Overclock.net › Forums › Industry News › Hardware News › [TweakTown] TSMC delays the mass production of 16nm FinFET chips by up to 6 months
New Posts  All Forums:Forum Nav:

[TweakTown] TSMC delays the mass production of 16nm FinFET chips by up to 6 months - Page 5

post #41 of 44
Quote:
Originally Posted by aberrero View Post

I guess it really depends what you consider to be Nehalem. Westmere was 32nm whereas Nehalem/Clarksfield was 45nm. Are we comparing quad cores? Desktop CPUs? Extreme Editions?

In any case, Nehalem is 6 years old now, and even if modern CPUs are 100% faster, that's still a lot less than what many of us would have predicted at the time. If they had doubled in performance every two years then current CPUs would be 8 times faster.
Not sure if you are refering to The moore's law but anyway... it just say that semiconductor will double transistor count every 18-24month. It doesnt talk about IPC gain(or any performance gain actually). On the mainstream platform, since 45nm, they keep shrinking those dual/quad core and use the extra transistor to add a beefier iGPU. On the enthusiast platform they just add more cores.

I wonder if they could use like 90% of the die size of an 18core Xeon and make a really huge single core LOL. Talk about beast IPC for games wink.gif
Edited by DarkBlade6 - 10/21/14 at 2:10am
post #42 of 44
If we are being pedantic. Moore's law is that the transistor count would double at the same price point. For a long time on CPUs that translated into pipeline a stages and logic improvements bringing ever increasing clock speeds and IPC improvements. Then it brought us extra cores, now the CPU die is shrinking and it's bringing us power saving and a little IPC.

GPUs theoretically can keep scaling for quite a while so long as the transistor counts keep coming. We are around 2000-3000 shaders and the initial simple scaling will work right up to around 1 shader per pixel on the screen. Many shader programs could also run in parallel so it can scale somewhat past that point as well, assuming the memory bandwidth can keep up.

Things are stagnant because tsmc hasn't delivered, but once they do I expect w nice big jump in performance again, currently no reason to believe that won't happen.
BC Primary
(20 items)
 
  
CPUMotherboardGraphicsRAM
6700k Gigabyte Gaming 7 MSI 1080 Gaming X 16GB Corsair 3200 
Hard DriveHard DriveHard DriveHard Drive
Hitachi 4TB Crucial MX100 512GB Crucial M4 512GB Samsung 950 pro 
CoolingOSMonitorMonitor
Custom watercooling 2xMCR320 1xMCR480 Windows 10 Asus Rog Swift PG279Q Benq XL2411T 
KeyboardPowerCaseMouse
Corsair K70 Red cherry Corsair HX1050 Little Devil 7 Zowie Evo CL EC2 
Mouse PadAudioAudioAudio
QCK Heavy Soundblaster ZX Sennheiser HD598 Schitt Magni/Mobi 
  hide details  
Reply
BC Primary
(20 items)
 
  
CPUMotherboardGraphicsRAM
6700k Gigabyte Gaming 7 MSI 1080 Gaming X 16GB Corsair 3200 
Hard DriveHard DriveHard DriveHard Drive
Hitachi 4TB Crucial MX100 512GB Crucial M4 512GB Samsung 950 pro 
CoolingOSMonitorMonitor
Custom watercooling 2xMCR320 1xMCR480 Windows 10 Asus Rog Swift PG279Q Benq XL2411T 
KeyboardPowerCaseMouse
Corsair K70 Red cherry Corsair HX1050 Little Devil 7 Zowie Evo CL EC2 
Mouse PadAudioAudioAudio
QCK Heavy Soundblaster ZX Sennheiser HD598 Schitt Magni/Mobi 
  hide details  
Reply
post #43 of 44
Thread Starter 
Quote:
Originally Posted by MapRef41N93W View Post

Sounds like you're describing AMD not NVIDIA there.
both did that... or you dont know what was a 770 GTX vs 680 GTX ,and they slighly downgraded a 670 GTX to a 760 GTX
Wanted: [WTB] GPU upgrade
$210 (USD) or best offer
  
Reply
Wanted: [WTB] GPU upgrade
$210 (USD) or best offer
  
Reply
post #44 of 44
Quote:
Originally Posted by aberrero View Post

I know, but the reason anybody confuses the two ideas in the first place is that they used to be synonymous.

"Performance doubles every 18-24 months"

"Smaller nodes mean higher clock speeds"

"Power consumption on a given node is directly proportional to transistor count"

All of these things used to be true but aren't anymore. The question is, is it actually impossible to make faster chips the way we used to, or does the industry just not care about the desktop anymore, innovating in mobile instead?


With CPUs, though mobile is the greater focus, I think Intel can do better if it weren't so overly the focus. Even though they've slowed down and are using their research and physical silicon budgets on iGPUs and low power-consumption, they're still making small but steady IPC improvements as well as introducing and improving new instructions. They've also specifically stated they aim to increase performance, but only while matching or furthering a correlating decrease in power consumption; this implies they could do more if they didn't have to adhere to those imposed limits and weren't so focused on more iGPU shaders. They're also probably playing things safe in terms of architectural innovations. I don't think it's likely CPUs could keep advancing as they once used to (depending on what time period we're talking about), but I'm certain Intel could do better than they have been. Better than 5-10% better IPC with lower overclocking headroom at least.

With GPUs, I don't think a mobile focus is the problem; rather, delayed node improvements and maximization of profit causing artficially expanded architecture generations. It's no coincidence AMD and Nvidia are now barely leapfrogging each other with new releases. Nvidia clearly have the capacity to release a significantly more powerful 28nm Maxwell part even given how small and power-efficient current 28nm GPUs are.
Edited by Serandur - 10/21/14 at 4:07pm
Ice-nine
(17 items)
 
  
CPUMotherboardGraphicsRAM
i7-6700K Asus Z170-DELUXE Gigabyte G1 GTX 980 Ti Corsair Vengeance LPX DDR4 3000 
Hard DriveHard DriveOptical DriveCooling
Samsung 850 Evo 500 GB Seagate Barracuda 7200 RPM 2 TB Asus DVD Drive DRW-24B1ST/BLK/B/AS Phanteks PH-TC14PE 
OSMonitorKeyboardPower
Windows 8.1 Pro AOC Q2770PQU Corsair K95 RGB - Cherry MX Blue EVGA Supernova G2 750W 
CaseMouseMouse PadAudio
NZXT Phantom 410 Gunmetal Edition Logitech G502 Proteus Core Razer Goliathus Speed Edition Small Cyber Acoustics Satellite CA-3602 
Audio
Oppo PM-3 Closed Back Planar Magnetic Headphones 
  hide details  
Reply
Ice-nine
(17 items)
 
  
CPUMotherboardGraphicsRAM
i7-6700K Asus Z170-DELUXE Gigabyte G1 GTX 980 Ti Corsair Vengeance LPX DDR4 3000 
Hard DriveHard DriveOptical DriveCooling
Samsung 850 Evo 500 GB Seagate Barracuda 7200 RPM 2 TB Asus DVD Drive DRW-24B1ST/BLK/B/AS Phanteks PH-TC14PE 
OSMonitorKeyboardPower
Windows 8.1 Pro AOC Q2770PQU Corsair K95 RGB - Cherry MX Blue EVGA Supernova G2 750W 
CaseMouseMouse PadAudio
NZXT Phantom 410 Gunmetal Edition Logitech G502 Proteus Core Razer Goliathus Speed Edition Small Cyber Acoustics Satellite CA-3602 
Audio
Oppo PM-3 Closed Back Planar Magnetic Headphones 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [TweakTown] TSMC delays the mass production of 16nm FinFET chips by up to 6 months