Intel Iris Pro 580 underwhelming performance

Nihilist

Newcomer
I was searching through the forum, but couldn't find much if any discussion on the gaming performance of the Intel Iris Pro 580 as featured in the Skull Canyon NUC. My initial impression/suspicion is that the Skull Canyon NUC cooling solution is inadequate and is stifling the iGPU's true potential. Thus, when gaming is concerned, the product fails to deliver on its promise and is not a real upgrade over GB-BXi7-5775.

From what little comparative benchmarks I could find, it manages to be only 10-15% faster than that of the Iris Pro 6200 found in 5775c/5775r. I based my estimate on the comparative results from: https://goo.gl/pikQyY.

This is confirmed again by comparing the Firestrike score of 1900 vs 1770, as evidenced in:
http://goo.gl/AUrsCu and http://goo.gl/2MrKqd. The latter link also gives 3d mark 11 comparison, where the Iris Pro 580 achieves a 14% improvement over the Iris Pro 6200.

This despite it having 72 CU against 48 of the latter, and a generational improvement. I would be very curious to know whether a better cooling solution and more TDP headroom would allow the Iris Pro 580 to deliver closer to 50% more performance compared to the Iris Pro 6200.

Perhaps we might have to wait for some of the 65W i7 6785r based SFF solutions to finally have this question answered.
 
http://www.anandtech.com/show/10343/the-intel-skull-canyon-nuc6i7kyk-minipc-review/7

its such a tiny system. Look at where all the heat is centralized

It looks like there is a side vent where it draws air in
upload_2016-8-7_14-53-55.png


maybe you can point a fan in that direction ?

https://www.amazon.com/dp/B00G05A2M...=UTF8&colid=BTT4EYXK0APH&coliid=IIF11VHX7HFJI

or an 80mm one

https://www.amazon.com/AC-Infinity-..._SR160,160_&psc=1&refRID=9XG28E1SK8EK8KSEQYQG

I had 4 140mm ones in my old tv stand but I've since mounted the tv on the wall and got one of those ikea 4 cube things. I've moved 2 of them over to blowing onto my router which gets quite hot and my fios modem
 
I'd say indeed TDP is a very big reason why the improvement is small. The 5775C/R cpus were 65W, and Skull Canyon is only 45W. Probably for this reason the max gpu clock was 1.15Ghz for the 5775 but is only 950Mhz for the 6770HQ.
So that it is an improvement at all is pretty good considering the TDP difference is very significant (it's all the same manufacturing node, after all). The raw (peak) performance isn't really much higher considering the max clock difference neither.
~40W gpus (considering they don't have a cpu part) from amd or nvidia aren't exactly spectacular neither (albeit a pascal chip with 40W TDP should be faster, without looking up any numbers I'd say maxwell might be about as efficient, though of course the latter has a manufacturing disadvantage).
 
I'd say indeed TDP is a very big reason why the improvement is small. The 5775C/R cpus were 65W, and Skull Canyon is only 45W. Probably for this reason the max gpu clock was 1.15Ghz for the 5775 but is only 950Mhz for the 6770HQ.
So that it is an improvement at all is pretty good considering the TDP difference is very significant (it's all the same manufacturing node, after all). The raw (peak) performance isn't really much higher considering the max clock difference neither.
~40W gpus (considering they don't have a cpu part) from amd or nvidia aren't exactly spectacular neither (albeit a pascal chip with 40W TDP should be faster, without looking up any numbers I'd say maxwell might be about as efficient, though of course the latter has a manufacturing disadvantage).

That does seem to explain a lot. So thanks to more TDP headroom, the 6785r can clock the GPU 20% higher up to 1.15 GHz. Assuming linear speedup this should put it at about 40% above that of 5775c/r

Too bad there are no announced products with this chip yet.
 
That does seem to explain a lot. So thanks to more TDP headroom, the 6785r can clock the GPU 20% higher up to 1.15 GHz. Assuming linear speedup this should put it at about 40% above that of 5775c/r

Too bad there are no announced products with this chip yet.
I do not know if these chips actually reach their max gpu clocks in practice. But even if they don't, it seems likely the effective clock would indeed be quite a bit higher on the 6785R (and 5775C/R) than on the 6770HQ (are there tools to monitor intel gpu clocks?).
 
I am enjoying my nuc, league of legends maxed out yields more than 150fps maxed out at 1080p. Sure all dx12 games dont run very well without decently reducing image quality but it is faster than kaveri or tx1.
 
yah, the intel integrated gpus really tank when they get too hot, and it's very easy for them to get too hot in laptops and small cases. Have the same issue with my macbook pro. Had to get a program to force the fan speeds to run higher.
 
yah, the intel integrated gpus really tank when they get too hot, and it's very easy for them to get too hot in laptops and small cases. Have the same issue with my macbook pro. Had to get a program to force the fan speeds to run higher.
From what I've observed with various notebooks and tablets and such over the years, I figure throttling is a problem with most of the non-desktop APU-based machines out there.

I bet there are OEM desktops with such pathetic heatsinks attached that throttling is a problem for them too. :D
 
Last edited:
From what I've observed with various notebooks and tablets and such over the years, I figure throttling is a problem with most of the non-desktop APU-based machines out there.

I bet there are OEM desktops with such pathetic heatsinks attached that throttling is a problem for them too. :D

Yah, I'm realizing I bought a laptop that has performance that's basically impossible to use.
 
My original AMD c50 Apu would throttle within minutes of any gaming or 3d workload. Found a way to undervolt the chip, and I got reduced temps, less throttling under medium workloads and slightly prolonged battery life. Lower clocked pstates with undervolting even drove up standby battery life considerably. Is that possible with the Intel?
 
I point a 140mm fan at the back of my original surface pro , keeps it about 8 degrees cooler and it doesn't throttle as much.

Hopefully a surface 5 doesn't have that issue lol
 
low-level apis that reduce cpu utilization would be great for laptops and portables that have difficulty managing heat.

Lower heat, better performance. That could than lead to improved visuals (greater IQ settings at same performance levels). All kinds of benefits. Shame to see Intel drop it on Windows. Even more baffling since they appear to be continuing to support it on Linux and Android. We can only hope that it's just a matter of miscommunication.

Regards,
SB
 
Yeah, if it really is the case that they are dropping it on Windows, I can only imagine that they don't care about wasting the power and potential of their GPUs on Windows since it's not generally a power constrained device. And they see the benefits for it on Android where they need all the power savings and performance they can get. And Linux, well, hmmm. :p

I don't know, it's just baffling.

Then again perhaps they feel Dx12 on Windows makes Vulkan irrelevant on the platform? After all Dx12 won't be available on Linux and Android, so Vulkan becomes significantly more important on those platforms. Doom so far being the lone exception of a major Vulkan only (no Dx12) application on Windows.

Regards,
SB
 
Don't forget The Talos Principle, even though it's Vulkan support is officially still beta, and it is arguably not a major application.
 
Last edited:
Back
Top