Im not suggesting a different die, im saying a cut down die with only Dual CPU Cores and cut down iGPU to 8x CUs could be more than 30% faster in Graphics than BristolRidge if paired with DDR-4 3200MHz memory.
I'm not sure AMD has any immediate plans to produce clusters of Zen cores smaller than 4, but otherwise . . . gosh I hope so! It'd better be faster than Carrizo/Bristol Ridge.
Ehm I mean to disable 2x CPU cores and left with only 2x Cores + HT(4x Threads). Same with the iGPU, disable 3x CUs and left with 8x CUs (if rummors of 11CUs is correct).
Why in the hell would they do such thing?
HBM is out of the question at least for Raven Ridge and it's possible successor in the consumer segment. AMD needs either to implement eDRAM or GDDR5X, or alternatively forget about the APUs completely. HBM is way too complex (expensive) and DDR4 frequencies won't reach high enough to provide sufficient bandwith on a 128-bit bus.
Sad but true.
The AMD Fusion Fund enables investment in companies that are developing innovative solutions and applications that leverage the features of the AMD Fusion family of APUs. The program is intended to accelerate time to market for solutions such as:
– Applications that are newly enabled or greatly enhanced by the synergy between CPU and GPU
– Tools that enable developers to easily develop, scale, optimize or test applications
– Technology or intellectual properties that can be a foundational building block of applications
– Form factors or devices that expand the scope for deployment of applications
HBM2 Doubles Bandwidth, Capacity
May 3, 2016
Author: David Kanter
High Bandwidth Memory 2 (HBM2) has several incremental but critical improvements over the previous-generation technology: it increases the memory capacity and bandwidth, adds ECC support, and improves memory-controller efficiency. These enhancements are essential for high-performance computing (HPC) and workstation tasks and will lead to wider industry adoption. Samsungs HBM2 chips are already in production, and the memory will appear in new graphics cards later this year.
Over the last decade, GDDR has been the high-performance graphics memory of choice, balancing bandwidth and cost for low-capacity systems. GDDR5 is the latest incarnation, but it cant scale much past 8Gbps; in practice, no GPU has achieved more than 320GB/s using GDDR5. The industry is currently developing GDDR5X, which will roughly double the bandwidth, but this memory is unlikely to reduce power consumption at higher data rates.
HBM uses 2.5D or 3D packaging to increase bandwidth and power efficiency by implementing a wide slow interface, but the interposer raises the cost. Since modern computing systems are almost always limited by power consumption, HBM gives architects an effective tool to improve system performance. For example, the first GPU to use HBM was AMDs Fiji, which delivered an incredible 512GB/s while using 3x less power than an equivalent GDDR interface. But HBM primarily serves in consumer graphics because of limitations in the first-generation technologylimitations that are alleviated by HBM2.
Why in the hell would they do such thing?
Every dual core (Single Module) APU today is coming from a Dual Module Quad Core die. There are Dual cores with 256 iGPU Shaders APUs, exactly half the Kaveri die. A6-7400K is one of them, this is a harvested Kaveri exactly in Half, but keeping the 128bit memory.
Not exactly true, since AMD just created Stoney Ridge to avoid this.
Dual ZEN Cores + HT and 384 shaders (6x CUs) with Dual channel DDR-4 memory. Something like that ??
All I want is an APU that has at least GTX 750 Ti performance with at least last gen i3 CPU perf levels at less than 100 watts of total system power draw.
Can we get there soon please? xD
1000MHz on a 11 GCU GPU would be a major waste of power. The relative memory bandwidth even at the maximum theoretical 3200MHz memory frequency would be ~13% lower than on a FX-8800P Carrizo (2133MHz). And even the FX-8800P is completely clogged up in terms of the available GPU bandwidth.
Anything higher than < 800MHz (at DDR-3200) would be waste of power. Since the maximum theoretical frequency supported by the controller is most likely not even supported / enabled on Raven, the highest "power efficient" frequency for the GPU would be even lower.
Games and applications are becoming more Compute heavy (DX-12/Vulkan, OpenCl etc). If we have 1000MHz on BristolRidge (8x CUs) at 28nm, there is no reason not have 1000MHz on Raven Ridge (11x CUs) at 14nm. This is for the Desktop high-End 95W TDP APUs, lower SKUs could have lower clocks. Example, the 8x CU 35W TDP Raven Ridge could only have 800MHz iGPU.
Both Fudzilla and SemiAccurate say it tops out at 65W, so not sure where are you seeing 95W models.
Games and applications are becoming more Compute heavy (DX-12/Vulkan, OpenCl etc). If we have 1000MHz on BristolRidge (8x CUs) at 28nm, there is no reason not have 1000MHz on Raven Ridge (11x CUs) at 14nm. This is for the Desktop high-End 95W TDP APUs, lower SKUs could have lower clocks. Example, the 8x CU 35W TDP Raven Ridge could only have 800MHz iGPU.