Tek Syndicate: Expert Conjecture and Speculation

[GOM3RPLY3R]

Hello All,

I would like to know how most of you feel about Tek Syndicate, and their benchmarks. I have compared their benchmarks to more than 10+ other sources (even from Toms), and I concluded that ~60-70% or more of their benchmarks are either Synthetic, Biased, or Falsely Contributed.

In addition, I would like to start a petition to ban the use of Tek Syndicate's benchmarks. (I would like to note that this is not fully out of rage, but just of well being. I have met many people (mostly AMD fans) that only use Tek Syndicate and maybe 1 or 2 other sources, for CPU threads and sometimes GPUs).

 

[MEC-777]

There could be other reasons why their benchmarks differ from the others. Not every test is exactly the same. There will always be variances.

Not sticking up for Tek Syndicate, just sayin.

 

[GOM3RPLY3R]

The one thing that I forgot to mention is that when they do their CPU tests for games, they always either use, a GTX 670 or a HD 7850 (from what I've seen). And then they push the settings to the max. That isn't really testing the CPU, that's testing the GPU. Doesn't make any sense. If you are going to test a CPU on a game, then you should do low settings with about 3 or more trials. If you are doing high settings then no less than 5 trials. Experiments are also based on trials. Something can't be proven right if other people try it and get difference results (which almost every other benchmark website is doing).

 

[Ranth]

The one thing that I forgot to mention is that when they do their CPU tests for games, they always either use, a GTX 670 or a HD 7850 (from what I've seen). And then they push the settings to the max. That isn't really testing the CPU, that's testing the GPU. Doesn't make any sense. If you are going to test a CPU on a game, then you should do low settings with about 3 or more trials. If you are doing high settings then no less than 5 trials. Experiments are also based on trials. Something can't be proven right if other people try it and get difference results (which almost every other benchmark website is doing).

http://www.youtube.com/watch?v=4et7kDGSRfc 00:47
And I quote:
"This is not a cpu review..." What they are doing is taking what people are using I5/8350 with 670/7870 and see what is the best cpu. As they say, are you gonna get those^ components to be playing at 800x600!? I could give a fat rat's ass about Intel being better at 800x600 if AMD is better running 1920x1080.

IT MAKES SENSE, you just have to figure out that it's about the performance of 2 systems and not 2 CPU's.

 

[juanrga]

Hello All,

I would like to know how most of you feel about Tek Syndicate, and their benchmarks. I have compared their benchmarks to more than 10+ other sources (even from Toms), and I concluded that ~60-70% or more of their benchmarks are either Synthetic, Biased, or Falsely Contributed.

In addition, I would like to start a petition to ban the use of Tek Syndicate's benchmarks. (I would like to note that this is not fully out of rage, but just of well being. I have met many people (mostly AMD fans) that only use Tek Syndicate and maybe 1 or 2 other sources, for CPU threads and sometimes GPUs).

Different does not mean wrong. Also degree of rightness is not obtained in a popularity contest: something is not always correct when it is said by more people; there are occasions when a minority is right and a majority is wrong.

The one thing that I forgot to mention is that when they do their CPU tests for games, they always either use, a GTX 670 or a HD 7850 (from what I've seen). And then they push the settings to the max. That isn't really testing the CPU, that's testing the GPU. Doesn't make any sense. If you are going to test a CPU on a game, then you should do low settings with about 3 or more trials. If you are doing high settings then no less than 5 trials. Experiments are also based on trials. Something can't be proven right if other people try it and get difference results (which almost every other benchmark website is doing).

Yes, in a CPU test you would eliminate GPU dependencies, for instance by playing at 720p, but the result is irrelevant for real gaming at 1080p or above. As Ranth has mentioned, they are worried about benchmarking the whole hardware under realistic settings rather than going purists with CPU tests and GPU tests.

I also love your attempt to go scientific, but ignoring what science says. There is nothing that justifies that making 3 trials or 5 trials you will be obtaining better results. There is not such magic numbers or rules of thumb. It is perfectly admissible that a 2 trial outcome was more accurate than a 3 trial one. Science is a bit more complex.

 

[GOM3RPLY3R]

Hello All,

I would like to know how most of you feel about Tek Syndicate, and their benchmarks. I have compared their benchmarks to more than 10+ other sources (even from Toms), and I concluded that ~60-70% or more of their benchmarks are either Synthetic, Biased, or Falsely Contributed.

In addition, I would like to start a petition to ban the use of Tek Syndicate's benchmarks. (I would like to note that this is not fully out of rage, but just of well being. I have met many people (mostly AMD fans) that only use Tek Syndicate and maybe 1 or 2 other sources, for CPU threads and sometimes GPUs).

Different does not mean wrong. Also degree of rightness is not obtained in a popularity contest: something is not always correct when it is said by more people; there are occasions when a minority is right and a majority is wrong.

The one thing that I forgot to mention is that when they do their CPU tests for games, they always either use, a GTX 670 or a HD 7850 (from what I've seen). And then they push the settings to the max. That isn't really testing the CPU, that's testing the GPU. Doesn't make any sense. If you are going to test a CPU on a game, then you should do low settings with about 3 or more trials. If you are doing high settings then no less than 5 trials. Experiments are also based on trials. Something can't be proven right if other people try it and get difference results (which almost every other benchmark website is doing).

Yes, in a CPU test you would eliminate GPU dependencies, for instance by playing at 720p, but the result is irrelevant for real gaming at 1080p or above. As Ranth has mentioned, they are worried about benchmarking the whole hardware under realistic settings rather than going purists with CPU tests and GPU tests.

I also love your attempt to go scientific, but ignoring what science says. There is nothing that justifies that making 3 trials or 5 trials you will be obtaining better results. There is not such magic numbers or rules of thumb. It is perfectly admissible that a 2 trial outcome was more accurate than a 3 trial one. Science is a bit more complex.

I'll agree with you on the trial amount. The only thing that I see people doing however, for example, 8350rocks, is/was using the CPU benchmark material and noting that it is perfectly fine, and those are the final results, when really, they only use one GPU, one driver, and use the highest settings. Running them at 1080p is probably the best resolution for today, however, when you are doing CPU based tests, Graphics are dependent on the GPU. Stressing the GPU with more work to do can directly / indirectly affect the CPU performance. Lowering the settings will put more work on the CPU (in other words, make it much easier on the GPU and have more of the FPS performance be achieved by the CPU).

I'm really just tired of people using it in the wrong manner. If we were on a subject of, "The 7970 Ghz Edition runs better on AMD CPU's than Intel," then that's different. Otherwise, with the trial number aspect aside, the best results from a benchmark (experiment) is achieved with as many variables the same as possible. Otherwise its like testing if fungi will disease Zebras faster than Giraffes, but the Zebras are in an extremely hot climate and the giraffes are in a cold. Obviously the fungi will have a better chance in the hot weather as cooler air neutralizes them. It. Just. Doesn't. Work.

EDIT: Upon posting this, an idea scratched my brain. Wouldn't it be possible to use just take the CPU mathematical codes that games use for like Physics, 3D Positioning, etc. from for example Battlefield 3, and run it in a software engine that doesn't specifically run a graphics based system, just puts axis points on a 3-D plain using numbers?

Basically running a game without any of the GPU based code, and just the CPU?

Just a thought.

 

[bobbybamf12]

All I know is my friends i5 can play dayz/planetside 2 better then mine 8350 where i struggle to stay above 30fps. So it's good to see the max fps a cpu can put out instead of how many when a gpu is holding it back.

 

[MEC-777]

Benchmarks are great for comparing individual components (GPU's vs GPU's and CPU's vs CPU's etc.) but when it comes to real-world performance, the benchmarking must be done using real-world settings and take into account the whole system. Sure, some things won't affect the outcome too much, like gaming with 8GB RAM vs 16GB RAM. (assuming the same freq/lat etc.)

Take two systems, both with comparable specs and the same mid-high-end GPU's. Only difference is one system is an 8350 and the other is a 3570k, for example. Play a particular game at 1080p at high or max settings. Some will run better on the one system and some on the other. Even running a more budget-oriented CPU may yield no discernible disadvantage with particular applications/games. That, to me and probably to many others, is just as useful as the benchmarks for the CPU's themselves.

It all has to be taken in context. I try to look for benchmarks of systems with similar specs to what I'm planning on building and look for tests of those systems running games I plan to run at the settings and resolutions I plan to run them at. If there is little to no difference in the performance of that system in that instance, whether it's using a Haswell 4670k or an Athlon x4 750k, then why would I spend $150 more if the AMD meets or exceeds my needs? (Not saying that's the case with those specific CPU's, I'm just them an example).

The question is; what will you be doing with your machine and what is your budget? That should dictate what components you should be comparing and using in your system. At that point, then you should be looking at benchmarks for the components in that price/performance bracket.

I have no beef with anyone's benchmarks, so long at you look at everything in context. You can't really state that someone's benchmarks are more or less accurate than someone else's, (unless you were there and know for a fact they did something wrong). It's not that simple.

All I know is my friends i5 can play dayz/planetside 2 better then mine 8350 where i struggle to stay above 30fps. So it's good to see the max fps a cpu can put out instead of how many when a gpu is holding it back.

Are you both running the same GPU? And what GPU is it? What mother board? What RAM/freq, lat, etc.? What settings are you both running the games at? What resolution? How many monitors? Are you guys overclocking?

You have to take into account the whole system and settings/res.

 

[juggernautxtr]

Logan always states "real world use settings'"
I find more and more synthetic test bench purists doing most of the complaining. no cpu will test exactly like another, no gpu will test exactly alike, no computer is running the exact same way as another. no board will test the same as another.
real world use can be more useful than a synthetic any day of the week.
synthetic are good for base line performance, once you hit real world use things change.