*Official Benchmark Thread* - Post it here or it didn't happen :D

[Mr. Fox]

5 hours ago, Rage Set said:

As the son of a Funeral Director/Embalmer, I grew up with a different spiritual fitness than others. I am sorry for your loss but to honor her, remember the memories you've created together...keep those memories close. In time you will begin to heal. 

That experience would be an interesting one, and valuable for perspective. I know some people do not believe in an afterlife in eternity and live as if our short time on this insane planet called Earth is the end of the road. I can't imagine how exasperating and futile it would be to go through life thinking it could end at any moment. It's not how long we are here, but what we believe and how we behave that sets the stage for what follows. Once you have that sorted, there is no reason to fear what is waiting on the other side of that doorway. Unless, of course, you have chosen to believe and act in a manner that should give you good reason to fear what is waiting for you.

5 hours ago, ryan said:

As for benchmarking I have no idea how to get 9200 again. I tried almost all the drivers nvidia offers, every tweak and overclock till it artifacts. and yet nothing having a hard time cracking 9k with stable clocks. but as @MyPC8MyBrainsaid my scores are great

I understand the frustration you feel. Been there, done that. I would suggest that you consider the distinct possibility that the 9200 score you achieved once upon a time may not be accurate because benchmarks (especially 3DMark) and Windows are both buggy. So, you may be struggling to replicate a score that simply can't be replicated unless the same bug or glitch, or group of bugs, reoccurs in the exact same manner. If you keep hitting the same high score over and over again regardless of tuning and tweaks, that tells me you have achieved consistency and identified the limits of the performance capacity of your laptop. That is not a reflection on you or indicative of your skill, it's the functional limit of your machine. That is an accomplishment all by itself. Most people have no idea what the functional limit of their computer is, and don't know how to find out. If you are not satisfied with the results, then the only solution at this point is to get something better. Therein lies the danger of a passion for benching. Nothing is ever good enough. You crave more and have to feed the number-chasing monster inside of you.

[MyPC8MyBrain]

39 minutes ago, ryan said:

I have no idea how to get 9200 again

there's another way to get a higher score, its technically legitimate test but I call it cheating,

there is one element you can change its value that will give you higher score in time spy and potentially others,

if you just looking to do it for the fun of it go for it otherwise its not a legitimate score,

 

there's precise steps you will need to take for the end result to be valid,

these settings must be applied 2 seconds before the bench starts (that is key),

this means you have to have everything ready to be applied but not applied until the right moment,

the right moment is as close as possible to when the bench itself starts, so time it beforehand,

during the test loading where its doing system info you want to apply the settings as close to when the actual bench starts, so apply 2 second before the screen flips into the bench scenery,

 

in nvidia profile inspector set "Texture Filtering - LOD Bias (DX)" value to -3
(you can also type a number not in the scope like -10)

if you did it right you will get a valid result that is way higher then you ever scored before, 

you will see the scenery will look different during the bench because the engine is not fully rendering textures beyond the level of detail (LOD) you will instruct it, 

 

don't pass this as legit bench, this is just for fun and giggles with friends,

 

 

 

 

[MyPC8MyBrain]

30 minutes ago, Mr. Fox said:

I would suggest that you consider the distinct possibility that the 9200 score you achieved once upon a time may not be accurate because benchmarks (especially 3DMark) and Windows are both buggy

100% agreed,

also consider variable from the time you had that score initially could have changed,
from just your room ambient temps (winter heater vs. summer AC) to fans getting little bit clogged over time,

applications running in the background, windows revisions, drivers updates, all can effect your bench score,

[Mr. Fox]

1 hour ago, MyPC8MyBrain said:

there's another way to get a higher score, its technically legitimate test but I call it cheating,

there is one element you can change its value that will give you higher score in time spy and potentially others,

if you just looking to do it for the fun of it go for it otherwise its not a legitimate score,

 

there's precise steps you will need to take for the end result to be valid,

these settings must be applied 2 seconds before the bench starts (that is key),

this means you have to have everything ready to be applied but not applied until the right moment,

the right moment is as close as possible to when the bench itself starts, so time it beforehand,

during the test loading where its doing system info you want to apply the settings as close to when the actual bench starts, so apply 2 second before the screen flips into the bench scenery,

 

in nvidia profile inspector set "Texture Filtering - LOD Bias (DX)" value to -3
(you can also type a number not in the scope like -10)

if you did it right you will get a valid result that is way higher then you ever scored before, 

you will see the scenery will look different during the bench because the engine is not fully rendering textures beyond the level of detail (LOD) you will instruct it, 

 

don't pass this as legit bench, this is just for fun and giggles with friends,

 

 

 

 

The LOD (Level of Detail) mod that you described is actually allowed for some benchmarks and not others according to HWBOT rules. While it is an accurate measurement, it is an apples to oranges comparison to scores that did not use that tweak. It is only cheating if you try to conceal it and submit it as a score that does not identify the LOD mod. 

 

What I think originally started it was 3DMark 11. It is one of the most demanding CPU+GPU performance tests of all time, and it employs tesselation on an extreme level. AMD/ATi GPUs struggled to run the benchmark well unless the GPU owner went into Catalyst Contro Center and disabled tesselation. That flipped the scenario and instead of AMD GPUs losing miserably against NVIDIA they destroyed NVIDIA scores. NVIDIA did not allow the option to disable tesselation. If you disabled tesselation in CCC on an AMD/ATi card, there were actually lots of totally missing graphic elements that were empty space on the screen and everything looked horrible. The absence of having any imagery to render allowed them to fly through the benchmark. 

 

There was a lot of debate on HWBOT on whether or not to allow  tesellation to be disabled so fake benchmarks scores could be submitted by AMD/ATi owners. If they did not allow it, trying to police the results would have been difficult if not futile. So, the LOD mod was allowed so that NVIDIA owners could level the playing field by using a similar cheat. The LOD mod with an NVIDIA GPU did not totally eliminate graphic elements completely like disabling tesselation did with an AMD/ATi GPU, but it removed the unfair advantage allowing AMD/ATi owners to cheat.

 

I am not sure why they allow the LOD mod on some benchmarks but not others.

[ryan]

just checked it out it did nothing....I don't know whats wrong with my laptop but its not accepting software or hardware mods

[Reciever]

So for some reason both of the Vega's dont like working anymore, going to try DDU in a bit and try again. They work fine until about 15-20 seconds and then the system crashes as a whole. I have a suspicion its Windows update doing what ever the hell it wants and crashing my system since the interval feels fixed.

 

The 3090 with the Galax vBIOS doesnt seem to pull more than 450w still, confirmed the vBIOS power limits in GPU-Z but I suppose I didnt try Furmark yet. I did score about 400 points higher iirc before I went to sleep last night but figured there was more in the tank so I didnt make any submissions.

 

Around 800 points in a couple of days, was hoping to take back the #33 position for the team but I may run out of hardware before that lol

[ryan]

anywho.

 

did you manage to get your 4090 @Mr. Fox

[Papusan]

Look at this old beauty bro Fox. The Caselabs chassis lacks modern connections in the front (due the old model design solutions), and I have the Corsair 1000D. But very tempting. Not everyday I see the old big glory is on the used market here home in Norway. Can't rembeber last time I have seen one.  6500 Nok = $647.31 

 

Review... https://www.xtremerigs.net/2013/01/31/caselabs-sth10/

[Mr. Fox]

1 hour ago, Reciever said:

So for some reason both of the Vega's dont like working anymore, going to try DDU in a bit and try again. They work fine until about 15-20 seconds and then the system crashes as a whole. I have a suspicion its Windows update doing what ever the hell it wants and crashing my system since the interval feels fixed.

 

The 3090 with the Galax vBIOS doesnt seem to pull more than 450w still, confirmed the vBIOS power limits in GPU-Z but I suppose I didnt try Furmark yet. I did score about 400 points higher iirc before I went to sleep last night but figured there was more in the tank so I didnt make any submissions.

 

Around 800 points in a couple of days, was hoping to take back the #33 position for the team but I may run out of hardware before that lol

I can vaguely remember hearing there were issues with the most current Catalyst drivers. Have you tried more than one driver version? The power limit can be as high at 1000W (Galax vBIOS) but you will be limited by voltage, so the power utilization won't go any higher that the corresponding available voltage allows. The problem might be the overclock is just too high for the available voltage. Not sure since I don't know all the details, just sharing my thoughts on possibilities.

41 minutes ago, ryan said:

anywho.

 

did you manage to get your 4090 @Mr. Fox

Should be here next Friday. It was a pre-order on Amazon and has not shipped yet.

37 minutes ago, Papusan said:

Look at this old beauty bro Fox. The Caselabs chassis lacks modern connections in the front (due the old model design solutions), and I have the Corsair 1000D. But very tempting. Not everyday I see the old big glory is on the used market here home in Norway. Can't rembeber last time I have seen one.  6500 Nok = $647.31 

 

Review... https://www.xtremerigs.net/2013/01/31/caselabs-sth10/

Those old gigantic Caselab cases are amazing. I always wanted one, but didn't want to pay the price needed to buy one.

[Mr. Fox]

This came today. Will investigate the pros and cons after the 4090 arrives (hopefully the end of next week).

Kind of sucks that I received an Amazon notification to preorder a Zotac 4090 AMP Extreme today for $100 less than the Suprim Liquid X. I almost ordered it and canceled the Suprim, but decided not to complicate things to save $100. But, I wanted the AMP more from the start. Seems like out of all the better 4090 models the AMP owners have less to gripe about overall... but, that's an anecdotal observation based on the idiosyncracies of the internet that nobody should rely on.

[ryan]

ok so Iv'e been testing out diff tweaks some raise the score some lower, with some helpful tweaks I gained about 5 points per tweak. heres my new normal...well overclocked but reproducable..

 

 

still experimenting with different drivers....perhaps one of you gentlemen could reccomend a timespy driver 🙂

 

basically with just tweaks handpicked I was able to raise my oc score by 31 points...yes yes this laptop is odd...I have never had so much trouble improving 3dmark scores in my life.

[Etern4l]

22 minutes ago, Mr. Fox said:

This came today. Will investigate the pros and cons after the 4090 arrives (hopefully the end of next week).

Kind of sucks that I received an Amazon notification to preorder a Zotac 4090 AMP Extreme today for $100 less than the Suprim Liquid X. I almost ordered it and canceled the Suprim, but decided not to complicate things to save $100. But, I wanted the AMP more from the start. Seems like out of all the better 4090 models the AMP owners have less to gripe about overall... but, that's an anecdotal observation based on the idiosyncracies of the internet that nobody should rely on.

 

1200W, oh ye Mighty Bro Fox? This is just for the shunt-modded GPU, right? ;)

[Reciever]

meanwhile I cant seem to crack 2100Mhz for an active clock during benchmarks, keeps going back to 2085Mhz on the rtx 3090

[Mr. Fox]

2 minutes ago, Etern4l said:

 

1200W, oh ye Mighty Bro Fox? This is just for the shunt-modded GPU, right? 😉

Probably so, but the shunt-modded 3060 Ti does not need anywhere near 1200W. I think it could replace the 1600W on the benching rig and 1200W would be ample for the 4090 and 13900K. It has native ATX 3.0 and PCI-e 5.0 support that is ideal for the 4090, but the PSU itself would be better suited to the Corsair 5000D Airflow chassis that the 3060 Ti is in. It will fit better and address the lack of room in the basement of the case with the new cable orientation. 

 

What I will probably do is temporarily install the 4090 in my work computer with the PSU permanently installed for the PSU review, then after the review, put the 3060 Ti back in the work computer and put the 4090 in my benching rig that has the 1600W PSU. I have tons of unused space in the gigantic Level 20 XT aquarium, so I do not need the space-saving feature of this new PSU design in that chassis. If that's what I do, I will order a Cablemods 12VHPWR cable to replace the three 6+2 power cables feeding the 3090 KPE. I really despise those stupid short pigtail adapters that come with 4090 GPUs. Those are so stinking idiotic.

[Etern4l]

9 minutes ago, Reciever said:

meanwhile I cant seem to crack 2100Mhz for an active clock during benchmarks, keeps going back to 2085Mhz on the rtx 3090

 

clkMax=f(T), I think. 

 

 

  

3 minutes ago, Mr. Fox said:

I really despise those stupid short pigtail adapters that come with 4090 GPUs. Those are so stinking idiotic.

 

Yeah, even the 3-way adapters are not great. I replaced mine with an original Corsair cable - so much nicer and more compact.

 

[Mr. Fox]

12 minutes ago, Reciever said:

meanwhile I cant seem to crack 2100Mhz for an active clock during benchmarks, keeps going back to 2085Mhz on the rtx 3090

That is a thermal throttle. It is designed to do that so nothing is wrong. (Thanks, NVIDIA.)

 

I have the same issue on the 3090 Kingpin. Only way around it for me is to run the water chiller without the radiator and give it about 30 minutes to run with the computer turned off so the water can get down to 0°C or colder. As soon as the water gets higher than about 10°C it starts losing clock speeds. Under normal thermal operating conditions, what starts off as 2235 or 2250 on core ends up landing around 2160 or 2175 average clock.

 

4 minutes ago, Etern4l said:

Yeah, even the 3-way adapters are not great. I replaced mine with an original Corsair cable - so much nicer and more compact.

That is exactly why I have not made up my mind yet. I intend to avoid using that tacky 4:1 pigtail. That is such an ugly low-cost solution.

[ryan]

why does everyone get locked clocks but me? on any computer everyone seems to get locked frequencies. When I set my gpu to 2000 it bounces around 1700-1780. never hits 2000mhz..I think im missing something..a setting that needs to be ticked I wish I had more than the INTRANET. would be nice to do a google search and find something....heck when I search science news on my google like only 20 pages come up and they are all hocus.

[Etern4l]

7 minutes ago, Mr. Fox said:

That is exactly why I have not made up my mind yet. I intend to avoid using that tacky 4:1 pigtail. That is such an ugly low-cost solution.

 

It just takes up 2-3 times the case space, and what if you have more than one GPU - it's a mess.

[Reciever]

4 minutes ago, Mr. Fox said:

That is a thermal throttle. It is designed to do that so nothing is wrong. (Thanks, NVIDIA.)

 

I have the same issue on the 3090 Kingpin. Only way around it for me is to run the water chiller without the radiator and give it about 30 minutes to run with the computer turned off so the water can get down to 0°C or colder. As soon as the water gets higher than about 10°C it starts losing clock speeds. Under normal thermal operating conditions, what starts off as 2235 or 2250 on core ends up landing around 2160 or 2175 average clock.

 

I think my body will give out before I reach 0c in my bedroom lol I'll wait 10-15 minutes before the next run and see if that will do the trick. I also reversed the PC orientation so that the GPU is pulling cold air straight from the window fan instead of being the furthest thing from.

 

2 minutes ago, ryan said:

why does everyone get locked clocks but me? on any computer everyone seems to get locked frequencies. When I set my gpu to 2000 it bounces around 1700-1780. never hits 2000mhz..I think im missing something..a setting that needs to be ticked I wish I had more than the INTRANET. would be nice to do a google search and find something....heck when I search science news on my google like only 20 pages come up and they are all hocus.

 

Laptops are very difficult to manipulate these days.

[ryan]

1 hour ago, Reciever said:

 

Laptops are very difficult to manipulate these days.

your telling me. I did 2 all nighters to find the cause of low performance and now that story is I wasted alot of time and \i need a new vbios if i want to progress

[Mr. Fox]

2 minutes ago, ryan said:

your telling me. I did 2 all nighters to find the cause of low performance and now that story is I wasted alot of time and \i need a new vbios if i want to progress

You're doing something with your laptop that it was never designed for. Do you know how your benchmark scores compare to the same or like configurations? If you are ahead of the pack, you may have reached Nirvana and don't know it.

[ryan]

it has some wiggle room\ for example I found out the legion 5 with a 115w 3060 scores 9500 in timespy so its all possible just a matter of finding out the loop hole or just accept what ive done like you said..benchmarking is super addictive

[Reciever]

 

Some new scores for the 3090 with the Galax vBIOS

 

 

Probably it for now, I plan to get one of my precisions up and running again for some cinebench runs, 20c/40t dual CPU workstation.

 

I also have a 4790k I recently picked mainly for Plex server but I can get some runs done with it.

[MyPC8MyBrain]

2 hours ago, Reciever said:

I cant seem to crack 2100Mhz for an active clock during benchmarks

prob redundant but worth checking, did you enable these options in your MSIAfterbusrner.ini

 

 

35 minutes ago, Reciever said:

Some new scores for the 3090

your timespy graphics score is kinda low even for my laptop 😞 

is your power plan in bios and widows on max performance profile?

 

@ryan you can compare how you stack up to other same spec devices here https://www.3dmark.com/search

 

 

[Reciever]

 

12 minutes ago, MyPC8MyBrain said:

prob redundant but worth checking, did you enable these options in your MSIAfterbusrner.ini

 

 

your timespy graphics score is kinda low even for my laptop 😞 

is your power plan in bios and widows on max performance profile?

 

@ryan you can compare how you stack up to other same spec devices here https://www.3dmark.com/search

 

 

If I had a GTX 970 that ran parity to a 3080Ti it would be quite the marvel indeed 🙂

 

I'll add my TS scores in just a second, screens are on a different PC