iGPU is built into the system-on-chip Intel GPUs, and the “lake” 8th generation – the processor R Kaby Lake, Coffee Lake, Amber Lake and Whiskey Lake. But what generation of Intel graphics treated these iGPU, I did not understand. 9.5+?
This is a continuation of a series about Coffee Lake, previous parts here.
The Coffee Lake processors were used, not all embedded processors for “lakes” of the 8th generation, but since Coffee Lake was not the last “lakes” of this generation used in Macs, briefly describe all of their options. The most notable changes compared to the previous graphics – renaming. For some reason, as soon as starts to smell roasted, decided to call for help marketers. More marketing than sales? I worked side-by-side with the marketing services they really needed and useful. But too much is never good. This renaming was, in my opinion, pure overkill.
GPU which is called Intel HD Graphics < number>, for example Intel HD Graphics 630, became known as UHD Intel Graphics < number>. For example, Intel UHD Graphics 630. They have become a little faster due to the increase of operating frequencies is approximately 50 MHz and improvements in the process, but no change in their architecture associated with the UHD, no one found. Great marketing move? That it is pointless, the very next day after processors with Intel UHD Graphics on Board fell into the hands of competent users have found out whom it was interesting. And those who are not versed in these delicate matters, and read only the headlines, I came to the conclusion that these GPUs aren’t good, and the effect of the renaming was rather negative.
Actually, all Intel UHD Graphics < number> was at least no worse than Intel HD Graphics < number> with the same numeric indices, on the contrary – they were better, but not essential. In addition to UHD Intel Graphics < number> processors Coffee Lake was built in and Intel Iris Plus, with other than before numeric indices, and with real improvements. If marketing renaming was not, would not become worse. Added support for HDCP 2.2, there are some changes related to UHD was still, though not too important in my opinion.
All GPU “8 lakes generation”, and about GT1
As in Kaby Lake, a schedule which called generation graphics 9.5 and which has been a real visible to the user the progress of all levels invented at Intel, it was only used three: GT1, GT2 and GT3e. GT3 and GT4e for several years was not used, but is still present in the technical documentation. “e” after the numerical index points to the use in this eDRAM, or “cache 4th level.” It is a kind of dynamic memory is more expensive than the usual memory types. The presence of this cache speeds up the graphics subsystem, and does not affect the volume of captured iGPU shared memory.
GT1 with 12 execution units, used in processors for computers which do not suffer excessive demands on the chart. As usual, in the 8th generation to this level is treated just one embedded GPU, the Intel Graphics UHD 610, the performance in GFLOPS 201,6. For Intel HD Graphics 610 Kaby Lake it was up to 172,8 211,2 GFLOPS, depending on the clock frequency. Can office desktops to do without support HDCP 2.2?
In GT2 it was already 24 execution units (or 23 in some versions), here on graphics performance varies a lot. Alone graphics processors at this level are used in ultrathin laptops, and cope with their tasks. In “lakes of the eighth generation” was used four models of graphics processors, subject to modification with a lower number of executable blocks of five models. Intel Graphics is UHD 615, 617 (such index in Kaby Lake was not), 620 and 630.
Intel UHD-level Graphics GT2
The performance of Intel Graphics UHD 615, depending on the clock frequency was varied in the range from 345,6 403,2 and GFLOPS. Exactly like the Intel HD Graphics 615. Intel UHD Graphics 617 performance was 403,2 – in Kaby Lake counterpart was not. Intel UHD Graphics 620, most running the GPU at Kaby Lake R, performance from 422,4 to 441,2 GLOPS – in Intel HD Graphics 620 it was to 384,0 403,2 GFLOPS. It could not go unnoticed – he did.
At Coffee Lake was used by the graphics processor Intel Graphics UHD 630, in Macs it was used in the complexes of discrete and integrated GPU. Produced in two versions, with 23 and 24 Executive units. The performance of the first of them from 404,8 to 423,6 GFLOPS, the second – from 403,2 up to 480 GFLOPS. The performance of Intel HD Graphics 630 to Kaby Lake, from 384,0 to 441,6 GFLOPS. Even the weakened version is better than the previous generation.
Plus Intel Iris Graphics, level GT3e
In the eighth generation of the most significant changes occurred in the most powerful and independent models iGPU, which have kept their names (Iris Plus Graphics) that changed numeric indexes and increased the size of dynamic memory (eDRAM) c 64 Gigabytes to 128. New – is well forgotten old. In Broadwell, in 2014, in the eDRAM GT3e is already set to 128 Gigabytes, then reduced this amount and here – again 128 GB. These changes were caused, and how it made sense – does not matter to me. In all variants GT3e with 48 execution units. You had a computer with this chip? Share in our Telegram chat.
Item GPUs level GT3e lakes eighth generation had two models, Plus the Intel Iris Graphics and Intel Iris 645 Graphics Plus 655. About the first of them can tell only that it was designed for lightweight laptops. Data about the performance I have. Intel Iris Graphics Plus 655, which was used in the 13-inch MacBook Pro 2018, easily coped with its graphical tasks, allowing to connect your computer to an external display with 4K resolution – but first of integrated graphics from Intel have long been allowed to do so, and secondly in the name of this GPU was not UHD.
The performance of Intel Iris Graphics Plus 655 from 806,4 up to 921,6 GHLOPS, depending on the clock frequency. The ancestor of this GPU can be considered Intel Iris Graphics Plus 650, with a capacity of 806,4 to 883,2 GFLOPS.
To be continued