If AMD has a deal with certain processors for bundled games, e.g. A10/A8, then those Power Packs will also have the codes. Tray processors typically do not come with codes, so it's part of the PIB/Tray meet-in-the-middle.
Still Richland? It's built on GF's old 32 nm process, has already somewhat outdated Piledriver CPU cores and rather outdated VLIW4 GPU part, originating back from pre-GCN Radeon HD 6970 Cayman GPU released in December 2010 - around 4 years ago (Richland GPU is ~ 1/4 Cayman GPU, similarly as Kaveri GPU is ~ 1/4 Tahiti GPU, and it all started with Llano GPU being ~ 1/4 Cypress GPU). I thought they are about to switch to Kaveri completely at this time in terms of manufacturing (after 3 years of production of 32 nm APUs, counting Trinity and Llano before Richland). HSA & GCN (both of which are absent in Richland), blah blah blah...
Have you looked at the benchmarks? From Llano on upwards there is not much distinction in overall processing power....although graphics has increased. I love the apu but got a bit turned off when they moved from Llano (which I have) to Trinity and had to boost the CPU by about 1ghz over the Llano just to have the benchmarks come out the same. And they are still beating down the same processor path...I just don't get it. I don't see how bull/pile/steamroler path greatly exceeds that of the Phenom path at 45 nm that they had over 3 years ago. I'm very frustrated by AMD's progress.
Perhaps if they can ever get a processor down to 22 nm and stop the shared portion of their core/module they will have a strong/competitive CPU portion of the APU. But Intel is already down to 14 nm and it could be another year+ for AMD to just get down to 22 nm.
I haven't owned an intel computer since the mid-1980's but have given up hope in the last few years that AMD will ever have a high end competitive CPU
Of course I know the benchmarks, in particular the x86 CPU cores benchmarks and I know the story about AMD modules not keeping up with their older Phenom II cores at the same frequency, and I agree with you. That's why I shifted from AMD to Intel Haswell and Ivy Bridge last year, while buying new PCs, by the way. But I meant really mostly not their CPU cores, but their new APU features in Kaveri (HSA and GCN), which they are promoting, on the one hand - but they still continue to produce Richland, completely lacking these features, on the other hand. So, they still continue to mass produce their old tech, while promoting their new more promising tech - that's what I meant when I wrote "blah blah blah" - they are not keeping up themselves with their own advertisements about new APU features, IMHO...
The GPU core is still the same as what Brazos/Llano was built on...which shows how ahead of its time Brazos was, and how behind AMD is now. Even their graphics division is still pimping a 3 year old architecture. I know GCN is now technically up to v1.2, but we're talking such incremental changes that it makes nVidia's rebadging of Kepler over the years look acceptable.
The real pain is going to be when nVidia launches Maxwell. This will be the equivalent to the graphics division of what Intel's Core microarchitecture did to their CPU division.
AMD is getting hammered. They won't be able to use ATI as a financial crutch much longer.
In terms of raw performance, I don't believe Maxwell will be quite the monster people think. In terms of power efficiency it should be very impressive, however. In the short term it's not going to bring AMD crashing down. If anything I would hope the competition will drive AMD to push out an impressive new GPU within the next year. Remember how rough ATI looked in the HD 2000/3000 days? A lot of people thought they were done for, and then they released the excellent HD 4000 series.
On the CPU/APU side, things are a lot tougher. Intel really is a juggernaut and they have been a lot more careful in recent years - they haven't been caught off guard in quite a while. With that being said, Keller does have his team working on a new ground-up architecture - though it won't bear fruit for some time. In the meantime all they can do is iterate on existing designs and integrate other improvements (HSA enhancements, stacked memory, etc).
HD 2000 and HD 3000 series had somewhat similar performance (e.g. top GPUs had both 320 shaders) with the corresponding HD 3000 GPUs being somewhat faster. But HD 3000 was built on a new (then) 55 nm process, unlike the older HD 2900's 80 nm, so HD 3000 series was much more power efficient and gave much better impression overall.
If you mean Trinity/Richland's GPU, then, no, it's not exactly the same as what Brazos/Llano was built on. Trinity/Richland GPU is VLIW4, while Brazos/Llano were VLIW5. VLIW4 in Trinity/Richland is slightly more efficient, than older VLIW5, as reviews and comparisons show.
Not too sure if this is a huge deal for boutique builders. Save $1-2 per system. However its an old chip so if someone buys a 6 pack and can only sell 5 systems they have extra inventory on hand that they will likely have to get rid of at a loss, making the pack useless.
I like this power pack idea for the lowest end SKUs the most because at this level of cost the extra dollar or two makes more of a difference.
However, at the same time I feel AMD's lowest end processors are a worse value in other ways. This due to AMD's binning process (on the big core models) which is much more extreme than Intel's approach. (Example: So much of the iGPU is disabled on the dual big core (single module) SKUs, which are all based on a quad core dies to begin with. Furthermore, refreshes like Richland which should have improved yields on 32nm don't change the extreme disabling of the iGPU at all)
Regarding the Power Packs for AM1, I think that would be a good idea for Sempron 2650 for the same reason I mentioned in my other response.
But with that mentioned, I think AMD needs to make another more pivotal choice when it comes to increasing value beyond Power Pack. For Sempron 2650, I would be referring to increasing clockspeed on the dual cpu cores. With two cpu cores on a quad core die rated at 25 watts, why are the clocks so slow? Even Intel gives me the full clockspeed for Bay Trail-D when they disable two cores and make a J1800 SKU. I don't understand why AMD doesn't do the same?
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
19 Comments
Back to Article
MrSpadge - Monday, September 15, 2014 - link
"or cause over-demand"With all due respect, I do not think AMD has to be afraid of over-demand for any of their current CPU products.
The_Assimilator - Monday, September 15, 2014 - link
I was going to go with something like "now AMD can honestly say they ship a SKU that's faster than one of Intel's". Well played sir.xdrol - Monday, September 15, 2014 - link
Over-demand only means that the demand is higher than what they are able to manufacture.For some AMD chips, this is indeed the case.
t.s. - Tuesday, September 16, 2014 - link
Which chip?DrApop - Monday, September 15, 2014 - link
What are "bundled codes"? Is that like a code for a game that goes with the processor or something?Ian Cutress - Tuesday, September 16, 2014 - link
If AMD has a deal with certain processors for bundled games, e.g. A10/A8, then those Power Packs will also have the codes. Tray processors typically do not come with codes, so it's part of the PIB/Tray meet-in-the-middle.TiGr1982 - Monday, September 15, 2014 - link
Still Richland? It's built on GF's old 32 nm process, has already somewhat outdated Piledriver CPU cores and rather outdated VLIW4 GPU part, originating back from pre-GCN Radeon HD 6970 Cayman GPU released in December 2010 - around 4 years ago (Richland GPU is ~ 1/4 Cayman GPU, similarly as Kaveri GPU is ~ 1/4 Tahiti GPU, and it all started with Llano GPU being ~ 1/4 Cypress GPU).I thought they are about to switch to Kaveri completely at this time in terms of manufacturing (after 3 years of production of 32 nm APUs, counting Trinity and Llano before Richland).
HSA & GCN (both of which are absent in Richland), blah blah blah...
DrApop - Monday, September 15, 2014 - link
Have you looked at the benchmarks? From Llano on upwards there is not much distinction in overall processing power....although graphics has increased. I love the apu but got a bit turned off when they moved from Llano (which I have) to Trinity and had to boost the CPU by about 1ghz over the Llano just to have the benchmarks come out the same. And they are still beating down the same processor path...I just don't get it. I don't see how bull/pile/steamroler path greatly exceeds that of the Phenom path at 45 nm that they had over 3 years ago. I'm very frustrated by AMD's progress.Perhaps if they can ever get a processor down to 22 nm and stop the shared portion of their core/module they will have a strong/competitive CPU portion of the APU. But Intel is already down to 14 nm and it could be another year+ for AMD to just get down to 22 nm.
I haven't owned an intel computer since the mid-1980's but have given up hope in the last few years that AMD will ever have a high end competitive CPU
TiGr1982 - Monday, September 15, 2014 - link
Of course I know the benchmarks, in particular the x86 CPU cores benchmarks and I know the story about AMD modules not keeping up with their older Phenom II cores at the same frequency, and I agree with you. That's why I shifted from AMD to Intel Haswell and Ivy Bridge last year, while buying new PCs, by the way.But I meant really mostly not their CPU cores, but their new APU features in Kaveri (HSA and GCN), which they are promoting, on the one hand - but they still continue to produce Richland, completely lacking these features, on the other hand.
So, they still continue to mass produce their old tech, while promoting their new more promising tech - that's what I meant when I wrote "blah blah blah" - they are not keeping up themselves with their own advertisements about new APU features, IMHO...
Samus - Monday, September 15, 2014 - link
The GPU core is still the same as what Brazos/Llano was built on...which shows how ahead of its time Brazos was, and how behind AMD is now. Even their graphics division is still pimping a 3 year old architecture. I know GCN is now technically up to v1.2, but we're talking such incremental changes that it makes nVidia's rebadging of Kepler over the years look acceptable.The real pain is going to be when nVidia launches Maxwell. This will be the equivalent to the graphics division of what Intel's Core microarchitecture did to their CPU division.
AMD is getting hammered. They won't be able to use ATI as a financial crutch much longer.
Alexvrb - Tuesday, September 16, 2014 - link
In terms of raw performance, I don't believe Maxwell will be quite the monster people think. In terms of power efficiency it should be very impressive, however. In the short term it's not going to bring AMD crashing down. If anything I would hope the competition will drive AMD to push out an impressive new GPU within the next year. Remember how rough ATI looked in the HD 2000/3000 days? A lot of people thought they were done for, and then they released the excellent HD 4000 series.On the CPU/APU side, things are a lot tougher. Intel really is a juggernaut and they have been a lot more careful in recent years - they haven't been caught off guard in quite a while. With that being said, Keller does have his team working on a new ground-up architecture - though it won't bear fruit for some time. In the meantime all they can do is iterate on existing designs and integrate other improvements (HSA enhancements, stacked memory, etc).
just4U - Tuesday, September 16, 2014 - link
Actually that was their 2x series Alex. The 3x came close to matching the geforce8TiGr1982 - Tuesday, September 16, 2014 - link
HD 2000 and HD 3000 series had somewhat similar performance (e.g. top GPUs had both 320 shaders) with the corresponding HD 3000 GPUs being somewhat faster. But HD 3000 was built on a new (then) 55 nm process, unlike the older HD 2900's 80 nm, so HD 3000 series was much more power efficient and gave much better impression overall.http://www.anandtech.com/show/2376
TiGr1982 - Tuesday, September 16, 2014 - link
If you mean Trinity/Richland's GPU, then, no, it's not exactly the same as what Brazos/Llano was built on. Trinity/Richland GPU is VLIW4, while Brazos/Llano were VLIW5. VLIW4 in Trinity/Richland is slightly more efficient, than older VLIW5, as reviews and comparisons show.whyso - Monday, September 15, 2014 - link
Not too sure if this is a huge deal for boutique builders. Save $1-2 per system. However its an old chip so if someone buys a 6 pack and can only sell 5 systems they have extra inventory on hand that they will likely have to get rid of at a loss, making the pack useless.Computer Bottleneck - Monday, September 15, 2014 - link
I like this power pack idea for the lowest end SKUs the most because at this level of cost the extra dollar or two makes more of a difference.However, at the same time I feel AMD's lowest end processors are a worse value in other ways. This due to AMD's binning process (on the big core models) which is much more extreme than Intel's approach. (Example: So much of the iGPU is disabled on the dual big core (single module) SKUs, which are all based on a quad core dies to begin with. Furthermore, refreshes like Richland which should have improved yields on 32nm don't change the extreme disabling of the iGPU at all)
Samus - Monday, September 15, 2014 - link
If Intel did this to the Pentium 3258(K) it'd be pretty popular.Computer Bottleneck - Monday, September 15, 2014 - link
Regarding the Power Packs for AM1, I think that would be a good idea for Sempron 2650 for the same reason I mentioned in my other response.But with that mentioned, I think AMD needs to make another more pivotal choice when it comes to increasing value beyond Power Pack. For Sempron 2650, I would be referring to increasing clockspeed on the dual cpu cores. With two cpu cores on a quad core die rated at 25 watts, why are the clocks so slow? Even Intel gives me the full clockspeed for Bay Trail-D when they disable two cores and make a J1800 SKU. I don't understand why AMD doesn't do the same?
Wolfpup - Tuesday, September 16, 2014 - link
For some reason I love the idea of a sexy black box with 6 CPUs in it! :-D