR2 & DDR3 memory controllers on die must cost die size and maintaining backwards compatibility to the level they do probably has some design/layout implications. Now, I think Intel could be making more of their CPUs backwards compatible, but it does mean they've been able to make clean breaks and move forward with optimal designs rather than compromises.


Yes... they "just" need to increase IPC. You make it sound so easy. IPC isn't just something that can be added like some kind of ingredient to the processor... it has to be designed in and takes time, money, and skill. As far as using 1/2 the HT speed... whos to say it needs more? Maybe theres a power cost associated with going higher and the performance gain is negligible?


"when" things start shifting to multi-threaded, AMD will still be seriously behind because their individual cores are so far behind. Intel could drop 6-8 core CPU's in the consumer segment if they wanted to, but they don't need to. If they ever needed to, AMD would be even further behind and putting 16 cores in a CPU and waving their arms saying "just wait, we'll need 16 cores soon and Intel only has 8!" I'm also not as optimistic about their GPU chances. Heterogeneous computing might help AMD right now, but Intel is already at near parity with their GPUs

Actually, Intel would not have it so easy to put a 6 or 8 core processor in the consumer space. As it is right now, their 6 core processors have a 140 Watt TDP and do not even come with a heatsink/fan, I would imagine an 8 core would have a much higher TDP.
 
"It would be a curious move by AMD to retire the FX Series after having broached 5GHz, earning the company a bit of bragging rights."

No it is not curious at all.

As I said previously, they are simply milking their 32nm fab capacity for everything its worth, and intending to ride am3+ into the grave with piledriver in the saddle.
 
"It would be a curious move by AMD to retire the FX Series after having broached 5GHz, earning the company a bit of bragging rights."

No it is not curious at all.

As I said previously, they are simply milking their 32nm fab capacity for everything its worth, and intending to ride am3+ into the grave with piledriver in the saddle.

So says your speculation and opinion. I do not see that as being the case since it would make more sense to produce the 8 core steamroller on an already established and solid platform. However, it would still be the last one to do so.

They are not going to kill off 30% of their sales especially when there is nothing to take its place yet. A two module, four core apu does not even come close to the 8 core in terms of processing power and what 8 threads can accomplish.

In the future, when they do unify the sockets and produce an 8 core on it, great, I will buy it. However, I am not going backwards on my main machines, no thanks.
 
Actually, Intel would not have it so easy to put a 6 or 8 core processor in the consumer space. As it is right now, their 6 core processors have a 140 Watt TDP and do not even come with a heatsink/fan, I would imagine an 8 core would have a much higher TDP.
Actually incorrect, its 130W TDP, and those are SB-E parts with quad-channel memory controllers based cores from a Tock-cycle ago. I think we can agree that quad-channel memory is pretty unnecessary, so remove it from the equation. If we want to assume a very quick simply scaling to a 6 core from a 77W IVB quad (which includes GPU TDP), you get <120W TDP for a hex-core IVB.Your statement that this might not be so easy for Intel completely belies the fact that AMD's 8-core parts have 130W TDP's. Intel put out their Gulftown hex-cores at 130W TDP and power consumption has only gone down since then.
 
Hmmmmm, I know I have seen an Intel 140 Watt TDP 6 core before. That said, I don't believe it will be as simply as you are making it sound. Doable, but not simple. There is a thread in the Intel section that indicates a 130 to 140 Watt TDP with a 6 to 8 core processor Haswell-E.

Therefore, it does not appear they would ever just slap on 2 or 4 more cores on the 4 core processor currently being used. Increasing cores without significantly increasing the TDP will definitely be a challenge. However, it does not really matter to me since I am going to go with the 8 core steamroller if/when it comes out. :D (The tech is interesting and fun, just that my wallet is not opening up for Intel, personal preference and all that.)
 
Hmmmmm, I know I have seen an Intel 140 Watt TDP 6 core before. That said, I don't believe it will be as simply as you are making it sound. Doable, but not simple. There is a thread in the Intel section that indicates a 130 to 140 Watt TDP with a 6 to 8 core processor Haswell-E.

Therefore, it does not appear they would ever just slap on 2 or 4 more cores on the 4 core processor currently being used. Increasing cores without significantly increasing the TDP will definitely be a challenge. However, it does not really matter to me since I am going to go with the 8 core steamroller if/when it comes out. :D (The tech is interesting and fun, just that my wallet is not opening up for Intel, personal preference and all that.)
I believe the 140W TDP is for 8 core parts. Anyway, I was merely attempting to rebut the idea that it would be difficult for intel to do. And why would it be? They already have the tech and many of the designs. The cores are already designed, they would need to re-do layouts and such, maybe chop out the GPU and retain the dual-channel memory controller or modify it for use with additional cores. The point is, they have everything in their repertoire already because they already make 6-8 core chips... All it would take is a little time and effort. Simply put, they just have no need to go in that direction for any real reason. AMD isn't so lucky. Their single core performance is so poor that they need to compensate with more threads solely for marketing reasons. I have no doubt that Intel could bring in their IB 22nm tech and put out a 6-core CPU with a lower TDP than AMD's parts have by a real margin.
 
What kind of ridiculous topic title is this? You posted something completely inaccurate and then added a bunch of conjecture.

Is there any way to downvote threads on here? I feel like you completely wasted my time with this.
 
What kind of ridiculous topic title is this? You posted something completely inaccurate and then added a bunch of conjecture.

Is there any way to downvote threads on here? I feel like you completely wasted my time with this.
Dude, its a tech forum.... 99% of the threads here are either "help me" or speculation
 
you wasted your own time by reading it and commenting :p

Anyways AMD has more multi-thread approach then Intel does at this point, most "legacy" and modern stuff is built for Intel which generally speaking is Single threaded by nature part of the reason their HT works quite well and why they kept it instead of shifting away from it after the P4 with HT.

AMD design however is a multi-threaded beast that needs all the "cores" used and threaded properly to show its potential as of yet very little is built with this in mind, so I think going forward if they reduce latency, up the IPC, increase HT speed(might mean they might never need to shift from dual channel) they have a great design on hand.

I guess will keep it simpler as this is the way I am understanding this, if built single threaded but can scale to many threads Intel wins as it can easily deal with the single threads due to their IPC efficiency and due to the way their architecture is built to HT they can easily scale the performance forwards, however, if built multi-threaded in mind then current FX chips do VERY well as they no longer try to shift the loading they just do the work, so Intel has an edge currently, but if the devs learn how to appropriately thread AMD architecture I think Intel may need to redesign their approach as many threads over many cores will be far more efficient then taking a thread and trying to split it.

I don't know but it does seem AMD might have the advantage "if" they can tweak FX weak sides, bolster FX good sides, and get the devs/software to truly take advantage of the design, FX is more of a hardware the requires software to make it work well, core I seems to be hardware that does not need near as much software to let it gain its advantage, almost like the older radeons vs their Nvidia counterparts, if one was threaded properly it had ALOT of oomph available BUT not many devs took the time to bother so their performance was just ok instead of great.
 
Read the article again then please. It specifically states AM3, not AM3+ is being phased out at the end of 2013. Anything else to add?

Edit: What a minute, you are the original poster and yet you did not seemingly read the article?!?!?!?!? I tend to look at what is actually written, not what someone wants it to say.

You are correct AM3 is dead , not AM3+. There will be additional offerings for AM3+ though what they will be is not clear. Digi Times was misleading by saying AMD has released a new road map. That is simply not true. The new road map will be released between the end of October and the early November AMD APU conference. I know this from John Taylor who is the Global Marketing manager for AMD.
 
on another note, AMD currently uses x16 width on the HT they can use x32 which will benefit the whole of the system not directly the cpu as it probably already has ample speed in this regard, they currently max at ~25gb speed when they could go up to ~52gb, the controller is already built to use it for desktop and server side, but I suppose they do not see a point of using it right now as it would make it more expensive on the cpu and board side I assume, but, not on an electrical increase side.

As it stands, I have seen a few times QPI which is similar is faster(as it gets clocked according to Bclck far as I read) however, Hypertransport in AMD chips is far more flexible in its use and can be tweaked in many number of ways that QPI simply was not built to be able to do..
 
What kind of ridiculous topic title is this? You posted something completely inaccurate and then added a bunch of conjecture.

Is there any way to downvote threads on here? I feel like you completely wasted my time with this.

lol, and then you take the time to post about it :rolleyes::p
 
Well you are wrong. Am3 is dead , NOT AM3+. You rumor mongerers never check out your sources.
 
Well you are wrong. Am3 is dead , NOT AM3+. You rumor mongerers never check out your sources.

The problem is AM3 seems to have been dead for some time. I mean didn't the production of all 45nm processors end sometime in 2012? So its surprising that it is announced that AM3 is dead in a few months.
 
Last edited:
Back
Top