Also, each Graphics Processing Cluster, or GPC, contains up to 4 SMX units in Kepler, and up to 5 SMM units in first generation Maxwell. Nvidia claims a 128 CUDA core SMM has 86% of the performance of a 192 CUDA core SMX. SMM allows for a finer-grain allocation of resources than SMX, saving power when the workload isn't optimal for shared resources. Texture units and FP64 CUDA cores are still shared. These units are connected by a crossbar that uses power to allow the resources to be shared. This is in contrast to Kepler, where each SMX has 4 schedulers that schedule to a shared pool of 6 sets of 32 FP32 CUDA cores, 2 sets of 16 load/store units, and 2 sets of 16 special function units. The layout of SMM units is partitioned so that each of the 4 warp schedulers in an SMM controls 1 set of 32 FP32 CUDA cores, 1 set of 8 load/store units, and 1 set of 8 special function units. The structure of the warp scheduler is inherited from Kepler, which allows each scheduler to issue up to two instructions that are independent from each other and are in order from the same warp. Nvidia also changed the streaming multiprocessor design from that of Kepler (SMX), naming it SMM. Accordingly, Nvidia cut the memory bus from 192 bit on GK106 to 128 bit on GM107, further saving power. Nvidia increased the amount of L2 cache from 256 KiB on GK107 to 2 MiB on GM107, reducing the memory bandwidth needed. These new chips provide few consumer-facing additional features Nvidia instead focused on power efficiency. my safe settings is 148 core and 425 memory note under watercooling.Main article: Maxwell (microarchitecture) First generation Maxwell (GM10x) įirst generation Maxwell GM107/GM108 were released as GeForce GTX 745, GTX 750/750 Ti and GTX 850M/860M (GM107) and GT 830M/840M (GM108). + note if u have evey good cooler that keeps the memory cool overclock the memory more then the core, u see much more out the card. Now take your max memory and then core and take about 3% off core and 1.5% off memory and see how it runs should be stable if u did alright testing. Now find max core overclock with memory at stock should be about 1000 - 1020 or so with the voltage u picked. Now run a stress test see what is self overclocks to. Number four up the voltage to what u feel safe at Number three set the temp to what u fell save at 87 is a ok number if u dont mind the fans. but right now it sounds like u are runnning in to the TDP or your card at your settings. Un evga click the voltage tab and set as u want +13 +25 or +38 up to u but not always more = better.
as u are seeing right now it down clocking becasue one u are to hot. It is the deal how to over clock u must fine the sweet spot.