AMD: Is SYSmark a Reliable Benchmark?

[Brian]

We all love drama between technology companies and the latest deliciousness comes courtesy of AMD. They have a 4 minute video on YouTube that questions whether or not SYSmark is a reliable, objective benchmark to use when evaluating performance between different systems. 

 

AMD's video opens with John Hampton, Director Computer Client Products, talking about how it is very important in choosing the right benchmark when evaluating a purchase because if you choose the wrong one , you can end up overpaying for technology or getting a lesser performing PC. To emphasize his point, he indirectly mentions Volkswagen's diesel debacle and uses the latter case as an example of how even established organizations can be misleading. 

 

We are then introduced to Tony Salinas, an AMD Engineering Manager, who runs through a couple tests in SYSmark using two notebooks, one that has an Intel i5 CPU and the other with an AMD FX CPU. SYSmark returns a score of 987 for the Intel system and 659 for the AMD one which is a delta of 50%. AMD notes that this is misleading and not reflective of real life performance. To prove this, they run a different benchmark called PCMark 8 on both systems and the results show the Intel i5 system scoring 4199 and the AMD equipped one 3908 which is a delta of 7%. AMD says this is a far more realistic indicator of true performance as PCMark 8 tests the CPU, GPU and video subcomponents rather than emphasizing the CPU like SYSmark does.

 

And just to drive the point home some more, AMD created their own in-house test using Microsoft Office Suite where they run a custom script on both notebooks and then measure the start and end of a task with a time stamp to gauge performance. Their in-house test shows the Intel system completing the task in approximately 61 seconds while the AMD based one finished it in roughly 64 seconds which is a delta of 6-7%.

 

As AMD's results line up with those found with PCMark 8, they contend that PCMark 8 is a more fair and balanced indicator of overall performance while SYSmark is not and therefore should be discarded by any potential customers as a tool for evaluating a purchase. 

 

Intel system using AMD in-house test: Test completed in ~61 seconds.

AMD system using AMD in-house test: Test completed in ~64 seconds.

 

They also take the time to remind us that the FTC has required SYSmark published benchmarks to contain fine print that notes their benchmark may only be optimized for Intel. Of course, this may have something to do with the fact that AMD resigned from BAPco consortium. Finally, AMD notes that a transparent benchmark should meet very specific criteria which includes being realistic, unbiased, objective and transparent, all of which SYSmark is not apparently. 

 

 

While it's fun to see these types of videos from AMD, they also highlight the fact that the company is facing an uphill battle on all fronts, whether it's CPU or GPU, and 2016 may be the year that makes or breaks them.

 

View full article

[Alex450]

After downloading the latest AMD driver I got an MS office error called Microsoft office error 30034-4

[joshthornton]

Sysmark has never been good really. Cinebench is the go-to for modern cpu testing.