intel Iris 6200

Post requests to add new GPUs to the official whitelist here.

Moderators: Site Moderators, PandeGroup

Re: intel Iris 6200

Postby bruce » Sun Jan 01, 2017 1:49 pm

Foldy has demonstrated that CPUs with OpenCL 1.2 will fail when running FAHBench, which is why FAH does not support older CPUs.

If you have the hardware that can support Double Precision and OpenCL 2.0 has been installed, please report the results of running FAHBench. Otherwise we can close this topic.
bruce
 
Posts: 21182
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: intel Iris 6200

Postby ComputerGenie » Sun Jan 01, 2017 1:52 pm

bruce wrote:...If you have the hardware that can support Double Precision and OpenCL 2.0 has been installed, please report the results of running FAHBench...

For the 530, the result was ~ 6.8 with the i7-6700K having a result of ~2.7 as a CPU
User avatar
ComputerGenie
 
Posts: 242
Joined: Mon Dec 12, 2016 4:06 am

Re: intel Iris 6200

Postby rwh202 » Sun Jan 01, 2017 4:03 pm

bruce wrote:Foldy has demonstrated that CPUs with OpenCL 1.2 will fail when running FAHBench, which is why FAH does not support older CPUs.

If you have the hardware that can support Double Precision and OpenCL 2.0 has been installed, please report the results of running FAHBench. Otherwise we can close this topic.


Sorry, I'm losing track of this thread - are we talking CPUs or iGPUs. I thought it was about iGPUs (specifically the Intel Iris generations).

As far as I know, Foldy has only demonstrated that FAHBench failed on his 2500k (an old Sandy Bridge CPU with an old incompatible iGPU)

FAHBench runs just fine on my laptop HD5500 iGPU (I'll have to wait to try on an Iris machine) - where has this supposed need for openCL 2.0 and DP come from? It's possible that Folding could be able to run with little or no effort on the majority of recent intel iGPUs.

Yes, I only score 4.1 vs 33 for the GTX960M, but it's still double what the quad core CPU manages. That's a high end CPU with a low-mid range iGPU. Most dual core machines with better iGPU will benefit more from having access to the iGPU.
rwh202
 
Posts: 269
Joined: Mon Nov 15, 2010 8:51 pm
Location: South Coast, UK

Re: intel Iris 6200

Postby foldy » Sun Jan 01, 2017 4:46 pm

Initially the thread was about whitelist Intel Iris 6200 iGPU but currently not supported.
Now we discuss the intel iGPUs a little more wide which hardware could be supported in future and what drivers are needed and so on.

1) The requirement for double precision comes from OpenMM? Supported since 5th GPU generation, Broadwell.

2) The requirement for OpenCL 1.1 comes from OpenMM. Supported since 3rd GPU generation, Ivy Bridge.
Also OpenCL 1.2 or 2.0 is already available in intel gpu drivers for Windows/Mac OSX/Linux CentOS(maybe others)

3) There may also the CPU part be listed as OpenCL device but that is not useful, only the iGPU part as OpenCL device is relevant.

rwh202 wrote:It's possible that Folding could be able to run with little or no effort on the majority of recent intel iGPUs.

a) The problem with the iGPUs is they are too slow to finish the work units for fast GPUs in time.
So if a FahCore for Intel iGPUs (and maybe AMD iGPUs) would be made it would need some work units tailored for these iGPUs to meet the deadlines.

b) There may also be problems in general when Intel GPU driver does something different than NVidia or AMD and then work units fail.
So creating a new FahCore for Intel GPUs may need some testing and bug fixing to guarantee the results are fine.

c) And bruce wants some numbers which iGPUs could run FahBench and are listed in FAH logfile.

Is my summary correct?
Last edited by foldy on Sun Jan 01, 2017 9:17 pm, edited 7 times in total.
foldy
 
Posts: 878
Joined: Sat Dec 01, 2012 3:43 pm

Re: intel Iris 6200

Postby foldy » Sun Jan 01, 2017 4:59 pm

Release Notes of Linux GPU driver say:
Code: Select all
OpenCL™ 2.0 GPU/CPU driver package for Linux* (64-bit)

The Intel intel-opencl-r4.0 (SRB4) Linux driver package  provides access to the GPU and CPU components of these processors:

    Intel® 5th, 6th, or 7th generation Intel® Core™ processors
    Intel® Celeron® Processor J3000 Series with Intel® HD Graphics 500 (J3455, J3355), Intel® Pentium® Processor J4000 Series with Intel® HD Graphics 505 (J4205), Intel® Celeron® Processor N3000 Series with Intel® HD Graphics 500 (N3350, N3450), Intel® Pentium Processor N4000 Series with Intel® HD Graphics 505 (N4200)
    Intel® Xeon® processor v4 or v5 with Intel® Graphics Technology (if enabled by OEM in BIOS and motherboard)

https://software.intel.com/en-us/articl ... ows_driver
foldy
 
Posts: 878
Joined: Sat Dec 01, 2012 3:43 pm

Re: intel Iris 6200

Postby _r2w_ben » Mon Jan 02, 2017 5:01 pm

foldy wrote:1) The requirement for double precision comes from OpenMM? Supported since 5th GPU generation, Broadwell.

A recent project announcement mentions
rafwiewiora wrote:Mixed precision is used (GPUs without DP are excluded).
It sounds like the server assignment code differentiates between projects requiring double precision and assigns to appropriate compatible GPUs.

From the FAHBench documentation
Precision

Computers have to round numbers after a certain point (e.g. when tracking the position or velocity of an atom). single precision uses 32-bit floating point numbers (more rounding, less accurate), whereas double uses 64-bit numbers (less rounding, more accurate).

Consumer GPUs are really bad at double precision calculations (so manufacturers can sell more expensive enterprise units to researchers). Luckily, molecular dynamics really only requires single precision to be useful. Folding@Home uses single precision.
_r2w_ben
 
Posts: 115
Joined: Wed Apr 23, 2008 3:11 pm

Re: intel Iris 6200

Postby bruce » Mon Jan 02, 2017 5:40 pm

Updating with a little more detail.

For quite a few years, consumer-grade] grade GPUs[/b} featured only single precision (SP) while [b]commercial grade GPUs also provided double precision (DP). Then consumer GPUs introduced a limited (read: "slow") support for DP. Modern consumer GPUs now provide improved DP capabilities. In fact, Intel seems to be following the same pattern with their integrated GPU.

Over the same time period, the complexity of proteins needing to be analyzed by FAH has increased..FAH now needs "mixed precision" which uses mostly SP but demands at least limited DP support for a small percentage of the calculations.

As I said eariler, we might be able to talk FAH into supporting Intel iGPUs but any efforts to support OpenCL1.2 (with SP only) would be a wasted effort since very few project would be available for assignment, and they'd dry up rather soon. That's why I believe we should limitet this discussion to OpenCL 2.0 (with DP).. If we follow my recommendation, we'll need a revised estimate of potential number of new folding slots that might be provided if FAH did add support for Intel OpenCL 2.0.

By the way, CPUs provided DP support before the Pentium was introduced.
bruce
 
Posts: 21182
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: intel Iris 6200

Postby foldy » Mon Jan 02, 2017 7:06 pm

So we need to guess how many donors have a 5th gen intel CPU with iGPU or newer. I guess between 1% to 10% of CPUs.
Last edited by foldy on Tue Jan 03, 2017 11:33 am, edited 1 time in total.
foldy
 
Posts: 878
Joined: Sat Dec 01, 2012 3:43 pm

Re: intel Iris 6200

Postby ComputerGenie » Mon Jan 02, 2017 7:14 pm

foldy wrote:So we need to guess how many donors have a 5th gen intel CPU with iGPU. I guess between 1% to 10% of CPUs.

Since the 6th gen has been the #1 selling CPU for the last 5 quarters, I'd add ~5,000,000 units to whatever that number is. :egeek:
User avatar
ComputerGenie
 
Posts: 242
Joined: Mon Dec 12, 2016 4:06 am

Re: intel Iris 6200

Postby Joe_H » Mon Jan 02, 2017 7:34 pm

ComputerGenie wrote:
foldy wrote:So we need to guess how many donors have a 5th gen intel CPU with iGPU. I guess between 1% to 10% of CPUs.

Since the 6th gen has been the #1 selling CPU for the last 5 quarters, I'd add ~5,000,000 units to whatever that number is. :egeek:


In my opinion, that is a completely unrealistic approach to estimating the potential iGPU's that would be available to folding. If you added a number more reflective of actual participation rates compared to CPU's sold, it would be likely be less than 0.1% of that. In addition, a large portion of that 5,000,000 units reflect sales of lower end CPU's that have correspondingly lower performance integrated video which will be less likely to meet the computational needs of folding.
Image

iMac 2.8 i7 12 GB smp8, Mac Pro 2.8 quad 12 GB smp6
MacBook Pro 2.9 i7 8 GB smp3
Joe_H
Site Admin
 
Posts: 3760
Joined: Tue Apr 21, 2009 4:41 pm
Location: W. MA

Re: intel Iris 6200

Postby ComputerGenie » Mon Jan 02, 2017 11:07 pm

Joe_H wrote:
ComputerGenie wrote:Since the 6th gen has been the #1 selling CPU for the last 5 quarters, I'd add ~5,000,000 units to whatever that number is. :egeek:


In my opinion, that is a completely unrealistic approach to estimating the potential iGPU's that would be available to folding. If you added a number more reflective of actual participation rates compared to CPU's sold, it would be likely be less than 0.1% of that. In addition, a large portion of that 5,000,000 units reflect sales of lower end CPU's that have correspondingly lower performance integrated video which will be less likely to meet the computational needs of folding.


1) That 5 million number is due, in no small part, to the estimates I was able to find on the i7-6700K alone (which as I said benches nearly 3x higher using iGPU than CPU).
2) Even if said "lower end CPU's" only produce 2x using iGPU as CPU, it's still 2x more than they can now.
3) At the moment that you decide to limit your mind to 1/10th of 1% of possibilities, you've doomed the project to stay within the realm of 1/10th of 1% of possibilities. :cry:
4) I pray that the devs, as a whole, disagree with you and decide that there is possibility to increase the project large enough to actually do some good in the world (rather than decide that it's "too hard", "too unlikely", etc. to expand and get new people involved).

As I've repeatedly said, and I will continue to say throughout my life: "When you think narrow, you get narrow results."
User avatar
ComputerGenie
 
Posts: 242
Joined: Mon Dec 12, 2016 4:06 am

Re: intel Iris 6200

Postby foldy » Tue Jan 03, 2017 9:02 am

I also wish we reach the million donors but that is a different problem. A fahcore for the iGPUs will not reach new donors in the first place but increase the performance of the current donors. That is why we calculate how many usable iGPUs the current donors may have.
foldy
 
Posts: 878
Joined: Sat Dec 01, 2012 3:43 pm

Re: intel Iris 6200

Postby ComputerGenie » Tue Jan 03, 2017 9:15 am

foldy wrote:.... A fahcore for the iGPUs will not reach new donors in the first place but increase the performance of the current donors. That is why we calculate how many usable iGPUs the current donors may have.

I get that, I do; however, it will in the "second place" and I think that if every issue is only looked at in relation to existing donors, solutions will never be found to add new donors.
If you were new and the software didn't work on your new (or newish) iGPU, would you want to use it? Especially if you came here and read that no one cares about the new stuff because there aren't enough "existing donors/users" with it? :?
User avatar
ComputerGenie
 
Posts: 242
Joined: Mon Dec 12, 2016 4:06 am

Re: intel Iris 6200

Postby bruce » Tue Jan 03, 2017 12:08 pm

Why would anyone be telling them that "nobody cares"?

The fact is that they care very much about increasing the science that's getting done but their funds are not unlimited. Since they can't do everything, they must establish priorities about the best way to increase the science.
bruce
 
Posts: 21182
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: intel Iris 6200

Postby ComputerGenie » Tue Jan 03, 2017 3:38 pm

bruce wrote:Why would anyone be telling them that "nobody cares"?

I don't know why you keep telling them that.

bruce wrote:The fact is that they care very much about increasing the science that's getting done but their funds are not unlimited. Since they can't do everything, they must establish priorities about the best way to increase the science.

The fact is that the answer (to so many thoughts, questions, ideas, and hardware options that will factually increase donors and WUs) is repeatedly "no" based on some imagined, unknown (perhaps even unknowable), massive quantity of x dollars in cost weighed against some narrow-minded, preconception that something will not be a great benefit.

As someone that's spent 10s of thousands of dollars on Bitcoin (only to finally break even at the 3 year mark), which unarguably has less to offer the world, I refuse to take the short view of anything (except how grammatically incorrect this sentence is).

The outright dismissal of ideas (or in this case millions of hardware units) based on an imaginary "too high" cost is beyond ludicrous.
Q: How do you know the cost is "too high" when you have no clue what the cost is, what the potential to raise funds is, or what the benefit is?
A: You don't.
User avatar
ComputerGenie
 
Posts: 242
Joined: Mon Dec 12, 2016 4:06 am

PreviousNext

Return to New GPUs (whitelist)

Who is online

Users browsing this forum: No registered users and 1 guest

cron