ALIVE @ Samford University

ALIVE @ Samford University

Brian Toone, Assistant Professor

ALIVE @ Samford University RSS Feed
 
 
 
 

Posts tagged benchmark

Overclocking update

One thing I have learned over the last week is that overclocking is challenging and somewhat addictive. I thought it was a simple tweak of the BIOS settings, but soon you find yourself tweaking the settings to try to get every last MHz out of your processor at the lowest possible voltage settings to avoid frying the motherboard and/or processor. “Frying” is a rather technical term referring to one of two things:

  1. Overheating - raising the processor core temp above its thermal specification rating for extended periods of time
  2. Overvolting - operating the processor at a voltage exceeding the maximum end of the VID Voltage Range

The other potentially bad thing that can happen with overclocking your processor is that your processor can stop operating correctly even while still being able to boot and load an entire operating system. This can trick you into thinking that your overclocked setup is working correctly when it really isn’t. A typical CPU can execute literally billions of instructions per second so if one or two or even twenty of those instructions executes incorrectly, the result may not even be noticable to a typical operating system or typical software applications that you run.

My architecture students should be able to tell you what happens if the clock rate is too high — signals do not have enough time to propagate to their final destination so an incorrect value will be written into memory or into a register before the newly calculated correct value arrives there.

That is why overclockers know that you can’t trust your setup until you run a few good benchmarks and at least one torture test. Two common torture tests that you can download and run for free are prime95 and Intel Burn Test.

I downloaded and installed prime95 and started to run it on what I thought was a stable 3.6 GHz overclock. It lasted less than a minute before crashing with a hardware failure. The way prime95 works is to perform a series of calculations and compare the results with known correct answers that are part of the data distributed with the program. If the calculated answer doesn’t match the known correct answer, then the only explanation is faulty hardware. Also, my CPU core temp started to approach the Intel thermal specification of 71deg celsius very quickly after raising the core voltage to increase stability. This is why overclockers need intricate cooling setups!

So I resolved to scale back my overclocked setup to 3.4 GHz, 800MHz memory speed, which is still a 20% overclock from the stock 2.83Ghz setting. With a core voltage setting of 1.26V, prime95 was able to run for 1 hour before I stopped it with temps just making it to 65deg celsius. I am using a cheap aftermarket air cooler so I am hesitant to let it run for longer without manually keeping an eye on the temp and shutting it down if it makes it up above 71deg celsius.

Prime95 run - just over 5.5 hours with a CPU temp warning
As you can see in the screenshot, Prime95 has made it just over 5.5 hours without error after I left it running overnight. The ASUS software PC Probe II has popped up the temperature warning. By stealing CPU cycles from Prime95 and then idling the CPU during the ensuing context switch, the temperature warning also acts a very minor temperature governor.

Computer upgrade: ATI HD 4670 graphics card

The last component of my computer upgrade arrived today in a surprisingly small box. I purchased an ATI HD 4670 graphics card based on the Tom’s hardware recommendation for entry level graphics cards. I didn’t have a lot of money to spend on this upgrade, but yet I wanted a card that would allow me to harness some of the parallel processing power via GPGPU programming. The HD 4670 has 320 streaming processors, which I am hoping to experiment with to increase the map rendering speed for topocreator.com.

After installing the card, I updated my Windows Vista Experience Index and it went from 3.3 up to a whopping 5.8 (5.9 is the maximum possible score). Note that my Primary hard disk - (2) 500 gb seagate drives in RAID 0 configuration - is rated at 5.9. This is significant for my PC Mark Vantage experiment further down in this post.


Windows Vista Experience Index - 5.9 is the highest score possible

Next, I re-ran PC Mark Vantage and was hoping for a figure somewhere in the 8000’s. Instead, I saw an increase from 6144 to 6655 - much lower than expected. Further investigation revealed that tests involving my HDD (hard disk drive) performance had very low scores. I discovered that Windows search was indexing my drive in the background, so I disabled Windows search and re-ran the benchmark. This saw a further increase to 6934 with slighly improved HDD performance. See screenshots below:

More »