Nvidia slams Intel: says ARM supercomputer may be more efficient

http://www.eetimes.com Nvidia believes ARM, not Intel, holds the future for supercomputing, with Sumit Gupta saying ARM CPUs are inherently more efficient than x86 processors. Noting that ARM architecture has arisen out of the embedded space where power limitations are key, Gupta said maximizing performance was better understood than it was in an Intel world where machines tended to be plugged in to wall sockets. X86's main advantage, in Gupta's opinion, was in making operating systems like Microsoft Windows run better, while ARM is much more suited to the HPC world of the future where power becomes a lot more precious.

Category:

Science & Technology

Tags:

• Nvidia
• intel
• arm
• architecture
• cpus
• gpus
• gpgpu
• sc11
• hpc
• windows
• operating system
• os
• Barcelona
• Cuda
• software
• x86
• power
• performance
• watt
• embedded

[VenatorFX2012]

@abram730 Nvida's Wafers Are to Large 500M.M. vs. 380M.M.! You don't quote any website dudes are you a joke? AMD had the Tek Lead with its ThunderBird Core-Athlon XP-Athlon FX series from 2000-2006 before Intel Released its Intel Core Duo, was cheaper to produce! AMD could have some driver dates that speed up performance in the works 10-15% I'd be careful before you call AMD out of exectution! By June AMD could have more faster proccessors out to lower the price 1/2cycle ahead of Nvidia!

[abram730]

@VenatorFX2012 Charlie Demerjian, has always praised AMD and held Nvidia to the fire. He claims the replacement for the 560, the gk104 is on a smaller die and beating the 7970. Well I have doubts they'll call it a 660 and run it against a 7870 if that is true. AMD costumers would revolt and demand refunds if they did that.

I guess we will see if it has 1536 CUDA cores and is on a smaller die.

Is English your second language?

[VenatorFX2012]

Hey GeckoDude2010 Don't bother with this Guy, Imagination Technologies is Faster, Plus Die size of ATI 7000 Series is smaller than Kepler Lineup! Giving Him spec fasts on the net! This Dude looks like a Like your car kinda stuff from Boogie Nights to start things up!This is the type of guy that likes to cough and fart simultaneously at a convention! later man no more interesting facts!

[abram730]

@VenatorFX2012 Is Imagination Technologies counting programmable GigaFlops? Sony and Microsoft calculated everything on their consoles and started claiming teraFLOPS of power on their GPU's.

Nvidia only lists programmable power and doesn't include HW specific calculations

Apples to Oranges.

They are using a smaller die on their 60's if leaks are correct. The larger die for high end extends into HPC and has high profits.

[VenatorFX2012]

@abram730 Seems Imagination Technologies Mobile Chip is Hitting 100GigaFlops and 1TerraFlop later on, Tegra 3 doesn't hit 100 GigaFlop Proccessing Power! Nvidia also needs to Die Shrink its Chip Size Wafer too Big vs. AMD with Maxwell! AMD has a 4Month Product Cycle Edge vs. Nvidia!

[abram730]

@GeckoDude2010 Sterne Agree is probably just being short sighted(Q1 desktops bad for Nvidia), although the timing seems a bit odd relating to other news and the sharp rise of NVDA shares.

Their excessive use of the word catalyst, was a bit funny.

Tegra 3 is demonstrating dynamic task migration, and that is a step away from dynamically shifting compute to the cloud.

Investors should understand what is going on, but they rarely do. They never make it past the buzz words.

[abram730]

@GeckoDude2010 Reaching out to the HPC community was a smart move on Nvidia's part. Not just for the revenue, but also the know how and feedback from a community composed of the brightest minds. AMD is following Nvidia in. Linking high end desktop and HPC is helpful and will yield good CPU acceleration. You can end up with a very flexible chip because you can dynamically shift computing resources based on task without the bottlenecks.

[abram730]

@GeckoDude2010 Basic GPGPU issues should be addressed with GCN. Both corps have a good future. I think the shared vision and good competition is a long term strength. I do think Intel has bigger long term issues as GPU's are gaining performance so much faster then CPU's. The demanding math most people need power for, can be done with GPU's and DX11.1 makes them more independent.

An i7-980 does something like 50-60 billion operations vs trillions. The door is there.

[abram730]

@VenatorFX2012 "GPGPU Instructions Seem to give AMD a 20%boost when Gpu is used to help out the CPU!"

"GPUs Are Only Up To 14 Times Faster than CPUs” says Intel"

500% wouldn't be unexpected when all is said and done for highly integrated tasks. GCN is less bursty and with more memory bandwidth in a higher TDP.

Heterogeneous MPSoC is really going to take off, pushed by, perf per watt in low power mobile, server, and HPC.

[abram730]

@VenatorFX2012 "You Joking Man" I was saying that you know what information I was referencing as your "7830x2" fit what I hadn't said. Competitive? yes. The GCN based parts will be a big jump. If AMD does a high end product beating Intel by orders of magnitude general consistence will change drastically. That will require more memory bandwidth and some of the improvements in the pipe. Why settle for just Competitive?