Will Cannonlake (10nm) be the first CPU that never becomes obsolete? Even if we make it to 5nm, that will only be like a ~10% gain in IPC based on jewtels track record.
>https://www.technologyreview.com/s/601441/moores-law-is-dead-now-what/
>>55368817
I expect 5% performance increase every generation from Intel now.
There's still no reason to upgrade from Haswell if you don't need more PCI-e lanes.
The biggest pusher now is chipset features.
>>55368879
What about video decoders and more instruction sets?
>>55368879
There's literally no reason to upgrade from Sandy Bridge.
>tfw memephene never
I want a huge jump in performance. I know Skylake is about 20% faster than Sandy, but that took several years. I'm not going to upgrade a 3570k for that low of a performance bump over such a long timespan.
No. The big three fabs (Intel, TSMC, Samsung) will all have their own 7nm processes around the end of the decade.
Zen is coming boys
16/32 core consumer CPU will save us all
>>55369798
>16/32 core
>consumer
Nigga please. Consumers will get 8/16 at most for at least $500. The huge multicore monsters are for servers. I don't mind, however; I want to upgrade from Ivy Bridge.
>>55369717
greetings ratotron
>>55369709
There is literally no reason to upgrade from Nehalem.
>>55369717
Greetings ratotron
>>55369717
greetings ratotron
>>55369798
I don't think programs ever use more than one (and the same) core.
Not everyone is Terry to support a gazillion of cores.