FLOPS

Fujitsu [富士铜:-)] board with SPARC64 VIIIfx processors. 

 

很可能是液冷——右中部朝上那个接头。SPARC64 VIIIfx, which was developed as a processor for the K computer, uses Fujitsu Semiconductor Ltd.’s 45-nm CMOS process for semiconductors and is composed of eight cores, a 6MB shared level 2 cache, and memory controllers. Peak performance of 128GFLOPS at an operating frequency of 2 GHz is achieved with power consumption as low as 58W. 128G / 8 / 2G = 8 FLOPS / cycle.

 Intel E8400:

• Code Name: Products formerly Wolfdale [Penryn]
• Lithography: 45 nm
• Total Cores: 2
• Processor Base Frequency: 3.00 GHz
• Cache: 6 MB L2 Cache
• Bus Speed: 1333 MHz
• TDP: 65 W Thermal Design Power (TDP) represents the average power, in watts, the processor dissipates when operating at Base Frequency with all cores active under an Intel-defined, high-complexity workload.
• TCASE: 72.4°C Case Temperature is the maximum temperature allowed at the processor Integrated Heat Spreader (IHS).
• Package Size: 37.5mm x 37.5mm
• Processing Die Size: 107 mm2
• # of Processing Die Transistors: 410 million. 平均每个晶体管占2.61x10^-7mm²，即261000平方纳米，45*45/261000=0.00775862.

【二手95新】英特尔XEON E31230 V2 1231 V3 1230 V3 12401271 xeon E3 1230 V2 - 京东 (jd.com)

英特尔® 至强® 处理器 E3-1230 v2（8M 高速缓存，3.30 GHz） 产品规范 (intel.com)

英特尔® 至强® 处理器 E3-1246 v3（8M 高速缓存，3.50 GHz） 产品规范 (intel.com)

...英特尔决定中止代号为Knights Landing（KNL）的Xeon Phi 7200系列处理器研发计划，这款CPU主要用于HPC应用，包括超级计算机。原因是市场对这类型CPU的需求并不太高。与此同时，英特尔将继续提供代号为Knights Mill（KNM）的深度学习解决方案。这些CPU具备64-72个HPC增强型Atom Silvermont内核，可同时处理多达四个代码线程，并配备16 GB高带宽MCDRAM内存。插座式Xeon Phi KNL处理器可提供3-3.4 TFLOPS的双精度浮点（FP64）性能。

Nvidia Ampere (only GeForce) PTX (Parallel Thread Execution, an intermediary assembler language by NVIDIA)

• FP64: 1/32
• FP32: 2 (FP32) + 0 (INT32) or 1 (FP32) + 1 (INT32)
• FP16: 16