Experts Exchange has released a series of infographics showing off some examples of Moore’s law, comparing the theoretical computing power of today’s legion of mobile devices versus the gaming consoles and supercomputers of yesteryear. The graphics are sure to delight any fan of visualized data.
Moore’s law for those who fell asleep in Computer Science 101 was a prediction/observation by Intel Corp. (INTC) cofounder and top circuit researcher Gordon E. Moore. In 1965 Moore wrote in an article [PDF]:
The complexity for minimum component costs has increased at a rate of roughly a factor of two per year. Certainly over the short term this rate can be expected to continue, if not to increase. Over the longer term, the rate of increase is a bit more uncertain, although there is no reason to believe it will not remain nearly constant for at least 10 years.
In 1975 he revised that estimate [PDF] to predict circuit density would double every two years (rather than the original rate of doubling annually), a shift that followed his prediction that the rate of circuit shrinkage would eventually drop to a somewhat slower pace.
A young Gordon E. Moore back in 1975. [Image Source: IEEE]
Moore’s law is commonly supplemented with his Intel research colleague David House’s riff/extension which predicts that every 18 months approximately, computing performance will double. Both predictions have largely held up, in spite of perpetual naysayers predicting the supposed “death of Moore’s law”.
(Click to Enlarge) [Image Source: Experts Exchange]
The infographics from Experts Exchange illustrate those trends. Among the examples they note:
- Apple, Inc.’s (AAPL) iPhone 4 (2010) has the theoretical processing power (1.6 GFLOPS) roughly equivalent to the Cray-2 supercomputer (1985) (1.9 GFLOPS)
- The Cray-2 was the refined supercomputer from master computer scientist and architect Seymour Cray. It used a 3D quad-processor circuit cooled by fluorinated coolant.
- The Cray-2 had roughly 50 million transistors, running at nearly 250 MHz
- The iPhone 4’s A4 system-on-a-chip (SoC) has only about 15 million transistors [source], but they run at around 800 MHz
- The key to such high speeds with passive cooling is the feature size. Transistors are 45 nm in the A4 SoC, while the Cray-2 was like built on a 1.5 um MOSFET process (based on comments by Cray himself). 1.5 um = 1500 nm, so the effective die size of the A4 is perhaps a thousand times as dense per plane (assuming similar scaling in the pitch size).
- The Apollo Guidance Computer (AGC) (1966) used on the Moon missions is roughly the power of a Nintendo Comp., Ltd. (TYO:7974) Entertainment System (NES) (the original “Nintendo”)
- The AGC was clocked at 2 MHz. A NES is clocked at 1.8 MHz
- The AGC has twice the RAM — 4 kilobytes (KB) compared to 2 KB in the NES.
- The AGC weighed 70 lb (32 kg) vs. the NES which weighed 11.7 lb (5.31 kg).
- A Samsung Electronics Comp., Ltd.’s (KRX:005930) (KRX:005935) Galaxy S6 smartphone (2015) has more theoretical processing power (34.8 GFLOPS) as five Sony Corp. (TYO:6758) PlayStation 2s (5x 6.2 GFLOPS) (2000)
- The GS6 also has eight times as many cores (the PS2 had just 1 CPU core).
- The GS6 CPU cores are clocked at 1.5 GHz (x4) and 2.1 GHz (x4) vs. 300 MHz in the PS2.
- The GS6 has as much memory as 96 PS2s, w/ 3 GB of DRAM vs. 32 MB in the PS2.
- One Apple Watch (2015) has the theoretical computing power (3 GFLOPS) of nearly 2 Apple iPhone 4 smartphones (2010).
- That means the the Apple Watch has the theoretical compute power of nearly 2 Cray-2 supercomputers.
- That power comes thanks to the onboard processor that can clock up to 1 GHz.
- Believed to be produced by Samsung, the Apple S1 “system in a package” (SIP) that powers the Apple Watch has an embedded application processor (AP) produced from 28 nm transistors — which probably equates to more thana two fold increase in transistor density [source] over the 45 nm process used for the A4.
- The Tianhe-2 Chinese supercomputer (2013) has roughly the computational power (33.86 peak PFLOPS) as 18,400 Sony PlayStation 4s (2014).
- It’d take 175,000 PS4s, though to equal the 1.4 petabytes (PB) of DRAM in the Tianhe-2 supercomputer.
- The Tianhe-2 is composed of 32,000 Xeon and 48,000 Xeon Phi multicore processors from Intel, making for a total of 3.12 million cores.
- The PS4 has a 1.6 GHz octacore chip from Advanced Micro Devices, Inc. (AMD)
- Tianhe-2 consumers 24 Megawatts of power (17.6 Megawatts go towards actual computations — the rest goes to cooling). That’s enough to power 175,000+ PS4s (which draw roughly 137 watts of power while actively gaming (source)).
Technically speaking Moore’s law and the House additions don’t cover magnetic storage, but the site also offers up this entertaining animated GIF infographic showing the progress on that front:
[Image Source: Experts Exchange]
Or to summarize all this in the words of legally-troubled rapper Bobby Shmurda, “All these computer / Oh, They want that hot s**t.”