Board index FlightGear Support Hardware

Iris GPUs

Joysticks, pedals, monitors.

Iris GPUs

Postby legoboyvdlp » Thu Jul 29, 2021 2:30 pm

I'm wondering if anybody has tried the new Iris GPU's yet. They claim to be faster than Nvidia's entry level mobile chips (I have a 920MX right now). However, they are still integrated chips so no dedicated VRAM; just that they have really fast processors. So I'm really curious if anyone has tried them out in FlightGear yet. Of course with the 11th gen Intel CPUs as well.
User avatar
legoboyvdlp
 
Posts: 7981
Joined: Sat Jul 26, 2014 2:28 am
Location: Northern Ireland
Callsign: G-LEGO
Version: next
OS: Windows 10 HP

Re: Iris GPUs

Postby The epic chicken » Fri Jul 30, 2021 3:42 pm

I don't think many people have them
In Thrust We Trust
User avatar
The epic chicken
 
Posts: 283
Joined: Wed Oct 28, 2020 11:24 am
Callsign: cva1920
Version: 2020.3.9
OS: windows 10

Re: Iris GPUs

Postby vnts » Sat Jul 31, 2021 10:02 am

Firstly, for anyone reading this in future actually considering buying a hardware/GPU of some sort, always check through the benchmarks like https://benchmarks.ul.com/compare/best-gpus , to see which of the various GPU/CPU options available for the laptops/systems are actually the best performing.

legoboyvdlp wrote in Thu Jul 29, 2021 2:30 pm:I'm wondering if anybody has tried the new Iris GPU's yet. They claim to be faster than Nvidia's entry level mobile chips (I have a 920MX right now). However, they are still integrated chips so no dedicated VRAM; just that they have really fast processors. So I'm really curious if anyone has tried them out in FlightGear yet. Of course with the 11th gen Intel CPUs as well.


The Iris integrated GPUs are Intels answer to AMD's APUs. According to google Intel has a low/high power Xe-LP/Xe-HP , as well as the discrete DG1 & DG2.

From google:
https://www.anandtech.com/show/16450/intel-iris-xe-video-cards-now-shipping-to-oems-dg1-lands-in-desktops wrote:As a quick refresher, the DG1 GPU is based on the same Xe-LP graphics architecture as Tiger Lake’s integrated GPU. In fact, in broad terms the DG1 can be thought of as a nearly 1-to-1 discrete version of that iGPU, containing the same 96 EUs and 128-bit LPDDR4X memory interface as Tiger Lake itself.


Looking at bunchmarks.ul https://benchmarks.ul.com/compare/best-gpus the Intel integrated Iris Xe Graphics G7 96EU has a rating of 1288 and the Intel Iris Xe Graphics G7 80EU has a rating of 1248. The DG1 is slower than even the 1030 apparently https://www.tomshardware.com/features/i ... enchmarked

(There are also Intel Xe-HPC for data centers etc. but these aren't what most people would use? https://en.wikipedia.org/wiki/Intel_Xe )

The roughly high-ish settings profile for FG (not counting the new OSM2City) needs a score of 3000 or faster (e.g. a 960 or 1050 Ti - a midrange card from from 5+ years ago) : https://wiki.flightgear.org/Hardware_re ... 2018.3_LTS

The Intel DG2 is somewhere around a 1050 which is lower than 3000 according to google, so pretty old : https://hexus.net/tech/news/graphics/14 ... territory/ . By comparison the current gen 3050 Ti is has a rating of 5356. It's probably aimed at the xx20 toxx40 low power GPUs.

The context for Intel's GPUs, aside from the competition by AMD, is that Intel were looking at the GPU market for HPC (e.g. scientific computation) a while back, and as a result have been investing in GPU development https://www.anandtech.com/show/12017/in ... -architect . This is their first offering, so it'll take a while to see how far Intel will go. Intel mainly just couldn't be bothered with GPUs much before.

The DG2 isn't really the answer to Apple's M1 made by ARM. The M1 is based on a similar design to AMD's consoles with memory on the same chip as the CPU and shared RAM/VRAM - called system-on-a-chip/SoC - but unlike consoles with the equivalent of a RTX 2080 GPU, the M1 GPU only has around the same performance as a mid-range GTX 1050 Ti from 5 years ago. Intel and AMD will respond with a new combined CPU/GPU/RAM in future. NVIDIA actually bought ARM recently , so they are likely looking at SoC solutions similar to M1/consoles. So 4 companies are in a position to create CPU/GPU on the same chip with shared RAM - bringing RAM on the chip really helps memory latency and performance, but it makes contributions less open for separate companies to competitively manufacture/design GPU or RAM.

Since Apple set the M1's GPU standard pretty low, the lack of strong competition is probably going to contribute to the response from Intel/AMD and even NVIDIA in future being muted (?). Newest consoles are putting some indirect competition on NVIDIA/Intel PC offerings as AMD have a console monopoly, but this consoles generation will probably get outdated and fall behind. The lack of competition will just mean the all-in-one systems people will have in future would be slower compared to if competition is stronger, which is not good from a 'giving people a better FG' perspective. The 4 potential competitors are a lot better than the monopolies on the best CPUs and GPUs a while back though.

Kind regards
vnts
 
Posts: 409
Joined: Thu Apr 02, 2015 1:29 am


Return to Hardware

Who is online

Users browsing this forum: No registered users and 5 guests