• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Stats and benchmark-performance vs. usability

Mushu

Junior Member
Hello everyone!

Over the past few weeks I've found myself staring at hundreds if not thousands of stats and benchmark-results for hardware, and for the most part I've been unable to decipher how just how they relate to real world usability.



So here's what I'm wondering:

What hardware stats and benchmarks do I need to look at to figure out a sweet-spot for usability? Eg. when looking at clock-speeds, at what frequency would I stop getting meaningful improvements to my user-experience if my usage mostly consists of multitasking on a windows system with tons of tabs open in firefox (some with flash), a couple of documents open in microsoft office, some music playing on winamp or spotify, maybe an MSN window or two... you get the picture.



Apart from trial and error, is there some way to get a good indication of where I'd need to be hardware-wise to get a pleasant experience with all my programs responsive?
 
Heya,

Almost all the benchmarks you're looking at are essentially synthetic. They produce a number which you can compare to other setups, but ultimately, it doesn't tell you a whole lot about a system. It just shows a number compared to another. And you simply assume that one is better than the other. But if you were to use both systems, you may not notice that difference what so ever. And that's because the benchmarks are showing specific synthetic tasks, where as when you use it, you're random and so the system will perform extremely differently.

For example, one of my favorite bencharmks are hard drive benchmarks. People show their RAID0, short stroked RAID0, SSD, etc, benchmarks of maximum and average sustained sequential reads/writes. And it looks impressive to see 400Mbps, 600Mbps, or even over 1Gbps speeds from some top setups. But what does that mean for you as an actual user? Nothing at all. Why? Because very few people routinely are throwing file sizes that huge, in sequence, to the drive. The only ones even coming close to that are people who are rendering or compiling video who routinely dump large files to the drive. Everyone else, will never notice if they burst copied a file at 100Mbps or 1Gbps at all. Simply because few files and few moments are you actually copying that huge of a file, that often, needing that kind of speed to keep you from noticing.

Games, OS's, and workstation applications for example do a lot of file access. They access files randomly, not sequentially. They access lots of scattered small file sizes. This is taxing. This makes the system feel sluggish compared to snappy. This is where a 800Mbps+ short stroked RAID0 array HDD setup seem like a completely joke when it performs horribly in random small file read/write tests. Random smile file read/write more accurately will give you an idea of what you as a person can expect since that reflects more of what you will be doing. And this is where a single SSD is hundreds to even a few thousand times faster than even the best short stroked RAID0 arrays. This is why an SSD is going to feel different during boot, during normal OS use, during games, during application use, etc. It can seek and read/write small files so much faster than even the super fast multi-drive short stroked RAID0 arrays. It's pathetic actually.

So when looking at drive benchmarks, only pay attention to random small file read/write (like 4Kb to 1Mb). Don't look at sustained sequential read/write of large files. That shows you nothing about your user experience. It just shows maximum throughput in ideal conditions--which never happens on your computer while you use it (except when you run that particular benchmark, a joke).

This kind of mentality can be applied to other benchmarks as well. Most of them are shooting for big numbers of some kind so that they can seem like their hardware is going to be better in some way.

If you want to see more response from your hardware for heavy application use, there's a few things you can do that there's just not a lot of benchmarks for:

1. Quadcore with L3 cache, like the i5 or i7. It does make a difference.

2. Lots of RAM, the more the merrier, but only if you disable page/swap filing in your Windows OS. If you leave page/swap file enabled (virtual memory) having more than 4Gb of RAM is essentially a waste of your time, your money and your HDD space (yes, HDD space). Disable Virtual Memory completely, and have a large sum of RAM like 8Gb to 12Gb, and you will see faster response times. Why? Because nothing is being dumped to a slow HDD for virtual memory and it's all forced into RAM, which is ridiculously faster. You need a lot of RAM though to be safe to do this because if you load more than the RAM has capacity, it will dump and crash whatever you're doing. So 8~12Gb is a safe bet. More is fine of course.

3. SSD instead of HDD. It nabs random small files faster than HDD can. It will improve loading times, seek times, everything. Even thumbnails load faster in explorer view.

4. Power management. Turn it off basically. You do not want your computer to power down when it goes idle, if you want to feel like your system is responsive and snappy. Specifically, turn off it's power-down feature of your HDD. This is a big one if you're using HDD. It's less of an issue for SSD since it doesn't spin down (it doesn't even have moving parts). But if you just have normal HDD, and you want to keep your system responsive, don't let it spin down at idle. Keep it spinning at all times. And if you're worried about that, appreciate that more wear/tear (harm) happens to a disk when it spins up. It's actually better long term for drives to simply remain spinning constantly (provided they have adequate cooling so they don't turn into an oven). Turn off Quiet & Cool, or any other CPU `down clocking' feature that will drop your clock when you're idling. You want it to be full steam ahead at all times, if you want responsiveness that you can feel.

Very best, 🙂
 
Back
Top