Log In | Register |

Thermal Imaging Benchmark

Posted Jan. 20, 2014 by philip

We recently took one of our builds and ran it through its paces. This was no ordinary benchmark run, however. We used a thermal imaging camera to record the system - showing the critical areas that need heat dissipation. We captured the video output using a frame capture device and combined the resulting images with system data such as temperatures, utilization, and system power consumption. Here is the result:

This is a format we're still experimenting with, so we'd love to get your feedback on this before we spend more time refining our tooling and benchmark infrastructure. Would you like to see more videos and content like this? Let us know what you think!

Breaking things down further, we plot out all the measurements over time:

System Power Consumption

FPS

CPU Usage

CPU Core Temperatures

GPU Usage

GPU Core Temperatures

GPU Fan Tachometer

GPU Memory Usage

System Memory Usage

Comments
Log in to rate comments or to post a comment.
Sorted by:
 
 
RyneSmith [7 Builds] 13 points 10 months ago

I really liked the format of this video however bench marking isn't something that I'm entirely focused on when building a computer, however I know there are those that are benchmark fanatics.

I think this is a great start to a format however I feel that videos like these should be geared a bit more towards game performance rather than benchmarks as most people on here worry about how well their system performs in game like scenarios.

Just my thoughts! Thanks for the video!

 
 
LeMonarque [3 Builds] 5 points 10 months ago

There's also the varying maximum thermal junction (TJ Max) or max temperature of different processing units. Temperatures aren't directly comparable across different units of hardware in the same category (4770K vs. FX-8350 temps, GTX 780 temps vs. R9 290 temps) because sometimes they're just designed with different max operating temperatures.

So to make thermal benchmarking a meaningful thing, the results should be given in relation to the TJ Max of the individual processor. Percentage of TJ Max or something like that (distance from TJ Max can be misleading when comparing one part with a very low TJ Max against another with a very high TJ Max).

Ambient temperature should also be recorded and used as either a result modifier or a simple filter ("Show me all %-TJ Max results at X ambient temp").

I think if it's done really thoroughly, thermal benchmarking would be an awesome addition to the site. No review site that I've seen does extensive thermal imaging, and those that do simple tests haven't really done them pre-290X. Something so unique will definitely drive more traffic to PCPartPicker! This is a great start!

EDIT: Something I didn't realize, the video shows perfectly why you don't need a 1200/1300/1500W PSU for 3 cards (bar the 290 or 290X). A kilowatt is clearly all you need! It's hard to get that info through to people sometimes though. Thanks PCPP!

 
 
skemble [17 Builds] 9 points 10 months ago

This is awesome! Not every day you get to see something like this. Keep up the good work PCPP!

 
 
DenverPCPP 8 points 10 months ago

Awesome Information! This is what system designers & builders need to see for cooling needs & plans.

Regardless of it being benchmarking or a gaming pc in action this is very informative and fun to watch.

Was the test done with a side panel in place?

How about general cooling tests/ examples going forward? Like:

  • How do new chassis sizes, layout, fan placement, and air flow designs effect cooling & temp's?
  • How do low, middle, and high price systems (CLC & AC) compare in the cooling department?
  • How do CLC vs. AC systems compare?
  • Do high-powered intake fans actually improve cooling or create eddies and hot-spots with-in a chassis.
  • How do the trendy new ultra small/ tight/ packed systems fair with temps?
  • Is there a sweet spot in fan cooling?
    - air volume per minute? Liquid volume per minute? - speeds? (500rpm vs 1000rpm vs 1500rpm vs etc.) - input vs. output fans?
    - # of fans & locations?
    - better to have 2 output or 2 input? - What about combinations of input and output?
    - balance - how much output needed for input fan amount?
    - fan location?
  • How do custom built & purchased vs. enthusiast built compare?
  • What is the real effect on performance from cooling? If can lower parts 30degrees during test, will it have an effect? 60? What if can cool quickly during any slight slowdown?
  • How quickly do the parts cool?

So many possibilities. Exciting possibilities. I really look forward to future tests and reviews. Most of the best information will probably come from non-benchmark tests.

 
 
JohnTsams [1 Build] 6 points 10 months ago

Awesome. You should show the system cool off for a few seconds after the test ends. I'm curious about cooling after the benchmark is completed... most people would probably find this boring though.

 
 
TheAtomicAss 5 points 10 months ago

I'm digging it. Heat is one of my biggest pet peeves in high-performance builds.

Maybe we could expect more CPU and GPU water-cooling options than just the AIO CPU coolers here on PPP? wink wink, nudge nudge... :)

 
 
manirelli [Staff] [2 Builds] 7 points 10 months ago

It is coming. We had to put it on hold a bit to shore up some other features and put in place some new features for the compatibility filter but I promise it is coming.

 
 
jmccarble [4 Builds] 3 points 10 months ago

Hey guys,

This is sweet! What's the name of the system benchmark you guys are running on the left hand side of the scree?

 
 
manirelli [Staff] [2 Builds] 2 points 10 months ago

The information on the left side is our custom display of system readings. The benchmark we are running is Unigine Valley

 
 
rankupgamers [1 Build] 2 points 10 months ago

Moar! Do some compact ITX builds. I think this could be an awesome tool!

 
 
TheShaz 2 points 10 months ago

Very interesting.

I agree with RyneSmith. Instead of Benchmarks, doing this playing different games with all the eye candy turned on.

A photograph of the system and its components before or after where the FLIR video is.

With this we can really test cooling for CPU and GPU setups. Use this for cases and fans to show proper and improper airflow.....yes many opportunities.

 
 
Skandranonsg [8 Builds] 1 point 10 months ago

The reason benchmarks are done is to keep consistence between tests. Let's say he's playing Bioshock Infinite and in one run he mows everything down, but in the second run three Crows spawn and create a ton of effects. No way to keep that consistent.

 
 
Wataweasle 2 points 10 months ago

First of all,,, Very Nice Job.

Next, I would like to get a glimpse of the NON-Flir view for just a second in the same view/window where the flir is... Why?? If I owned this mobo I might see a hot spot (say an exposed mosfet or something) then I could forward to the "regular sight view" and identify it, and maybe decide to put a heatsink on it or something. I think they call this actionable intelligence.

Again, great to see the effort. Nice visual impact.

 
 
Skandranonsg [8 Builds] 2 points 10 months ago

I think this would be a great way to benchmark chassis, fans, coolers, etc., but poor for the chips for the reasons LeMonarque described.

 
 
unclesalty711 2 points 10 months ago

Kickass video. Now do it again, but with liquid cooling :P Push that comp to the limit

 
 
LukePRTR [1 Build] 1 point 10 months ago

NICCEE!

 
 
fnapoli [1 Build] 1 point 10 months ago

The FLIR in particular is very cool!

 
 
rakenrol14 1 point 10 months ago

EPIC... just... awesome! thanks pcpp team!

 
 
Marcus_CA 1 point 10 months ago

Innovative to say the least.

 
 
30wololo 1 point 10 months ago

Fun idea! Might be smart to place the camera a fair bit closer- they can show more detail than just vague blobs! If your model allows custom temp ranges, it might be worth setting it to 40-85 or something too, as bits cooler than that aren't hugely interesting/relevant.

 
 
joejccva71 1 point 10 months ago

Hey guys, do you mind running another test? Can you overclock your CPU to 4.5ghz and overclock the GPU as well and then running another one of these benchmarks? I'd like to see the load CPU and GPU temps while OC'd. :) Pretty please?

 
 
Trejin1 1 point 10 months ago

That was awesome! keep up the good work Pcpartpicker team!!!

 
 
oajarmeh95 [1 Build] 1 point 9 months ago

where can i download this benchmark ?

 
 
philip [Staff,Submitter] [9 Builds] 2 points 9 months ago
 
 
oajarmeh95 [1 Build] 1 point 9 months ago

thanks i already knew from your replay on Youtube i am the same guy who asked you, sorry about that it was a mistake :)

[comment deleted]
 
 
infinigun 1 point 9 months ago

What is the name of the music? It's awesome.

 
 
philip [Staff,Submitter] [9 Builds] 1 point 9 months ago

That's actually the music from the benchmark itself!

 
 
infinigun 1 point 9 months ago

Oh! did't know that. i have never benchmarked.

 
 
msmolt 1 point 9 months ago

So were can we download this new program ? or is it done yet?

 
 
philip [Staff,Submitter] [9 Builds] 1 point 9 months ago

You can find the benchmark we ran here: http://unigine.com/products/valley/

 
 
kark9 1 point 9 months ago

thermal camera benching at that scale has very little application and its more of a novelty thing in my opinion. most of the data could have been collected by othr means. the thermal camera is fairly useful for checking cooling performance of a gpu cooler or mobo chips. for example you can see where the mobo cant dissipate its heat. or compare various gpu brands (e.g. msi vs asus) coolers by capturing thermal reading for vrm's, mem-chips and other voltage circuitry components. removing the back plate and fan shroud should help with getting data althu it still wouldn't be easy.

 
 
mattataki [1 Build] 1 point 9 months ago

Great work on that one, I am from the ones that likes benchmarks. I think this type of benchmark is a good initiative and it was well done. I like the fact that it's very visual. I would see it useful to benchmark some smaller form factor cases. I noticed that the GPU's temps propagated very quickly. You'd come to expect that from 3 Titans in SLI. One thing though, the SLI bridge made it seem cooler where it sat on. Is it fact or is it acting as a shield. Then also, I noticed that under the water block, near the end, we could capture a temp rise. A little green blob surfacing out. This means that The GPUs affected the CPU temps, especially near the end. Slightly but it had an impact. Of course, we all knew that all components in a case will heat other parts and some will steal the slightly cooler air by sucking it with their fan(s). Now, in small form factor cases, it is logical that the parts are more cramped together and thus increasing the global temps of all the parts. There are more and more ITX form factor gaming cases that will handle a titan and water cooling. And I hear more and more that the mATX is rising up to the new gaming standard. We also heard a lot from Steam machines and people building their own similar ones. That is where I would see a big interest in thermal imaging.

Though, I think the main motivation of this initiative is still good and you should keep up with it. You already have the equipment and you kicked the project pretty nicely. Keep up with the good work!

Thanks for sharing the project.

EDIT: Will this method only be used with the Unigine Valley benchmark? And also, will you add more parts to the benchmarks? In 2012, you released the benchmarks on the website but it seemed to be less of a thing and the list of parts feels quite little.

 
 
overlandpark6me 1 point 9 months ago

Enjoyed watching that. well done.

 
 
Cpufreak101 1 point 9 months ago

I do like the science behind this on how it can be used for research but to me it is just... Relaxing. It calms my nerves, seeing the slow transition from slow to hot, helps patience, and also the unigine valley video in General is also relaxing, it is a perfect simulation of nature and nature is proven to help calm down and relax.

 
 
L7Mcmacdaddy 1 point 9 months ago

I watched this, then looked over my shoulder at my gtx 780 3gb graphics card and I thought "Wow you look like crap now don't you?"

[comment deleted]
 
 
Gooberdad [2 Builds] 1 point 8 months ago

That was cool. I especially like the information given afterwards in the charts. I thought three of those cards, and all the other components would pull much more power. It was fascinating to see how hot those badboys get. Thanks Phillip and the PcPartpicker Team.

 
 
kevinez 1 point 8 months ago

Nice test format. Really enjoyed the real time format of the video combined with the charting below. The video really gives a great understanding of the test and the adding the thermal camera shows areas that may need work.

Recent Posts

November 2014

October 2014

September 2014

August 2014

July 2014

June 2014

May 2014

April 2014

March 2014

February 2014

January 2014

December 2013

November 2013

October 2013

August 2013

July 2013

June 2013

May 2013

April 2013

March 2013

January 2013

December 2012

November 2012

October 2012

September 2012

August 2012

July 2012

April 2012

March 2012