Recent content by Rufus

I hope you use this as a primary source: Transcript of the recording: http://archive.computerhistory.org/resources/access/text/2014/05/102746834-05-01-acc.pdf

Page 9 shows throughputs for Geforce vs Quadro. Only difference I see (other than due to number of SMs) is FP16 Tensor w/ FP32 accumulate. https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/technologies/turing-architecture/NVIDIA-Turing-Architecture-Whitepaper.pdf

You mean RISC-V cores https://riscv.org/wp-content/uploads/2017/05/Tue1345pm-NVIDIA-Sijstermans.pdf

Not unless someone makes a new coin where the mining is based on ray tracing or tensor calculations. The only mining algorithm that really matters is Ethereum - it's 80% of the GPU mining revenue. Ethereum's algorithm is "memory hard" which requires memory bandwidth and a just a few XORs. So...

Interesting look back on the origins of Ray Tracing and Global Illumination: https://blogs.nvidia.com/blog/2018/08/01/ray-tracing-global-illumination-turner-whitted/

This type of movement isn't unusual. Just look at Mike Houston - left AMD as a Fellow to start at Nvidia in the same month. https://www.linkedin.com/in/mike-houston-537693/

A year ago I would have agreed with you. Today Ryzen can serve this market just as well as Intel can, so all this combo chip does is undercut their own CPU.

Re-reading Kyle's article about Raja trying to spin RTG off in order to provide custom parts for Intel is fascinating in light of the Core+Radeon chip anouncment: https://www.hardocp.com/article/2016/05/27/from_ati_to_amd_back_journey_in_futility It's even more so if the rumor of him going to...

Correct. The metric that matters is USD/day per MH/s. https://bitinfocharts.com/comparison/ethereum-mining_profitability.html#1y

And a follow up presentation from CppCon on designing Volta for C++:

Direct link so it doesn't get lost in the future: http://en.community.dell.com/techcenter/high-performance-computing/b/general_hpc/archive/2017/09/29/hpc-applications-performance-on-v100

The new vega ops: https://github.com/RadeonOpenCompute/ROCm_Documentation/blob/master/GCN_ISA_Manuals/Vega_Shader_ISA_28July2017.pdf V_XAD_U32: D.u32 = (S0.u32 ^ S1.u32) + S2.u32 V_SAT_PK_U8_I16: D.u32 = {16'b0, sat8(S.u[31:16]), sat8(S.u[15:0])} V_LSHL_ADD_U32: D.u = (S0.u << S1.u[4:0]) + S2.u...

AMD finally released a beta driver fixing the ETH >2GB DAG performance issue. http://www.legitreviews.com/amds-new-mining-block-chain-optimized-driver-tested_197095 http://support.amd.com/en-us/kb-articles/Pages/Radeon-Software-Crimson-ReLive-Edition-Beta-for-Blockchain-Compute-Release-Notes.aspx

Point. From a software or program point of view it's deadlocked since no state is changing. From a hardware point of view it's livelocked since instructions are being issued. I'm a hardware guy so I think more on that side. But whatever you call it I think that producer/consumer example is...

The producer waiting for the consumer to yield execution is exactly what's happening. The consumer thread will continuously read *flag, see it's 1, read it again, see it's 1. The producer never executes. The system as a whole is executing instructions so the system is live, but no thread is...