AMD’s Kaveri: Pre-Launch Information
by Ian Cutress on January 6, 2014 8:00 PM EST
On the back of AMD’s Tech Day at CES 2014, all of which was under NDA until the launch of Kaveri, AMD have supplied us with some information that we can talk about today. For those not following the AMD roadmap, Kaveri is the natural progression of the AMD A-Series APU line, from Llano, Trinity to Richland and now Kaveri. At the heart of the AMD APU design is the combination of CPU cores (‘Bulldozer’, ‘Steamroller’) and a large dollop of GPU cores for on-chip graphics prowess.
Kaveri is that next iteration in line which uses an updated FM2+ socket from Richland and the architecture is updated for Q1 2014. AMD are attacking with Kaveri on four fronts:
Redesigned Compute Cores* (Compute = CPU + GPU)
Kaveri uses an enhanced version of the Richland CPU core, codename Steamroller. As with every new CPU generation or architecture update, the main goal is better performance and lower power – preferably both. AMD is quoting a 20% better x86 IPC with Kaveri compared to Richland when put clock to clock. For the purposes of this information release, we were provided with several AMD benchmarking results to share:
These results border pretty much on the synthetic – AMD did not give any real world examples today but numbers will come through in time. AMD is set to release two CPUs on January 14th (date provided in our pre-release slide deck), namely the A10-7700K and the A10-7850K. Some of the specifications were also provided:
| AMD APUs | ||||
|
Richland A8-6600K |
Richland A10-6800K |
Kaveri A10-7700K |
Kaveri A10-7850K |
|
| Release | June 4 '13 | June 4 '13 | Jan 14th '14 | Jan 14th '14 |
| Frequency | 3900 MHz | 4100 MHz | ? | 3700 MHz |
| Turbo | 4200 MHz | 4400 MHz | ? | ? |
| DRAM | DDR3-1866 | DDR3-2133 | DDR3-2133 | DDR3-2133 |
| Microarhitecture | Piledriver | Piledriver | Steamroller | Steamroller |
| Manufacturing Process | 32nm | 32nm | ? | ? |
| Modules | 2 | 2 | ? | 2 |
| Threads | 4 | 4 | ? | 4 |
| Socket | FM2 | FM2 | FM2+ | FM2+ |
| L1 Cache |
2 x 64 KB I$ 4 x 16 KB D$ |
2 x 64 KB I$ 4 x 16 KB D$ |
? | ? |
| L2 Cache | 2 x 2 MB | 2 x 2 MB | ? | ? |
| Integrated GPU | HD 8570D | HD 8670D | R7 | R7 |
| IGP Cores | 256 | 384 | ? | 512 |
| IGP Architecture | Cayman | Cayman | GCN | GCN |
| IGP Frequency | 844 | 844 | ? | 720 |
| Power | 100W | 100W | ? | 95W |
All the values marked ‘?’ have not been confirmed at this point, although it is interesting to see that the CPU MHz has decreased from Richland. A lot of the APU die goes to that integrated GPU, which as we can see above becomes fully GCN, rather than the Cayman derived Richland APUs. This comes with a core bump as well, seeing 512 GPU cores on the high end module – this equates to 8 CUs on die and what AMD calls ’12 Compute Cores’ overall. These GCN cores are primed and AMD Mantle ready, suggesting that performance gains could be had directly from Mantle enabled titles.
Described in AMD’s own words: ‘A compute core is an HSA-enabled hardware block that is programmable (CPU, GPU or other processing element), capable of running at least one process in its own context and virtual memory space, independently from other cores. A GPU Core is a GCN-based hardware block containing a dedicated scheduler that feeds four 16-wide SIMD vector processors, a scalar processor, local data registers and data share memory, a branch & message processor, 16 texture fetch or load/store units, four texture filter units, and a texture cache. A GPU Core can independently execute work-groups consisting of 64 work items in parallel.’ This suggests that if we were to run asynchronous kernels on the AMD APU, we could technically run twelve on the high end APU, given that each Compute Core is capable of running at least one process in its own context and virtual memory space independent of the others.
The reason why AMD calls them Compute Cores is based on their second of their four pronged attack: hUMA.
HSA, hUMA, and all that jazz
AMD went for the heterogeneous system architecture early on to exploit the fact that many compute intensive tasks can be offloaded to parts of the CPU that are designed to run them faster or at low power. By combining CPU and GPU on a single die, the system should be able to shift work around to complete the process quicker. When this was first envisaged, AMD had two issues: lack of software out in the public domain to take advantage (as is any new computing paradigm) and restrictive OS support. Now that Windows 8 is built to allow HSA to take advantage of this, all that leaves is the programming. However AMD have gone one step further with hUMA, and giving the system access to all the memory, all of the time, from any location:
Now that Kaveri offers a proper HSA stack, and can call upon 12 compute cores to do work, applications that are designed (or have code paths) to take advantage of this should emerge. One such example that AMD are willing to share today is stock calculation using LibreOffice's Calc application – calculating the BETA (return) of 21 fake stocks and plotting 100 points on a graph of each stock. With HSA acceleration on, the system performed the task in 0.12 seconds, compared to 0.99 seconds when turned off.
Prong 3: Gaming Technologies
In a year where new gaming technologies are at the forefront of design, along with gaming power, AMD are tackling the issue on one front with Kaveri. By giving it a GCN graphics backbone, features from the main GPU line can fully integrate (with HSA) into the APU. As we have seen in previous AMD releases and talks, this means several things:
- Mantle
- AMD TrueAudio
- PCIe Gen 3
AMD is wanting to revolutionize the way that games are played and shown with Mantle – it is a small shame that the Mantle release was delayed and that AMD did not provide any numbers to share with us today. The results should find their way online after release however.
Prong 4: Power Optimisations
With Richland we had CPUs in the range of 65W to 100W, and using the architecture in the FX range produced CPUs up to 220W. Techincally we had 45W Richland APUs launch, but to date I have not seen one for sale. However this time around, AMD are focusing a slightly lower power segment – 45W to 95W. Chances are the top end APUs (A10-7850K) will be 95W, suggesting that we have a combination of a 20% IPC improvement, 400 MHz decrease but a 5% TDP decrease for the high end chip. Bundle in some HSA and let’s get this thing on the road.
Release Date
AMD have given us the release date for the APUs: January 14th will see the launch of the A10-7850K and the A10-7700K. Certain system builders should be offering pre-built systems based on these APUs from today as well.
Facebook
Twitter
RSS
133 Comments
View All Comments
mikato - Wednesday, January 8, 2014 - link
I agree. The new tech features are where I see the big upgrade advantage. I don't know if the CPU side will be as much of an improvement as everyone wants, IPC is up but clock speed down. However, the CPU performance has been quite good since the last flagship APU A10 6800K. Gamers are looking for another 6 or 8 core though to put any CPU disadvantage to rest, and no integrated GPU required. With AMD not having such processor in any plans right now, gamers building systems have to accept the APU or go Intel. AMD may be choosing to do this to push the new tech and get adoption. That's probably the focus they need and the long game they need to play. AMD will lose gamers, but they are a small percentage of customers. All just theories. Perhaps if AMD can fully move FPU work to the integrated graphics on APUs, then the need for more cores evaporates.extide - Tuesday, January 7, 2014 - link
I envision a point where CPU's don't even have an FPU on them anymore. They utilize the iGPU for all floating point math, scalar AND vector. Seems to me that this is where AMD is trying to go.mikato - Wednesday, January 8, 2014 - link
Yep, I believe this has been part of the plan since their new architecture began (module based architecture and APUs with solid GPUs integrated on die go together). You have to admire their long term vision for being the underdog in the market.JohnHardkiss - Tuesday, January 7, 2014 - link
Is it true that the Kaveri A10-7800 will have the same igpu as the Kaveri A10-7850k? When looking at this leaked data it seems to be the case: http://wccftech.com/amd-a10-7800-kaveri-apu-benchm... . Two points:1) What intel haswell cpu will the A10-7800 compare to, and
2) It is said to be released on 14th january as well, but I can't find any new news about it.
Anyone any insights about these points? Thanks.
azazel1024 - Tuesday, January 7, 2014 - link
A couple of things to keep in mind. The current Iris pro Intel 4570R is actually pretty cheap. As cheap as these? Doubtful, but they are priced at $288 a tray. Not exactly $400+ to get in to Iris pro on the desktop.Next, for performance disadvantage in single thread. Yeah, it is pretty bad. Kaveri makes up some of it. However, keep in mind, AMD is claiming a 20% boost in IPC, but it is also cut back 10% on clock speed. So actual performance is only advancing about 10% or so. Its something, but it isn't any more than Intel's move from Ivy to Haswell was in effect. Which means AMD hasn't really gained ground between generations from Richland to Kaveri as Intel went from Ivy to Haswell. There might have been a slight improvement, but slight.
Intel on the other hand has cut AMD's lead on iGPUs in the move from Ivy to Haswell. Intel generally seems to have gained more in the move from Ivy to Haswell than AMD is gaining in the move from Richland to Kaveri. AMD still generally has the lead, but Intel is cutting it down.
Broadwell is claimed to be a pretty huge gain in iGPU for Intel again, which if that proves to be true, Intel might actually have a lead in performance or be very, very close behind.
Unless Kaveri brings a lot better power use tech in to their APUs, Richland was pretty far behind Ivy and Haswell makes it darned right embaressing the difference in idle, light work load and performance per watt under heavy load. I don't see that Kaveri is making that much better. They have slightly better TDP, better efficiency under load, but likely they still won't be as efficient in performance per watt as Ivy in most work loads. Under light/idle its likely to be pretty bad...which factors in to the price advantage for AMD if you are looking at business machines or machines that are going to be on most/all day long.
A savings of $10 or 15 in power over a year doesn't sound like much, but if a machine has a 3 year expected service life, or even 4 or 5 that gets to be a lot of power savings. A $190 processor ends up being as cost effective as a $235 processor after 3 years.
Kaveri seems to be a good step forward, but it isn't a performance "win" to "dethrone" Intel. It pretty much leaves AMD in the same bucket. They can rule on entry level systems, gamers on a steep budget and some HTPC systems depending on the HTPC requirements (thinking gaming for an HTPC on a modest budget or very constrained chasis that can't accept a dGPU).
In general Kaveri doesn't seem to be better for workstations than higher end Intel offerings, and I don't mean Ivy Bridge-E either. Also for a machine where cost is slightly less of an issue, probably still better performance with a higher end i5 or i7 Haswell processor and a dGPU.
Or on a modest machine, a $130 odd Haswell i3 is probably going to give generally better system performance than a similar priced Kaveri will, and most users aren't going to care that the Kaveri might have 20-50% better iGPU performance. Still not a lot of stuff that is GPGPU capable through OCL and most of the stuff that is, either the user won't notice a hair of difference in performance or else it isn't stuff average users of modest machines are going to be using (and higher end machines again would benefit more from a $100 or 200 dGPU and something like an i54670).
UtilityMax - Wednesday, January 8, 2014 - link
Very good analysis. Intel is behind in iGPU area, but not by much. HD Graphics 4600 and Iris Graphics show that Intel does have good potential for developing a good iGPU. If the Kaveri APU concept does take off, Intel won't take very long time to respond. Maybe just a year. And in the end, APU gamers is just a fraction of a PC gamer market, and PC gamer market is just a fraction of the overall PC market. Basically, the AMD APUs are marketed to folks who care enough about gaming to want an APU instead of Intel Core, but not enough to be willing to buy a discrete video card. How many are there? To be honest, a $170 APU that alone can power a capable gaming rig is intriguing, but I dount the leap in performance will be huge compared to Richland A10.Another hurdle that AMD needs to overcome is Intel's aggressive pricing. In the end, the $130 Core i3 is a pretty damn good CPU for most folks who just run productivity and a few multimedia apps. The A10 APU would have been a steal if AMD offered it at the same price as Core i3. But at $160-170, the "productivity" folks will walk away with the Core i3, while many gamers will scratch their head about buying this APU vs Intel Core I3 with dedicated low-end GPU. Power users will skip both and jump straight to Core i5 or i7.
yankeeDDL - Wednesday, January 8, 2014 - link
If I may comment here ...Intel's GPU offering is comparable to AMD only in terms of Iris Pro. Everything else is much behind (http://www.anandtech.com/show/6993/intel-iris-pro-...
The i7-4850HQ is > 300usd more expensive than the 5800K, and I don't quite buy the yearly cost estimates: unless you play 24/7, the delta cost in electricity is going to be much lower.
There are other savings in AMD's platform at MoBo level, for various reasons, so the saving doesn't only come from the CPU.
If Mantle really provides the 45% improvement it is going to make the difference even larger.
Basically, except for the few Iris Pro parts, it really makes no sense to compare Intel with AMD's APUs, from a GPU's perspective, while a 20% penalty on the CPU side seems much more manageable.
If (and it's a big "if") the HSA really gets some traction at software level, the CPU's shortcomings can be easily offset by the GPU.
If you think about it, already today most intensive every-day apps are already leveraging the GPU (web browing, spreadsheets, Flash/Silverlight and even image processing).
So for everyday tasks I doubt anyone can really see any speed advantage comparing a ~$130 AMD APU with a ~$500 Intel APU. Throw in there the gaming advantage from AMD's platform and I see a fairly decent prospect for AMD. Of course, they need to execute: there have to be some decent PC/laptop offerings with good APUs and balanced configurations (no more of those 2GB craps, please).
The real questions, from where I sit, are: does HSA really work as expected? Is the gap to Hashwell CPU really reduced by ~20%? Is the memory bandwidth going to bottleneck the GPU's performance, or will it really be able to hit the levels of the 7750?
So I wouldn't say that
mikato - Wednesday, January 8, 2014 - link
It begs the question why Intel doesn't bring their good integrated graphics to lower end CPUs for an appropriate price. Maybe they don't see money in it from their perspective. Joe Schmo wouldn't buy a slower CPU for a little more $, and probably the system builders see that as well. AMD has some additional reasons to do this with their long game. They will eventually flip the switch to make their superior GPUs translate into much more powerful overall APUs. They get closer with each architecture update, and with more HSA adoption. It will be interesting to see how things play out. We'll certainly know more come Excavator.http://www.pcper.com/reviews/Editorial/AMDs-Proces...
UtilityMax - Wednesday, January 8, 2014 - link
What I said is that Intel has more than enough in-house capability to build a good iGPU. If APU market really takes off, which is not given, Intel has the capacity to respond very fast. And Intel is not that much behind right now. The A10-6800K APU is on average about 30% faster than HD Graphics 4600. So the rift is there, but it isn't that bit. AMD said previously that the Steamroller APU will improve GPU performance by 20-30%. So if you believe AMD's words, then the Kaveri A10s may be faster than Intel HD Graphics by some 50% on average. As for Mantle, it probably will take a long time to take off and become mainstream. Only new titles will use it, but I suspect people who intend to play on APUs have in mind older games as well, and those may not benefit from Maltle. It's possible that Intel will try to respond. It's also possible that the APU market will remain so small that Intel won't bother.SoBizarre - Wednesday, January 8, 2014 - link
Can it do 8 taps Jinc in MadVR?