AMD CPU speculation... and expert conjecture

[ColinAP]

No. My claim about bulk Kaveri wasn't on any roadmap. At contrary, some people took an public GF roadmap and said me that Kaveri was SOI

Some people also linked to this article to prove me wrong

http://www.pcper.com/reviews/Editorial/GLOBALFOUNDRIES-2013-and-Beyond

I recall that after official presentation of Kaveri some people here was still pretending that Kaveri was SOI because one slide from AMD mentioned "SHP".

Same about my claim on 20nm bulk. No roadmap or anything as that. At contrary, some people here used an article that suggested AMD was returning to SOI for 20nm

http://news.softpedia.com/news/AMD-and-GlobalFoundries-Interested-in-FD-SOI-for-the-20-nm-Process-267519.shtml

Idem about my claim about FINFETs on bulk being the future and SOI being 'dead'. No roadmap or anything like that. At contrary, the replies that I received included a dozen of links to articles that supposedly disproved me. This was one of the articles mentioned

http://www.advancedsubstratenews.com/2013/07/globalfoundries-on-cost-vs-performance-for-fd-soi-bulk-and-finfet/

Thus, contrary to your claim, I didn't believe what companies said, but only what passed my expert-eye filter. I am rather sure you can find some intermediate label between "technology Nostradamus" and someone who is right from pure chance. But the label that you can find for me is irrelevant, the relevant concepts are "bulk" and "FINFET".

You really are full of it. It was known back in March 2012 that AMD was moving to bulk for Kaveri.
http://www.brightsideofnews.com/news/2012/3/6/amd-ditches-soi-kaveri-goes-bulk-at-gf2c-more-details-from-the-new-wsa.aspx

Whether people here believed it or not, whether they misunderstood other details like the ambiguous "SHP" slide is irrelevant.

 

[jdwii]

Athlon II X3 440(which can be overclocked quite well) could be grabbed at 70$ and that leaves 70$ for the GPU which i could almost always find a 5670 DDR3 or 6570 DDR3 during llano's life. Also when both components have separate heat-sinks they also allowed better overclocking which improves performance even more making llano during its time even less important for gamers. Again i own a Llano desktop and laptop(love the laptop) their amazing but for gamers on a budget they could have had more per dollar. However i feel like that will go away soon enough. Really all the A10 7850K needs to do is drop the price by 30-50$ and they have a winner.

Actually i built my friend a PC back in the day with a Athlon II x3 and it unlocked to a Phenom II x4 and i also found a 4830(DDR5) for just 65$ on newegg at the time was a lucky thing and not repeatable by any means just brings back memorys

 

[jdwii]

No. My claim about bulk Kaveri wasn't on any roadmap. At contrary, some people took an public GF roadmap and said me that Kaveri was SOI

Some people also linked to this article to prove me wrong

http://www.pcper.com/reviews/Editorial/GLOBALFOUNDRIES-2013-and-Beyond

I recall that after official presentation of Kaveri some people here was still pretending that Kaveri was SOI because one slide from AMD mentioned "SHP".

Same about my claim on 20nm bulk. No roadmap or anything as that. At contrary, some people here used an article that suggested AMD was returning to SOI for 20nm

http://news.softpedia.com/news/AMD-and-GlobalFoundries-Interested-in-FD-SOI-for-the-20-nm-Process-267519.shtml

Idem about my claim about FINFETs on bulk being the future and SOI being 'dead'. No roadmap or anything like that. At contrary, the replies that I received included a dozen of links to articles that supposedly disproved me. This was one of the articles mentioned

http://www.advancedsubstratenews.com/2013/07/globalfoundries-on-cost-vs-performance-for-fd-soi-bulk-and-finfet/

Thus, contrary to your claim, I didn't believe what companies said, but only what passed my expert-eye filter. I am rather sure you can find some intermediate label between "technology Nostradamus" and someone who is right from pure chance. But the label that you can find for me is irrelevant, the relevant concepts are "bulk" and "FINFET".

You really are full of it. It was known back in March 2012 that AMD was moving to bulk for Kaveri.
http://www.brightsideofnews.com/news/2012/3/6/amd-ditches-soi-kaveri-goes-bulk-at-gf2c-more-details-from-the-new-wsa.aspx

Whether people here believed it or not, whether they misunderstood other details like the ambiguous "SHP" slide is irrelevant.

Actually if you been here you would notice that people here did and possibly still believe that making his point relevant,

 

[ColinAP]

Actually if you been here you would notice that people here did and possibly still believe that making his point relevant,

My point isn't that he didn't "predict" anything that wasn't already known.

I predict that the sun will rise tomorrow. I further claim that the moon will be full at some point in the next 28 days. You can disagree with me if you like, but you'd be wrong every time. It doesn't make my claims worth anything though.

 

[juggernautxtr]

Bulk must not be that bad, Intel is using it on their processors too, only finfet/bulk,
did anyone see the graphene article? IBM is getting that worked out pretty well so far.

 

[juanrga]

No. My claim about bulk Kaveri wasn't on any roadmap. At contrary, some people took an public GF roadmap and said me that Kaveri was SOI

Some people also linked to this article to prove me wrong

http://www.pcper.com/reviews/Editorial/GLOBALFOUNDRIES-2013-and-Beyond

I recall that after official presentation of Kaveri some people here was still pretending that Kaveri was SOI because one slide from AMD mentioned "SHP".

Same about my claim on 20nm bulk. No roadmap or anything as that. At contrary, some people here used an article that suggested AMD was returning to SOI for 20nm

http://news.softpedia.com/news/AMD-and-GlobalFoundries-Interested-in-FD-SOI-for-the-20-nm-Process-267519.shtml

Idem about my claim about FINFETs on bulk being the future and SOI being 'dead'. No roadmap or anything like that. At contrary, the replies that I received included a dozen of links to articles that supposedly disproved me. This was one of the articles mentioned

http://www.advancedsubstratenews.com/2013/07/globalfoundries-on-cost-vs-performance-for-fd-soi-bulk-and-finfet/

Thus, contrary to your claim, I didn't believe what companies said, but only what passed my expert-eye filter. I am rather sure you can find some intermediate label between "technology Nostradamus" and someone who is right from pure chance. But the label that you can find for me is irrelevant, the relevant concepts are "bulk" and "FINFET".

You really are full of it. It was known back in March 2012 that AMD was moving to bulk for Kaveri.
http://www.brightsideofnews.com/news/2012/3/6/amd-ditches-soi-kaveri-goes-bulk-at-gf2c-more-details-from-the-new-wsa.aspx

Whether people here believed it or not, whether they misunderstood other details like the ambiguous "SHP" slide is irrelevant.

You are replying to only one of the three points mentioned in my post.

Your BSN link shows a wrong (outdated) roadmap where Kaveri is the successor to Trinity. I also recall how the people who disagreed with me said that the author of the quotes in your BSN link was no longer working for AMD and thus that the quotes couldn't be taken as valid anymore. Some of that people said here that Kaveri was being delayed to coincide with the start of risk production of the supposed 28nm SOI process at GF. Check the pcper link.

There was similar hot discussions about the Kaveri/Steamroller fabrication process in other forums beyond this. I was in minority here defending the idea about bulk. It everything was so evident and well-known, as you pretend, then only a very small minority of people would be expecting SOI Kaveri and refuting the claim about bulk...

I like how everything is evident, known, and expected, but only a posteriori. It was not so evident, known, or expected during those long pages of discussion in different forums including this one.

Finally, let me emphasize that there is nothing ambiguous on the AMD slide about the fabrication process for Kaveri. The original SHP process based on SOI was canceled and replaced by a new SHP process based in bulk silicon that GF tweaked during many months for Kaveri. SHP means Super High Performance. It is not a synonym for SOI as some pretended then.

 

[juggernautxtr]

No. My claim about bulk Kaveri wasn't on any roadmap. At contrary, some people took an public GF roadmap and said me that Kaveri was SOI

Some people also linked to this article to prove me wrong

http://www.pcper.com/reviews/Editorial/GLOBALFOUNDRIES-2013-and-Beyond

I recall that after official presentation of Kaveri some people here was still pretending that Kaveri was SOI because one slide from AMD mentioned "SHP".

Same about my claim on 20nm bulk. No roadmap or anything as that. At contrary, some people here used an article that suggested AMD was returning to SOI for 20nm

http://news.softpedia.com/news/AMD-and-GlobalFoundries-Interested-in-FD-SOI-for-the-20-nm-Process-267519.shtml

Idem about my claim about FINFETs on bulk being the future and SOI being 'dead'. No roadmap or anything like that. At contrary, the replies that I received included a dozen of links to articles that supposedly disproved me. This was one of the articles mentioned

http://www.advancedsubstratenews.com/2013/07/globalfoundries-on-cost-vs-performance-for-fd-soi-bulk-and-finfet/

Thus, contrary to your claim, I didn't believe what companies said, but only what passed my expert-eye filter. I am rather sure you can find some intermediate label between "technology Nostradamus" and someone who is right from pure chance. But the label that you can find for me is irrelevant, the relevant concepts are "bulk" and "FINFET".

You really are full of it. It was known back in March 2012 that AMD was moving to bulk for Kaveri.
http://www.brightsideofnews.com/news/2012/3/6/amd-ditches-soi-kaveri-goes-bulk-at-gf2c-more-details-from-the-new-wsa.aspx

Whether people here believed it or not, whether they misunderstood other details like the ambiguous "SHP" slide is irrelevant.

You are replying to only one of the three points mentioned in my post.

Your BSN link shows a wrong (outdated) roadmap where Kaveri is the successor to Trinity. I also recall how the people who disagreed with me said that the author of the quote in your BSN link was no longer working for AMD and thus that link with wrong roadmap couldn't be taken as valid anymore. Some of that same people said here that Kaveri was being delayed to coincide with the start of risk production of the supposed 28nm SOI process at GF. Check the pcper link.

There was similar hot discussions about the Kaveri/Steamroller fabrication process were made in other forums beyond this. I recall that I was in minority here defending the idea about bulk. It is was everything so evident and well-known as you pretend then only a very small minority of people would be expecting SOI Kaveri...

I love as everything is evident, known, and expected, but only a posteriori. It was not so evident, known, or expected during those long pages of discussion in different forums.

Finally, there is nothing ambiguous on the AMD slide about fabrication process for Kaveri. The original SHP process based on SOI was canceled and replaced by a SHP process based in bulk that GF tweaked during many months for Kaveri. SHP means Super High Performance. It is just some people that took it as synonym for SOI.

I think people are more confusing that it SOI, because of the SOI substrate,

 

[Cazalan]

I am thinking more along the lines of multiple apu sockets. A pair of 7850k's would be a 8 core gaming beast. I also think it would be of great benefit to any anyone already accustomed to buying 4-6 socket motherboards to scale performance. Especially going forward as more emphasis is place on gpu workloads.

Sockets end up being too specific and limiting in size, use, pin count and more importantly power delivery. With PCIe 4.0 and XDMA you can get the bandwidth you need and fully customizable layout.

 

[Cazalan]

Executing the OS in a TCU is far from optimal. Thus the KL Phi has to have a LCU inside if it is designed to work on standalone mode.* How Intel exactly achieves this is not still known, but I expect something similar to a big.LITTLE configuration.

Silvermont cores being OoO are fully capable of running a server OS, or the popular term now microserver. The prior one was not really capable of that because they were in order cores. They are LCUs. Lots of them (72+) running slowly (below Avoton speeds) to keep under 300W. Actually quite similar architecturally to the popular Blue Gene/Q line from IBM which also uses quad threaded cores.

 

[gamerk316]

did anyone see the graphene article? IBM is getting that worked out pretty well so far.

The lack of a native bandgap is always going to be a major design issue. We're getting there, but still a LOT of work to be done.

http://spectrum.ieee.org/semiconductors/devices/first-graphene-integrated-circuit
http://spectrum.ieee.org/semiconductors/processors/researchers-pencil-in-graphene-transistors

 

[juggernautxtr]

I am thinking more along the lines of multiple apu sockets. A pair of 7850k's would be a 8 core gaming beast. I also think it would be of great benefit to any anyone already accustomed to buying 4-6 socket motherboards to scale performance. Especially going forward as more emphasis is place on gpu workloads.

Sockets end up being too specific and limiting in size, use, pin count and more importantly power delivery. With PCIe 4.0 and XDMA you can get the bandwidth you need and fully customizable layout.

A dual socket apu can be easily added now and not need pcie 4, the interconnects are available already on c32 and g34 mb. If amd plans to use apu as server chips then a multi socket board will be needed. I just want it at consumer level price like intel did with the abit bp6

I think it was yuka that suggested such a thing, $400 dollars of cpu and graphics 8 core cpu power and dual graphics at under 200 watts stock. makes an over clocked 8350 with dgpu look pretty sad.
over clocked they would probably still look better.

 

[Cazalan]

A dual socket apu can be easily added now and not need pcie 4, the interconnects are available already on c32 and g34 mb. If amd plans to use apu as server chips then a multi socket board will be needed. I just want it at consumer level price like intel did with the abit bp6

I'm a fan of the older dual processor boards as well. I still have a BP6 lying around and a couple Athlon MP systems. It's just cost prohibitive these days. Back then power regulation was a joke. You need much higher quality VRMs these days.
If you put it on a PCIe card they can go in any system. And you can use more than 1 at a time. G34 is good but overkill once the RAM goes on package.

 

[Cazalan]

did anyone see the graphene article? IBM is getting that worked out pretty well so far.

The lack of a native bandgap is always going to be a major design issue. We're getting there, but still a LOT of work to be done.

http://spectrum.ieee.org/semiconductors/devices/first-graphene-integrated-circuit
http://spectrum.ieee.org/semiconductors/processors/researchers-pencil-in-graphene-transistors

Lots of graphene hype but not much product yet. Weren't they supposed to make heatsinks 25% more efficient too?

 

[Cazalan]

Actually if you been here you would notice that people here did and possibly still believe that making his point relevant,

There was a lot of HOPING and PRAYING it wasn't on bulk, because of the known clock speed reductions. We of course got bulk and the clock speed reduction *predictions* came true. And the candle a few people still hold for an enthusiast class (CPU) part from AMD got dimmer.

 

[juggernautxtr]

Actually if you been here you would notice that people here did and possibly still believe that making his point relevant,

There was a lot of HOPING and PRAYING it wasn't on bulk, because of the known clock speed reductions. We of course got bulk and the clock speed reduction *predictions* came true. And the candle a few people still hold for an enthusiast class (CPU) part from AMD got dimmer.

Bulk can't be all that bad if Intel is using it on their processors, only in finfet, not sure if GFs' SHP is finfet or not.
haven't seen any real description of the process.

 

[wh3resmycar]

^ During that time you could build a Tri core athlon machine and a 6570HD video card(both of which were 70$), This is around Llano times.

or step up to a 6670/6750.

 

[Cazalan]

Bulk can't be all that bad if Intel is using it on their processors, only in finfet, not sure if GFs' SHP is finfet or not.
haven't seen any real description of the process.

I wasn't saying bulk is bad. It's a compromise. Just that AMD's biggest IPC gain in recent history was hobbled by a process change. Two steps forward, 1 step back.

 

[Cazalan]

memory on package your talking intel phi and that will always be too expensive for consumers , g34 is almost out of reach. HSA would seem a good match for a multi apu system. The power saving would be substantial with something like the 45watt a10 7700.

Volta is supposed to have it. The card after Maxwell which is a year or so after Phi. It's not going to have 16GB like Intels Phi, but it doesn't need that much either.

 

[juggernautxtr]

Bulk can't be all that bad if Intel is using it on their processors, only in finfet, not sure if GFs' SHP is finfet or not.
haven't seen any real description of the process.

I wasn't saying bulk is bad. It's a compromise. Just that AMD's biggest IPC gain in recent history was hobbled by a process change. Two steps forward, 1 step back.

i think the choice made was because of the gpu and hsa elements of the chip, the power design envelope could be a reason too.
just wish GF get in gear and actually do AMD right instead of giving it in the backside every turn. GF has done more damage to AMD than AMD has done to themselves.

 

[palladin9479]

Athlon II X3 440(which can be overclocked quite well) could be grabbed at 70$ and that leaves 70$ for the GPU which i could almost always find a 5670 DDR3 or 6570 DDR3 during llano's life. Also when both components have separate heat-sinks they also allowed better overclocking which improves performance even more making llano during its time even less important for gamers. Again i own a Llano desktop and laptop(love the laptop) their amazing but for gamers on a budget they could have had more per dollar. However i feel like that will go away soon enough. Really all the A10 7850K needs to do is drop the price by 30-50$ and they have a winner.

Actually i built my friend a PC back in the day with a Athlon II x3 and it unlocked to a Phenom II x4 and i also found a 4830(DDR5) for just 65$ on newegg at the time was a lucky thing and not repeatable by any means just brings back memorys

Again a Llano is going to be better CPU wise then an Athlon II regardless. Llano is an enhanced Athlon II CPU. The dGPU was a DDR3 Turks Pro (HD 6570) bolted onto the side with a few shader cores sliced away, the sharers weren't a big deal as it was the memory interface that served as the bottleneck. The cost of the Llano was, on average, lower then the costs of a Athlon II x3 440 + 6570. I've demonstrated the prices of the various parts along with what you'd need to beat them, and it's more expensive then what the Llano was priced at. I've found that $140~150 USD is the magic line, anything higher starts to lean towards dCPU + dGPU, anything lower is solidly in the APU arena.

Also I treat sales, ebay and "special deals" the same way I treat Microcenter pricing, if you can get it great and you should definitely go for it. Never make recommendations or general purpose determinations based on them, as Microcenter just demonstrated, you can also find the APU's at discounted prices. The A10-7850K is priced at $185 USD, at that price it's non-competitive. At $130 there is absolutely nothing that can compete with it. Because it's MSRP is so high I wouldn't recommend any build with it as a general rule. If you got it for cheap then great for you, make it work the best you can. But don't go around saying the aformentioned APU's were bad because you built this one rig, for this one person, using on sale parts that got you the same or slightly less performance for cheaper. And yes your Athlon II X3 was less performance then the A8-3870K which opened at $140 USD MSRP. GDDR5 (the ones that will actually provide better graphics performance) dGPU's costed $85~$100 USD MSRP at that time (actually I can't remember any being under $100 USD). That leaves you 40~55$ USD for the CPU, and that's assuming you got really lucky on an $85 USD GDDR5 dGPU.

What your doing is remember the review performance of the GDDR5 version of the dGPU's, the ones the venders send to the websites, then using the purchasing price of the DDR3-1600 versions. The DDR3 GPU's had the same performance limits that Llano did, the 128-bit DDR3-1600 memory interface. There was a very active group of people tuning those APU's for maximum performance and we found that overclocking the iGPU core didn't do squat but overclocking the CPU-NB and Memory had a very noticeable effect on performance. Those dGPU's suffer from the same problem.

Funny you should mention the HD5670.
HD5670:
GPU Core: 775
Memory Speed: 800, 128-bit DDR3-1600
GPU Config: 400:20:8
MSRP: ~$85 USD (DDR3) $100 USD (GDDR5)

A8-3870K:
CPU Speed: 3Ghz
Cores: 4 K10.5
L2 Cache: 1MB per core, 4 MB total
L3 Cache: None
GPU Core: 600
Memory Speed: 800/933 128-bit DDR3-1600 / 1866
GPU Config: 400:20:8
MSRP: $140 USD

Congrats you just bought an overclocked HD6550D. You'll now have to pair it with a $55 USD CPU.

Here is that CPU you were mentioning

Athlon II x3 440
Clock Speed: 3Ghz
Cores: 3 K10
L2 Cache: 512KB per core, 1.5MB total
L3 Cache: None
Memory: 128-bit DDR3-1333
MSRP: $84

So you have a $84 + $85 USD solution that performs the same as a $140 USD solution. To get any real advantage you'd have to go with the $100 USD GDDR5 GPU for a $84+$100 USD solution.

This applies to all APU's. There is a distinct price band that appears around $140~160 USD (depending on available products on the market at that time). Anything under it favors APU, anything over it favors dGPU + dCPUs. I use MSRP because any sale that applies to one product may also apply to other products, and as always if you can get what would be a more expensive combo at a more competitive price from a particular vender, then by all means go for it.

As for Kaveri, the MSRP on the 7850K is too high, it's outside of that competitive price band. The A8-7600 on the other hand is at $120 MSRP which is actually under the price band and better then any conceivable combo on the market. If AMD lowered the MSRP on the 7850K to ~$150 USD, right at the price band then it would be competitive.

 

[palladin9479]

I am thinking more along the lines of multiple apu sockets. A pair of 7850k's would be a 8 core gaming beast. I also think it would be of great benefit to any anyone already accustomed to buying 4-6 socket motherboards to scale performance. Especially going forward as more emphasis is place on gpu workloads.

Sockets end up being too specific and limiting in size, use, pin count and more importantly power delivery. With PCIe 4.0 and XDMA you can get the bandwidth you need and fully customizable layout.

A dual socket apu can be easily added now and not need pcie 4, the interconnects are available already on c32 and g34 mb. If amd plans to use apu as server chips then a multi socket board will be needed. I just want it at consumer level price like intel did with the abit bp6

Would be far to expensive.

Guys' here's the issue with SMP layouts. You need coherency between all CPU's involved. A multi-core CPU has a single MMU and internal coherency to the various cache's. Each core is capable of looking up the contents of another cores cache, though the prefetcher and MMU should be taking care of that. On a multi-socket layout you need to manage that coherency not only between multiple cache's but also in a NUMA environment with multiple physical memory topologys.

That's all doable and HPC has been doing it for a long time, but it's not cheap. There is a reason multi-socket boards are a few hundred USD.

 

[de5_Roy]

Actually if you been here you would notice that people here did and possibly still believe that making his point relevant,

There was a lot of HOPING and PRAYING it wasn't on bulk, because of the known clock speed reductions. We of course got bulk and the clock speed reduction *predictions* came true. And the candle a few people still hold for an enthusiast class (CPU) part from AMD got dimmer.

whatever little i know about fabrication and chips makes me hopeful about an enthusiast sr-fx. but i think, from the same limits of knowledge, that sr and exc shouldn't allow traditional cpus. by traditional cpus i mean per amd's designs e.g. decoupled northbridge, no shared imc (pointless without the igpu, too high latency for dgpu) etc. while process is still limited no thanks to glofo. both sr and exc are about more integration which allows for deeper implementation of hsa which subsequently should drive amd's next cpu/apu/soc roadmap.
iirc, amd showed how sr cores and the gcn igpu works individually but never showed how they are connected together... or what's inside the u.n.b.
i'd still like an sr-fx. if amd can make it am3+socket compatible, all the better. theoretically, even bulk shp should allow for 6-8m/12-16c big dies. from the die shot, looks like 4M/8C should be easy with decent yields despite no improvement in the per core performance.

 

[Reynod]

No. My claim about bulk Kaveri wasn't on any roadmap. At contrary, some people took an public GF roadmap and said me that Kaveri was SOI

Some people also linked to this article to prove me wrong

http://www.pcper.com/reviews/Editorial/GLOBALFOUNDRIES-2013-and-Beyond

I recall that after official presentation of Kaveri some people here was still pretending that Kaveri was SOI because one slide from AMD mentioned "SHP".

Same about my claim on 20nm bulk. No roadmap or anything as that. At contrary, some people here used an article that suggested AMD was returning to SOI for 20nm

http://news.softpedia.com/news/AMD-and-GlobalFoundries-Interested-in-FD-SOI-for-the-20-nm-Process-267519.shtml

Idem about my claim about FINFETs on bulk being the future and SOI being 'dead'. No roadmap or anything like that. At contrary, the replies that I received included a dozen of links to articles that supposedly disproved me. This was one of the articles mentioned

http://www.advancedsubstratenews.com/2013/07/globalfoundries-on-cost-vs-performance-for-fd-soi-bulk-and-finfet/

Thus, contrary to your claim, I didn't believe what companies said, but only what passed my expert-eye filter. I am rather sure you can find some intermediate label between "technology Nostradamus" and someone who is right from pure chance. But the label that you can find for me is irrelevant, the relevant concepts are "bulk" and "FINFET".

You really are full of it. It was known back in March 2012 that AMD was moving to bulk for Kaveri.
http://www.brightsideofnews.com/news/2012/3/6/amd-ditches-soi-kaveri-goes-bulk-at-gf2c-more-details-from-the-new-wsa.aspx

Whether people here believed it or not, whether they misunderstood other details like the ambiguous "SHP" slide is irrelevant.

ColinAP you now join jdwii on the last warning list for personal attacks.

One more and your on a holiday for two weeks with noob2222..

In this thread I leave these here as evidence so other mods can simply pull the trigger as the traffic is so high, and the posts frequent.

Embarrasing eh?

Juanrga ... if you didn't brag so much people wouldn't attack you either ... so look up the word "humility" and buy some.

 

[jdwii]

Athlon II X3 440(which can be overclocked quite well) could be grabbed at 70$ and that leaves 70$ for the GPU which i could almost always find a 5670 DDR3 or 6570 DDR3 during llano's life. Also when both components have separate heat-sinks they also allowed better overclocking which improves performance even more making llano during its time even less important for gamers. Again i own a Llano desktop and laptop(love the laptop) their amazing but for gamers on a budget they could have had more per dollar. However i feel like that will go away soon enough. Really all the A10 7850K needs to do is drop the price by 30-50$ and they have a winner.

Actually i built my friend a PC back in the day with a Athlon II x3 and it unlocked to a Phenom II x4 and i also found a 4830(DDR5) for just 65$ on newegg at the time was a lucky thing and not repeatable by any means just brings back memorys

Again a Llano is going to be better CPU wise then an Athlon II regardless. Llano is an enhanced Athlon II CPU. The dGPU was a DDR3 Turks Pro (HD 6570) bolted onto the side with a few shader cores sliced away, the sharers weren't a big deal as it was the memory interface that served as the bottleneck. The cost of the Llano was, on average, lower then the costs of a Athlon II x3 440 + 6570. I've demonstrated the prices of the various parts along with what you'd need to beat them, and it's more expensive then what the Llano was priced at. I've found that $140~150 USD is the magic line, anything higher starts to lean towards dCPU + dGPU, anything lower is solidly in the APU arena.

Also I treat sales, ebay and "special deals" the same way I treat Microcenter pricing, if you can get it great and you should definitely go for it. Never make recommendations or general purpose determinations based on them, as Microcenter just demonstrated, you can also find the APU's at discounted prices. The A10-7850K is priced at $185 USD, at that price it's non-competitive. At $130 there is absolutely nothing that can compete with it. Because it's MSRP is so high I wouldn't recommend any build with it as a general rule. If you got it for cheap then great for you, make it work the best you can. But don't go around saying the aformentioned APU's were bad because you built this one rig, for this one person, using on sale parts that got you the same or slightly less performance for cheaper. And yes your Athlon II X3 was less performance then the A8-3870K which opened at $140 USD MSRP. GDDR5 (the ones that will actually provide better graphics performance) dGPU's costed $85~$100 USD MSRP at that time (actually I can't remember any being under $100 USD). That leaves you 40~55$ USD for the CPU, and that's assuming you got really lucky on an $85 USD GDDR5 dGPU.

What your doing is remember the review performance of the GDDR5 version of the dGPU's, the ones the venders send to the websites, then using the purchasing price of the DDR3-1600 versions. The DDR3 GPU's had the same performance limits that Llano did, the 128-bit DDR3-1600 memory interface. There was a very active group of people tuning those APU's for maximum performance and we found that overclocking the iGPU core didn't do squat but overclocking the CPU-NB and Memory had a very noticeable effect on performance. Those dGPU's suffer from the same problem.

Funny you should mention the HD5670.
HD5670:
GPU Core: 775
Memory Speed: 800, 128-bit DDR3-1600
GPU Config: 400:20:8
MSRP: ~$85 USD (DDR3) $100 USD (GDDR5)

A8-3870K:
CPU Speed: 3Ghz
Cores: 4 K10.5
L2 Cache: 1MB per core, 4 MB total
L3 Cache: None
GPU Core: 600
Memory Speed: 800/933 128-bit DDR3-1600 / 1866
GPU Config: 400:20:8
MSRP: $140 USD

Congrats you just bought an overclocked HD6550D. You'll now have to pair it with a $55 USD CPU.

Here is that CPU you were mentioning

Athlon II x3 440
Clock Speed: 3Ghz
Cores: 3 K10
L2 Cache: 512KB per core, 1.5MB total
L3 Cache: None
Memory: 128-bit DDR3-1333
MSRP: $84

So you have a $84 + $85 USD solution that performs the same as a $140 USD solution. To get any real advantage you'd have to go with the $100 USD GDDR5 GPU for a $84+$100 USD solution.

This applies to all APU's. There is a distinct price band that appears around $140~160 USD (depending on available products on the market at that time). Anything under it favors APU, anything over it favors dGPU + dCPUs. I use MSRP because any sale that applies to one product may also apply to other products, and as always if you can get what would be a more expensive combo at a more competitive price from a particular vender, then by all means go for it.

As for Kaveri, the MSRP on the 7850K is too high, it's outside of that competitive price band. The A8-7600 on the other hand is at $120 MSRP which is actually under the price band and better then any conceivable combo on the market. If AMD lowered the MSRP on the 7850K to ~$150 USD, right at the price band then it would be competitive.

Perhaps i think you're missing the main point and i'll say it again i never even seen the Athlon II x3 which can be overclocked to a good 3.6ghz easily for more than 70$ during that time and i could always find a 6570-5670 for around 70$ on newegg you are using listed pricing i'm using the actual prices that i could always get for them, throw in the need for cheaper ram and you can even get more performance per dollar.

 

[jdwii]

Also more information on LLano vs phenom clock per clock
http://www.anandtech.com/show/4444/amd-llano-notebook-review-a-series-fusion-apu-a8-3500m/2
Reality is it was only 3% faster on average according to that site(although i hate anandtech)
And even Amd said 6% just to put that in perspective Llano was clocked at 3.0Ghz max that would mean with a 6% improvement in IPC a 4 core Llano would equal a 3.2Ghz processor phenom processor. Now according to these article from tomshardware
http://www.tomshardware.com/reviews/athlon-l3-cache,2416-9.html
Theres an average performance increase of 5.8% when the K10 has L3 cache meaning a Athlon is only 5.8% weaker compared to the Phenom on average.
Now add 6% and 6% and you get 12% LLano is around 12% more powerful clock per clock compared to the athlon i'm comparing meaning a 3.0Ghz llano would equal a 3.4Ghz athlon.....Games only used 3 cores or less during that time meaning llano was a slight boost in speed but overall nothing to brag about and surly not enough to justify it for its processor.