Will zen offer the same value as vishera?
Is it going to be Ivy bridge ipc with ddr4?
How about APUs? How much faster will they be with ddr4? Why can't I find any proper dx12 or vulkan benchmarks for APUs?
What do we even know about zen so far?
>>56281023
If it has good overclocking and overclocking provides more than just marginal returns then my expectations are that they will be bringing Intel prices down or forcing Intel to actually provide performant updates rather than just tick-tocking.
If Zen is good and didn't get bogged down single threaded in games it'll probably be offered in midrange/budget gaming PCs.
>>56281135
But will it offer better value, or will i3 still hold that crown? What do we even know so far?
>>56281023
>Will zen offer the same value as vishera?
Su has stated that their upcoming Summit Ridge will have a higher ASP than prior FX parts. The FX 8350 had a launch MSRP of $195.
At the very least it'll cost more than that, likely a lot more.
>Is it going to be Ivy bridge ipc with ddr4?
Its going to be Excavator with 40% higher IPC, and DDR4.
>How about APUs? How much faster will they be with ddr4?
There is no word yet on how Raven Ridge's memory configuration will address the IGP's need of bandwidth. Socket AM4 only supports dual channel memory, and two channels of 4000mhz DDR4 only provide 64GB/s. Not enough by a long shot.
They haven't commented on the use of HBM anywhere in their material as of yet, so that seems very unlikely.
That aside, they'll still be a huge improvement over current chips on the CPU front. IGP should still show subtantial gains, but will remain memory bottlenecked without HBM.
>Why can't I find any proper dx12 or vulkan benchmarks for APUs?
Probably because no one cares.
>What do we even know about zen so far?
Everything aside from its multicore scaling and SMT throughput
http://www.anandtech.com/show/10591/amd-zen-microarchiture-part-2-extracting-instructionlevel-parallelism
>>56281163
The increased bandwidth of ddr4 over ddr3 should help though.
Could they make some special APU though, that has something like 1GB memory on board and uses ram as low priority storage or something?
>>56281879
>The increased bandwidth of ddr4 over ddr3 should help though.
The increased bandwidth comes entirely from frequency. 2400mhz DDR3 in dual channel provices 38.4GB/s 64GB/s for 4000mhz DDR4. The fastest DDR4 out there still isn't enough to feed an IGP in dual channel mode.
The Radeon 7750 with 8CU just like AMD's Kaveri and Carrizo, the GDDR5 variants have 72GB/s to provide adequate bandwidth to the CUs without choking them.
Raven Ridge has more CU, and would have considerably less bandwidth if it only has DDR4 supply it.
As I said HBM is the only solution to this, but AMD haven't commented on it at all. So its unlikely that they'll feature HBM on these chips.
>>56281933
That's an almost 50% increase in memory bandwidth. It should help a lot, not far off from the 72GB/s of the 250x which will be a pretty nice improvement.
Almost 50% according to this tomshardware review.
I'm really looking forward to seeing some dx12 or vulkan benches for low level shit like APUs though.
http://www.tomshardware.com/reviews/radeon-r7-250x-graphics-card-review,3747-3.html