New MSI GTX 1060 6GB Getting Lousy Fire Strike Score

[Sasqui]

I don't dare go outside in the sun for more than 30 minutes because my skin is neon white and burns way too easily.

On topic, I'm going to test my card and see if it matches other 1080TI scores because maybe it's an NVIDIA bug.

Look at the OP same system with a 1070 got near 15k

I have dumb question for OP:

Did you plugged 6 pin power cable from PSU to graphics card before turning on pc?

Lol yes

 

[crazyeyesreaper]

use a driver cleaner like DDU clean out all AMD driver crap, Reinstall Nvidia drivers do a clean driver install. Go into the NV Control panel globally set power / performance from optimal to maximum.

Been having issues like these with Nvidia GPUs in a laptop I am reviewing currently. The Drivers are just borked. Try rolling back to a slightly older driver and you should be able to verify.

 

[Laki89]

Look at the OP same system with a 1070 got near 15k

When you saying same system it also means same power supply
That leads to conclusion what you said in first post: you got bad card

Just for test purposes, i would try to power on pc with 6 pin cable unplugged to see what will happen
In most of the cases result will be what toothless said so you don't need to worry about

 

[Sasqui]

use a driver cleaner like DDU clean out all AMD driver cra

Please read the OP

When you saying same system it also means same power supply
That leads to conclusion what you said in first post: you got bad card

Just for test purposes, i would try to power on pc with 6 pin cable unplugged to see what will happen
In most of the cases result will be what toothless said so you don't need to worry about

Interesting idea. Funny, there are people here that don't think it's worth the time test a new card. (Really)

 

[Toothless]

I'm going to just admit that I'm dumb and I can't find other benchmarks for my card and that test to see if it's a NVIDIA bug or not. Hurdedur no teeth=no brain.

EDIT: Nevermind I found it, 3k less than the top scores of similar rigs. I'd go with that PCI power theory from Laki over there.

 

[jboydgolfer]

Folks, two different 290x's, a 1070 and the 1060 were tested in this exact rig numerous times with stock 4970k speeds, results were consistent between card and on multiple runs see the OP, the only one with the shit score is the 1060.

Here's Hampton beach in NH this afternoon. Not a bad view

I'm hoping to make it up to Maine later on in the summer finances permitting of course.... I forgot you live close to me NE ftw

 

[Sasqui]

I'm going to just admit that I'm dumb and I can't find other benchmarks for my card and that test to see if it's a NVIDIA bug or not. Hurdedur no teeth=no brain.

EDIT: Nevermind I found it, 3k less than the top scores of similar rigs. I'd go with that PCI power theory from Laki over there.

The 1070 was getting around 15k every run. The 1060 should see about 12k from every place I looked.

I'm going to check the cable when I get home

 

[Laki89]

I'd go with that PCI power theory from Laki over there.

To be exact, my theory is: power regulators (not all) on card are bad / broken / burned / short circuited

 

[Sasqui]

To be exact, my theory is: power regulators (not all) on card are bad / broken / burned / short circuited

I'm hoping I did something stupid like not push the connector in all the way. I'll repost in a bit...

 

[Sasqui]

Ok, the cable was snugly in there. Just to be sure, I grabbed the second pair of PCI-e cables and plugged one of those in. I haven't run 3DMark yet, because I saw this in idle

@W1zzard - do you know more details about the significance of that number? Does it count total draw from the PCIe slot + the cable?

 

[Vya Domus]

I'd say you should try other benchmarks like Unigine Heaven/Superposition and see if you get much lower scores in those as well. Firestrike seems really unreliable to me.

Use something to check the clocks in real time while running benchmarks and see if it throttles. I remember some Kepler cards having a BIOS bug where they would be stuck to a certain low power limit no matter what.

 

[Laki89]

Having in mind that GPU-Z is not made by nVidia, so it doesn't display a correct number in all possible states of GPU can be
I for some reason believe that your card is on 50% of power ( not 60% like i said before, sorry )
If W1zzard knows more details on that, it would be implemented in GPU-Z already, so i think that comma is not on right spot
so instead 494.7% you actually have 49.47% of TDP usage and performance similar to GTX 1050

 

[jboydgolfer]

Ok, the cable was snugly in there. Just to be sure, I grabbed the second pair of PCI-e cables and plugged one of those in. I haven't run 3DMark yet, because I saw this in idle

@W1zzard - do you know more details about the significance of that number? Does it count total draw from the PCIe slot + the cable?

P4 pointed that out earlier too

 

[Sasqui]

P4 pointed that out earlier too

Yea, I saw that. Was really hoping it was a cable problem

 

[jboydgolfer]

Yea, I saw that. Was really hoping it was a cable problem

you bought a real Gem there huh

 

[crazyeyesreaper]

Please read the OP



Interesting idea. Funny, there are people here that don't think it's worth the time test a new card. (Really)

I read the OP, I am currently testing and reviewing hardware with Nvidia GPUs same problem - better results in say Firestrike etc by 1000-4000 points with the CP edits. Drivers tended to use optimal power, in the CP changing to Maximum I saw improvements using older drivers same manual configuration results in 45%+in SLI scaling improvement. Also been dealing G-sync issues etc etc. Nvidia drivers don't crash but depending on configuration shit can be easily broken.

Only option i can recommend is grab an old drive and install win 10 on it and run Firestrike / benchmarks. Could be OS issue at this point. Also ran into that with system I am testing. OS reinstall fixed a bunch of problems that clean drivers etc could not. While it could be GPU its more likely if it was having problems it would crash or the driver would encounter a problem and fail. So its more than likely a software problem. Which leads back to driver or OS issue.

 

[Sasqui]

I read the OP, I am currently testing and reviewing hardware with Nvidia GPUs same problem - better results in say Firestrike etc by 1000-4000 points with the CP edits. Drivers tended to use optimal power, in the CP changing to Maximum I saw improvements using older drivers same manual configuration results in 45%+in SLI scaling improvement. Also been dealing G-sync issues etc etc. Nvidia drivers don't crash but depending on configuration shit can be easily broken.

Only option i can recommend is grab an old drive and install win 10 on it and run Firestrike / benchmarks. Could be OS issue at this point. Also ran into that with system I am testing. OS reinstall fixed a bunch of problems that clean drivers etc could not. While it could be GPU its more likely if it was having problems it would crash or the driver would encounter a problem and fail. So its more than likely a software problem. Which leads back to driver or OS issue.

Thanks. I have my other PC (system in stats) that this card was ultimately going to go into. Before any RMA, I did plan to give it a shot. You got any particular driver version to recommend? I just grabbed the latest from nvidia website yesterday. It's appears to be the same package as for the 1070, but I did a separate download for the 1060 just in case.

 

[crazyeyesreaper]

Not particularly I have tried 7-8 drivers now and depending on configuration i can see performance swings as large as 6000 3dmarks in Firestrike with SLI 1080s single GPU 1000-2000. This is Score not GPU only. Seeing crazy fluctuations. Seems to be a crap shoot to get it solved at the moment.

Best luck I have had is 381.65 which while scores are lower seems to be stable with the best overall performance, but I am using Windows 10. So your mileage may vary. My underlying thought is Nvidia screwed the Win 7 driver. Thus the try an old drive with Win 10 on it no need to activate just get it running with drivers and run Firestrike. Eliminates the OS / software from the equation. Another system is a good test as well. However I find a separate drive with a fresh os works best for trouble shooting a specific configuration.

If it is a Win 7 issues I would use the loophole for for special needs to upgrade to Win 10 for free.

 

[eidairaman1]

Ok Use a 1070 from here on out...

 

[W1zzard]

The GPU load value is read directly from the NVIDIA driver and displayed without modification

 

[Toothless]

The GPU load value is read directly from the NVIDIA driver and displayed without modification

Sooooo borked driver?

 

[GreiverBlade]

and they said :1060 = 980 level ... come again then .... the 2 score i saw here are not even above a 970 ...

The 1060 should see about 12k from every place I looked.

actually ... your 7k and the other at 9k ... 12k is ... well seemingly the norm, from what you've seen, but ... for me it seems also still low ... that would put it on par with a 970 actually yep that would be normal: a 980 score (well .... not really that the 970 3.5gb is far from a 980 4gb/1060 6gb tho, if it was a 1060 3gb, maybe )

mmhhh not even 12k, 7k and 9k seems like a normal run variance (well 9k anyway)
VALID RESULT
SCORE 10 561 with NVIDIA GeForce GTX 1060-6GB(1x) and Intel Core i5-6500
SCORE 12 059 with NVIDIA GeForce GTX 980(1x) and Intel Core i5-6600K
SCORE 9 271 with NVIDIA GeForce GTX 970(1x) and Intel Core i5-4590 Processor
ok the CPU differ (tho they are all not far from each other in therm of perfomances)
something is seriously wrong... with benchmark ... i recommend to let down your worries ... : your card is fine ...

or ... after seeing the TDP % :

Sooooo borked driver?

could be.... it's nvidia ... no wonder i am still on 376.53

ohhh @Sasqui ... before even thinking of RMA: retrograde your driver if you didn't already try ... nvidia latest driver can, sometime, be total crap



did your card give you unsatisfying result in day to day gaming? because : Benchmark are not a mean to test a card ... (run variance can be a pest sometime)

*downloading 3dmark to compare my 1070 ...

 

[Toothless]

and they said :1060 = 980 level ... come again then .... the 2 score i saw here are not even above a 970 ...

actually ... your 7k and the other at 9k ... 12k is ... well seemingly the norm, from what you've seen, but ... for me it seems also still low ... that would put it on par with a 970 actually yep that would be normal: a 980 score (well .... not really that the 970 3.5gb is far from a 980 4gb/1060 6gb tho, if it was a 1060 3gb, maybe )

mmhhh not even 12k, 7k and 9k seems like a normal run variance (well 9k anyway)
VALID RESULT
SCORE 10 561 with NVIDIA GeForce GTX 1060-6GB(1x) and Intel Core i5-6500
SCORE 12 059 with NVIDIA GeForce GTX 980(1x) and Intel Core i5-6600K
SCORE 9 271 with NVIDIA GeForce GTX 970(1x) and Intel Core i5-4590 Processor
ok the CPU differ (tho they are all not far from each other in therm of perfomances)
something is seriously wrong... with benchmark ... i recommend to let down your worries ... : your card is fine ...

or ... after seeing the TDP % :

could be.... it's nvidia ... no wonder i am still on 376.53

ohhh @Sasqui ... before even thinking of RMA: retrograde your driver if you didn't already try ... nvidia latest driver can, sometime, be total crap



did your card give you unsatisfying result in day to day gaming? because : Benchmark are not a mean to test a card ... (run variance can be a pest sometime)

*downloading 3dmark to compare my 1070 ...

Uh, every single benchmark I've seen has the 1060 between the 980 and 970 and it's not that big of a gap per all three, and still most likely a driver issue since a lot of people confirmed this newest driver is crap.

 

[Vya Domus]

and they said :1060 = 980 level ... come again then .... the 2 score i saw here are not even above a 970 ...

It is the same level , my score is identical to what I see for 980s , 14k. Seems like Firestrike reports normal scores sometime and other times it doesn't.

 

[GreiverBlade]

It is the same level , my score is identical to what I see for 980s , 14k. Seems like Firestrike reports normal scores sometime and other times it doesn't.

aherm... 14k is a 1070 generally

View attachment 89473

This is on my system , there dose seem to be something wrong , graphics score is almost half of mine. Make sure you didn't dialed back the power limit or messed with something within NV control panel.

this is a 970 level 980 is 12k and indeed a 1060 is rather in between with a 10k average, but you confirm the driver issue (although that score is in the norm of other 970 on the 3dmark website, but for your score it's probably a run variance since 1060 and not 970)

and still most likely a driver issue since a lot of people confirmed this newest driver is crap.

never said the opposite nvidia drivers are crap ... i know that since i had my 980 back (~ 2yrs iirc)

also yep 1060 in between 970 and 980 as you can see in my quoted post.

tho as i said ... Firestrike should not be used to see "does my card do fine" since most high score in the GPU category you would seek, would not be stock, unless there is a filter per mhz Core/vRAM that would enable you to see differences between similar clocked card, that's why i wrote to Sasqui to downgrade the driver before even thinking "my gpu is bad i need to RMA" when it's likely behaving correctly but has drivers issues.

my driver are stable in any games i play, performances are top notch, but firestrike crash in the middle of the demo ... is there an issue with my 1070? likely not ...