Firestrike inconsistency

[MrG13]

So I have recently upgraded to a decent PC package and was given a copy of 3dmark free with the graphics card. Decided to give it a whirl and am getting some very strange results. Main thing is how inconsistent the whole things is. I have used a few different overclocks using the Asus hero MOBO to see if I can change the results much but as you can see from the results below it seems to perform better at a lower clock on the cpu than a higher clock...which I would never have suspected?! 

so i started by putting the MOBO onto TPU 1 and leaving it for a while, then the ram didnt seem to be hitting the 3200mhz it was supposed to so I read you have to enable XMP so I did that and got some better results, the MOBO also has a software program that does a 5 stage OC and I ran that which I think gave me the highest score but can't recall, either way I tried it again and couldn't get above 4700 score on ultra. 

 

The fans were either maxed out on full for the test or was I ran the built in fan software on the MOBO and left it on normal, although I didn't see any thermal throttling in the tests when the MOBO 5 stage OC was running it did hit 4800MHz and then have to restart as it hit about 80 celcius and then tuned down core 0 and 1 to 4,700Mhz and core 2 and 3 to 4,600MHz

 

Anyway here are some links to my results and a quick summation: 

 

XMP over clock on MOBO bios (most recent)

CPU: 4,698MHz

Score: 4655

http://www.3dmark.com/fs/7996239

 

Default Bios

CPU: 4,300Mhz

Score: 4753

http://www.3dmark.com/fs/7989639

 

can't remember which OC this was but I think it was just the Asus software 5 stage OC

CPU: 4,697Mhz

Score: 4776

http://www.3dmark.com/fs/7989064

 

a lot is recorded on my 3d mark profile and the scores are for firestrike ultra for the links above, I am using a asetek AIO water cooler for the CPU and then a Inno3D 980TI Hybrid for my graphics card, haven't attempted to OC that as with the CPU I don't want to be possibly causing any other problems...

 

Now I wont lie, I am a complete noob with OC and this high performance tech, coming from a intel E5300 and a 650Ti with 4GB of ram, this is a massive step up in performance, which is great! 

CPU is a 6700k (and its comfortably running at a 4,700MHz overclock with perfect stability), no thermal issues with the water coolers attached with temps not going above 60C in any benchmarks. 

 

To summarise is this normal, should I be running these benchmarks this much on new hardware and then finally could it be something wrong with the new hardware?!

One thing that may be worth mentioning, with the CPU cooler it came with some normal standard thermal paste so I bought some artic silver and then reapplied that which since I have had higher scores but in real world, I haven't noticed any raise in temps or better performance (apart from the benchmark scores. 

 

If you need more details let me know!

 

(included a pic of the new system all set up because it looks shiny and green and to thank you for taking the time to read all of this lol)

[MrG13]

*Bump*

 

And combined comparison link on 3dmark site...

 

http://www.3dmark.com/compare/fs/7989064/fs/7989639/fs/7996239

[Equlizer]

I may be off here but from what i've noticed overclocking,  if you go to high, your score gets lower (unless you are on water or ln2)

[Michamus]

If I understand correctly, you are overclocking your CPU in an effort to improve your 3DMark GPU score? OCing your CPU is only really going to effect CPU benchmarks.

[MrG13]

Equlizer

I am running a water block on the CPU and the graphics card is a hybrid, (i.e. water for GPU and I think also the VRAM and rest is cooled by the fan)

Michamus

3Dmark also includes physics test so uses the CPU and if you look at the hall of fame all the high hitters have an extremely high cpu clock. My OC was only an auto OC using the built in software from the MOBO. 

 

I have been talking to a few people (on Lukes stream this evening) though and its all seems in the margin of error for firestrike to give different results (sometimes up to 500 points different), even if you don't change anything so I am starting to think the CPU had nothing to do with it and maybe its just the software being itself but would need to see more people with the same thing before I could call it conclusive...or get info from the pro's hence the call for help  

[Michamus]

36 minutes ago, MrG13 said:

margin of error for firestrike

That's actually where I was going to go next. As for those Hall of Famers, they're overclocking so as to better facilitate the 4 way SLI setups they're running.

[MrG13]

32 minutes ago, Michamus said:

That's actually where I was going to go next. As for those Hall of Famers, they're overclocking so as to better facilitate the 4 way SLI setups they're running.

The more i look into it the more it seems to be the software, which is reassuring as it means there isn't any faults with my new hard ware so win win but looking at all those high scores is making me want to sell my first born and get 4 way SLI too  but I game too much and it has too many issues at this time to be worthwhile... well at least until direct x 12 becomes more main stream and utilizes card correctly! 

[sof006]

I would ignore synthetic benchmarks... Better off running a game and checking if your FPS goes up or down or stays the same. I'm most cases if your overclock is stable your FPS will improve (slightly) or might make your FPS more stable while gaming. Synthetic benchmarks are literally only there for enthusiasts who are looking for "bragging rights". 

[MrG13]

1 hour ago, sof006 said:

I would ignore synthetic benchmarks... Better off running a game and checking if your FPS goes up or down or stays the same. I'm most cases if your overclock is stable your FPS will improve (slightly) or might make your FPS more stable while gaming. Synthetic benchmarks are literally only there for enthusiasts who are looking for

"bragging rights". 

To be honest I am currently running 2x 1080p monitors so everything I throw at it stays at a constant 60fps so until I get a new 1440 60hz plus monitor it may be difficult to notice, not that I am complaining as gaming on this is amazing! But for now it is hard to see any difference with the monitors being a bottle neck. 

 

Judging by the results I would totally agree with you, not that bragging rights is a bad thing, after all without competition everything would get a bit stagnant, and also for a good 'ballpark' figure on where a system is sitting, I guess they have their uses. 

 

I honestly had no idea the benchmarks were 'this' inaccurate though as no one ever runs 2 or more firestrike scores when benchmarking builds. So they must either do a few, and then post the highest, or they do one and then chuck that up online and not worry so much. The biggest gaps I have seen so far is about 700 which isn't really THAT much of a big difference...

 

BUT

 

One thing which does strike me though is sometimes when people are rating graphics cards, for example, you see the AMD vs Nvidia the benchmark scores are separated by a few hundred and always in Nvidia's favour.... so it wouldn't be hard to fiddle the results to make either or the top end card which is what most people pay attention too, so in reality is AMD and the bottom score for the Nvidia so the gap is even bigger, or is it more of the opposite, either way it's worth taking note of an I know I will be factoring in a minimum of 700 point swing either way when looking at tables in the future....