Sticky

Tom's Hardware Superposition Benchmark Thread

Noticed that there wasn't a Superposition thread on Tom's. So I'm starting one. The moderators will be sent a Google Drive link to edit the master table to update it when I'm not around or don't have time.

Some rules/notes you should regard and keep in mind before you post:

1. Use standard Nvidia / AMD settings, especially don't tamper with the tesselation settings.
2. Post the screenshot, with both GPUZ and the Superposition window in the same image.
3. The only allowed tweak is overclocking. No driver tweaks please.
4. You can benchmark any preset, but 1080p Extreme is preferred as it will be the result logged
5. Please try to keep a low post count unless you have something important to say. No double+ posts.
6. Fake results will be ignored
7. This thread is community (non-moderator) maintained. Results are not always updated immediately.
8. Feel free to resubmit your score as long as it's 200+ marks above the previous score (usually within 5% of the previous). Any lower than 200 is within margin of error.
9. You will be notified via PM when your submission has been added to the master GPU chart. If you do not wish to receive a notification, just state so in your post.
10. SLI/Crossfire benchmarks are allowed.
11. Video cards that cannot run 1080P Extreme due to memory limits can be run at 1080p Medium, where 35% of that score will be used.
12. Lastly, no fanboying or excessive bragging. Do not say things like "My card is better than yours, Hahaha!" A light competitive attitude is fine, but excessive bragging and purposely bringing down other participants in an overly competitive attitude is not tolerated.

Here's the master results table:



Here's my GTX 1070 score for Superposition 1080p EXTREME (now outdated)

402 answers Last reply
More about tom superposition thread
  1. Consider the suggested thread as started!

    Here are some scores from when I first installed my Vega:

    Pro Mode:


    Game Mode:


  2. Was able to get a hair over 200 points more with a 1500Mhz OC on the old girl. Also had the fan cranked to 100% which explains the max temp of 56°C.



  3. After some performance tweaks and a couple more tries:
  4. Stock.



    Overclocked: GPU 2113 Memory 5702 and no voltage adjustment needed. +212 points.



    2113 GPU and 5940 Memory (No voltage adjustment)


    2133 GPU and 6000 Memory




    2126 GPU and 6039 Memory

  5. System was stock for this.

  6. Resubmission. GPU 2113 Memory 5702

  7. jankerson said:
    Resubmission. GPU 2113 Memory 5702



    Just had to be first in everything, huh?
  8. EquineHero said:
    jankerson said:
    Resubmission. GPU 2113 Memory 5702



    Just had to be first in everything, huh?


    LOL :D

    When I was doing runs there really wasn't much difference in scores on my system, sometimes within a point or 2. It was consistent like 4443, 4444, 4445, 4443 etc at stock clocks. So when I started OCing I found it really does takes a lot of difference in clocks to actually make a notable difference in score on my machine. Started turning it up and doing runs until I went over 200+ points, started to become a mission to hit over 200+..... Didn't have to adjust the voltage though so that was good. (No way was I going to bump the voltage)

    Superposition is really something, not for the faint at heart or weak GPU's.
  9. GPU @ 2214

  10. Vellinious said:
    GPU @ 2214


    Holy shoot. That's better than a stock 1080Ti.

    Isn't the 8GB 1080 just a really high-binned 1070? Or are 1070s just low-binned 1080 8GBs?
  11. 1070s have less shaders and texture mapping units, and don't have the gddr5x memory.
  12. EquineHero said:
    Vellinious said:
    GPU @ 2214


    Holy shoot. That's better than a stock 1080Ti.

    Isn't the 8GB 1080 just a really high-binned 1070? Or are 1070s just low-binned 1080 8GBs?


    Yeah, water cooling helps.

    Great score really.

    Most I can get is 4725 out of mine, well the most I an going to try and get, I posted the results so it's on record.

    Yes, there is a big difference between the GTX 1070 and GTX 1080's.
  13. I'll be pushing it harder this winter. My rig is setup on casters, so I can roll it out to the garage in sub-zero temps, really get the clocks ramped up.
  14. Vellinious said:
    I'll be pushing it harder this winter. My rig is setup on casters, so I can roll it out to the garage in sub-zero temps, really get the clocks ramped up.


    Saw a few scores on the site between 4800 and 5000 and higher with GTX 1080's, crazy clocks though.

    I am settling on 2113 and 5940 on mine... Never got hot at all and no voltage messing around, so I am sticking with that.

    4725 is fine with me for an air cooled GTX 1080 FTW2. :D
  15. jankerson said:
    Vellinious said:
    I'll be pushing it harder this winter. My rig is setup on casters, so I can roll it out to the garage in sub-zero temps, really get the clocks ramped up.


    Saw a few scores on the site between 4800 and 5000 and higher with GTX 1080's, crazy clocks though.

    I am settling on 2113 and 5940 on mine... Never got hot at all and no voltage messing around, so I am sticking with that.

    4725 is fine with me for an air cooled GTX 1080 FTW2. :D


    I just did the power limit mod on my 1070 by shorting the power chokes. I haven't run Superposition yet but I did run the GTA 5 benchmark at 2K pixels (4320x900) and saw above 100 FPS the whole time.

    Thing is to run the Superposition bench I need to disable nVidia Surround...what a hassle...
  16. EquineHero said:
    jankerson said:
    Vellinious said:
    I'll be pushing it harder this winter. My rig is setup on casters, so I can roll it out to the garage in sub-zero temps, really get the clocks ramped up.


    Saw a few scores on the site between 4800 and 5000 and higher with GTX 1080's, crazy clocks though.

    I am settling on 2113 and 5940 on mine... Never got hot at all and no voltage messing around, so I am sticking with that.

    4725 is fine with me for an air cooled GTX 1080 FTW2. :D


    I just did the power limit mod on my 1070 by shorting the power chokes. I haven't run Superposition yet but I did run the GTA 5 benchmark at 2K pixels (4320x900) and saw above 100 FPS the whole time.

    Thing is to run the Superposition bench I need to disable nVidia Surround...what a hassle...


    Nice. :)

    Now to see what it does in Superposition. :D
  17. jankerson said:
    EquineHero said:
    jankerson said:
    Vellinious said:
    I'll be pushing it harder this winter. My rig is setup on casters, so I can roll it out to the garage in sub-zero temps, really get the clocks ramped up.


    Saw a few scores on the site between 4800 and 5000 and higher with GTX 1080's, crazy clocks though.

    I am settling on 2113 and 5940 on mine... Never got hot at all and no voltage messing around, so I am sticking with that.

    4725 is fine with me for an air cooled GTX 1080 FTW2. :D


    I just did the power limit mod on my 1070 by shorting the power chokes. I haven't run Superposition yet but I did run the GTA 5 benchmark at 2K pixels (4320x900) and saw above 100 FPS the whole time.

    Thing is to run the Superposition bench I need to disable nVidia Surround...what a hassle...


    Nice. :)

    Now to see what it does in Superposition. :D


    It's not doing me any good without a good cooler. I'm still using the stock sink until I get my Rajintek Morpheus II Core Edition, so any time the temps go above 60C the clock drops. Not to mention at 2500MHz the card artifacts like crazy. You can only go so far on the stock heat sink and 2100MHz is about it.

    I'm sure most people here have seen my custom shroud I made from three 92mm fans. Just gotta wait for my Morpheus.

    The choke mod involves shorting the parts of the card that report power consumption, which I did, and I noticed it actually works while monitoring clocks during the bench.
  18. Ah, OK.

    Hopefully that will help then.
  19. Vega Frontier Edition after first driver update:
  20. COLGeek said:
    Vega Frontier Edition after first driver update:



    Updated
  21. Getting closer to that +200 mark re-submission requirement...Gotta get a custom BIOS or a better cooler...60C, fans 100%, 2100MHz.
    I love my Xeon though, these things are near future-proof.
    Last score was 3842. Come winter I am testing outside.

  22. Turned on HBCC and set GPU to turbo mode, had a MASSIVE bump in performance apparently

  23. Rogue Leader said:
    Turned on HBCC and set GPU to turbo mode, had a MASSIVE bump in performance apparently



    Table updated.

    I dunno if Tom's forum S/W supports spoilers, but if it does, put the images in a spoiler tag for reduced bandwidth usage on metered connections.

    HBCC is a turbospeed cache that acts in the same manner as CPU cache. It's in between the GPU and the vRAM in the same way L3 cache is between CPU and RAM. Very high speed cache that speeds up read/write buffers a lot for most apps.
  24. Rogue Leader said:
    Turned on HBCC and set GPU to turbo mode, had a MASSIVE bump in performance apparently



    That's was a massive difference, over 1,000 points.
  25. jankerson said:
    Rogue Leader said:
    Turned on HBCC and set GPU to turbo mode, had a MASSIVE bump in performance apparently


    That's was a massive difference, over 1,000 points.


    and I have to set everything as high as possible to even get 100 points...
  26. maybe I need to think about getting a 1070 card... :)

  27. The Paladin said:
    maybe I need to think about getting a 1070 card... :)



    It's almost 200% as good as the 970. Which is why I get almost double your score.

    Do it, it's fantastic and even with Skyrim 2K and Fallout 4's 60GB Hi-res DLC I never, ever max out the 8GB of vRAM. It's a freakin fast card.
  28. oh yeah no doubt, its also 500$... can you loan it for me for one to three years on a low interest :)
  29. The Paladin said:
    oh yeah no doubt, its also 500$... can you loan it for me for one to three years on a low interest :)


    I'm getting the 1070Ti as soon as I can so if you want to buy it off me for $400...Check out my custom cooler in the pic below...

    Edit:
  30. EquineHero said:
    jankerson said:
    Rogue Leader said:
    Turned on HBCC and set GPU to turbo mode, had a MASSIVE bump in performance apparently


    That's was a massive difference, over 1,000 points.


    and I have to set everything as high as possible to even get 100 points...


    Yeah, I got +282 points going from stock to 2113 and 5940......

    So a 1,000 point difference.... GAWD.... Massive.
  31. Whats funny is the system uses noticeably more power with HBCC on. It averaged around 50W-60W or so more playing Mass Effect Andromeda. Before I was just about maxing the wattage from my UPS, now the overcurrent alarm goes off all the time, so I turned HBCC off for now and ordered a bigger UPS!
  32. Rogue Leader said:
    Whats funny is the system uses noticeably more power with HBCC on. It averaged around 50W-60W or so more playing Mass Effect Andromeda. Before I was just about maxing the wattage from my UPS, now the overcurrent alarm goes off all the time, so I turned HBCC off for now and ordered a bigger UPS!


    Guess the lights all dim once that thing starts to work. LOL :D

    Well the good thing is you have a space heater now in the room. :D

    What are your 3D Mark Scores?
  33. jankerson said:
    Rogue Leader said:
    Whats funny is the system uses noticeably more power with HBCC on. It averaged around 50W-60W or so more playing Mass Effect Andromeda. Before I was just about maxing the wattage from my UPS, now the overcurrent alarm goes off all the time, so I turned HBCC off for now and ordered a bigger UPS!


    Guess the lights all dim once that thing starts to work. LOL :D

    Well the good thing is you have a space heater now in the room. :D

    What are your 3D Mark Scores?


    lol yeah prettymuch

    Firestrike - http://www.3dmark.com/fs/13959146

    Timespy - http://www.3dmark.com/spy/2611814

    Firestrike improved by 700 points with HBCC on, Timespy about 200 pts. I think the improvements are based on how texture rich the test is. Thats probably why Superposition had such a huge jump the textures on it are insane.
  34. Rogue Leader said:
    jankerson said:
    Rogue Leader said:
    Whats funny is the system uses noticeably more power with HBCC on. It averaged around 50W-60W or so more playing Mass Effect Andromeda. Before I was just about maxing the wattage from my UPS, now the overcurrent alarm goes off all the time, so I turned HBCC off for now and ordered a bigger UPS!


    Guess the lights all dim once that thing starts to work. LOL :D

    Well the good thing is you have a space heater now in the room. :D

    What are your 3D Mark Scores?


    lol yeah prettymuch

    Firestrike - http://www.3dmark.com/fs/13959146

    Timespy - http://www.3dmark.com/spy/2611814

    Firestrike improved by 700 points with HBCC on, Timespy about 200 pts. I think the improvements are based on how texture rich the test is. Thats probably why Superposition had such a huge jump the textures on it are insane.


    Good scores. :)

    Yeah it much be something with Superposition, crazy BM.

    Here are mine, both stock and OCed because I don't run my card OCed over what the boost and auto OC does.

    Stock:

    Fire Strike - https://www.3dmark.com/fs/13947500

    Time Spy - https://www.3dmark.com/spy/2147013

    OCed:

    Fire Strike - https://www.3dmark.com/fs/13907988

    Time Spy - https://www.3dmark.com/spy/2576590
  35. Asus Strix GTX 1070 Gaming OC (ROG STRIX-GTX1070-O8G-GAMING)

  36. Rogue Leader said:
    Whats funny is the system uses noticeably more power with HBCC on. It averaged around 50W-60W or so more playing Mass Effect Andromeda. Before I was just about maxing the wattage from my UPS, now the overcurrent alarm goes off all the time, so I turned HBCC off for now and ordered a bigger UPS!

    I've got a Corsair CX600 PSU, I ran my 1070 alongside a 580 both at the same time for a faster Blender export...Silicon power supply lottery is a thing I guess. The daughterboard in my Corsair CX600 is actually from a GS800 minus two 8-pin PCI plugs, so I soldered some on. The board is also labelled "800" where Corsair usually puts the wattage, so I guess someone at the factory screwed up and covered their tracks.

    EDIT: fixed a typo

    jankerson said:

    Guess the lights all dim once that thing starts to work. LOL :D

    Well the good thing is you have a space heater now in the room. :D

    What are your 3D Mark Scores?


    Gawd, have you felt the heat three Fermi cards can generate? I had dual 460s for a bit and I had to lower the heat when I used blender or even played Minecraft.

    SinxarKnights said:
    Asus Strix GTX 1070 Gaming OC (ROG STRIX-GTX1070-O8G-GAMING)

    Is that overclocked at all? That's practically the same score as my max-clocked 1070 (until I get my Morpheus II). Trying to break 4000 with a 1070 is hard :(
  37. EquineHero said:
    SinxarKnights said:
    Asus Strix GTX 1070 Gaming OC (ROG STRIX-GTX1070-O8G-GAMING)

    Is that overclocked at all? That's practically the same score as my max-clocked 1070 (until I get my Morpheus II). Trying to break 4000 with a 1070 is hard :(

    Yes it is very near the absolute max i can get unless i add some voltage to it. Also there are two 5ths on the chart.
  38. EquineHero said:
    Rogue Leader said:
    Whats funny is the system uses noticeably more power with HBCC on. It averaged around 50W-60W or so more playing Mass Effect Andromeda. Before I was just about maxing the wattage from my UPS, now the overcurrent alarm goes off all the time, so I turned HBCC off for now and ordered a bigger UPS!

    I've got a Corsair CX600 PSU, I ran my 1070 alongside a 580 both at the same time for a faster Blender export...Silicon power supply lottery is a thing I guess. The daughterboard in my Corsair CX600 is actually from a CX800 minus two 8-pin PCI plugs, so I soldered some on. The board is also labelled "800" where Corsair usually puts the wattage, so I guess someone at the factory screwed up and covered their tracks.

    jankerson said:

    Guess the lights all dim once that thing starts to work. LOL :D

    Well the good thing is you have a space heater now in the room. :D

    What are your 3D Mark Scores?


    Gawd, have you felt the heat three Fermi cards can generate? I had dual 460s for a bit and I had to lower the heat when I used blender or even played Minecraft.

    SinxarKnights said:
    Asus Strix GTX 1070 Gaming OC (ROG STRIX-GTX1070-O8G-GAMING)

    Is that overclocked at all? That's practically the same score as my max-clocked 1070 (until I get my Morpheus II). Trying to break 4000 with a 1070 is hard :(


    Umm no. There is no such thing as a CX 800. There's a CX 750 but it's based on a completely different platform so there's no cross compatibility. Not to mention the old group regulated CX series is terrible and certainly not something you want to be stressing more than their onboard connectors already allow. The new CX(M) x50 units are better ( dumbed down version of the discontinued RM line ) and have DC to DC regulation. Still not something I'd be modding so they could provide more power than intended from the factory. That's one of the worst ideas I've ever heard.

    TL: DR : I wouldn't trust a CX 600 with a 1070 from the factory much less with some home brewed mod. You can expect 3 years out of a CX x00 just like the warranty.
  39. EquineHero said:

    I've got a Corsair CX600 PSU, I ran my 1070 alongside a 580 both at the same time for a faster Blender export...Silicon power supply lottery is a thing I guess. The daughterboard in my Corsair CX600 is actually from a CX800 minus two 8-pin PCI plugs, so I soldered some on. The board is also labelled "800" where Corsair usually puts the wattage, so I guess someone at the factory screwed up and covered their tracks.


    Interesting.
    Corsair doesn't even mention a CX800. :pt1cable:

    http://www.corsair.com/en-us/power-supply-units/cx-series
  40. Because there's never been one. The CX 750 is a completely different platform that has different internals. If anything it was worse than the regular CX series. It failed in HardOCP's testing. I truly can't imagine modding a CX for any reason much less to make it capable of providing more power. Those units are only rated for 30C!

    http://www.realhardtechx.com/index_archivos/Page447.htm

    It's a fire waiting to happen.
  41. anort3 said:
    Because there's never been one.


    Oh, I know.
    I was just displaying that fact.
  42. EquineHero said:
    Rogue Leader said:
    Whats funny is the system uses noticeably more power with HBCC on. It averaged around 50W-60W or so more playing Mass Effect Andromeda. Before I was just about maxing the wattage from my UPS, now the overcurrent alarm goes off all the time, so I turned HBCC off for now and ordered a bigger UPS!

    I've got a Corsair CX600 PSU, I ran my 1070 alongside a 580 both at the same time for a faster Blender export...Silicon power supply lottery is a thing I guess. The daughterboard in my Corsair CX600 is actually from a CX800 minus two 8-pin PCI plugs, so I soldered some on. The board is also labelled "800" where Corsair usually puts the wattage, so I guess someone at the factory screwed up and covered their tracks.


    Er, there was no CX800, I don't know what that "800" on the board is but it has nothing to do with the output.

    Being that the CX600 is a known problem PSU, pushing it how you are (and adding connectors) is a recipe for disaster, that honestly is a house fire waiting to happen.
  43. Using the same PCB for different products and changing what's populated is *extremely* common.

    And changing exact product specifications and names prior to launch is also very common.

    It's possible that when that board was designed, there was expected to be a CX800, and it was going to be used for everything 430W up. Or it's from someone else's 800W PSU and it saves on design costs. Or anything.
  44. Someone Somewhere said:
    Using the same PCB for different products and changing what's populated is *extremely* common.

    And changing exact product specifications and names prior to launch is also very common.

    It's possible that when that board was designed, there was expected to be a CX800, and it was going to be used for everything 430W up. Or it's from someone else's 800W PSU and it saves on design costs. Or anything.


    Nope. The CX 430-600 use the Channel Well DSA III platform. It was created specifically for Corsair and is unique to those models.

    http://www.orionpsudb.com/platforms
  45. Sure, the platform may have been. But that doesn't necessarily mean that every part was. The modular board could have been grabbed from something else, or they intended to use the modular board on a future expansion of the lineup that never happened.
  46. Or it came off Factory Assembly Line #800.
  47. Either way, they don't stamp that wattage rating on the side of a PSU for their health.
  48. anort3 said:
    Because there's never been one. The CX 750 is a completely different platform that has different internals. If anything it was worse than the regular CX series. It failed in HardOCP's testing. I truly can't imagine modding a CX for any reason much less to make it capable of providing more power. Those units are only rated for 30C!

    http://www.realhardtechx.com/index_archivos/Page447.htm

    It's a fire waiting to happen.


    Deleted my other reply because the formatting was horrible.

    I meant to say GS800. I had CX stuck in my head from typing it so I just typed it as CX800 instead.

    Rogue Leader said:

    Er, there was no CX800, I don't know what that "800" on the board is but it has nothing to do with the output.

    Being that the CX600 is a known problem PSU, pushing it how you are (and adding connectors) is a recipe for disaster, that honestly is a house fire waiting to happen.


    I think we've established this.

    This is the very first revision, even says REV 1.0 on the PCB. I've had this since it was released, and it's powered almost every card I've had. Dual GTX460SE's, a Gigabye 580, GTX 650, GTX 660, GTX 1050Ti, 1050Ti FTW, GTX 770, GTX 950 FTW, dual GTX 550Tis, the list goes on. This PSU has proven to be extremely reliable, I wouldn't have done the mod if I wasn't confident.

    Someone Somewhere said:
    Using the same PCB for different products and changing what's populated is *extremely* common.

    And changing exact product specifications and names prior to launch is also very common.

    It's possible that when that board was designed, there was expected to be a CX800, and it was going to be used for everything 430W up. Or it's from someone else's 800W PSU and it saves on design costs. Or anything.


    I believe this is a GS800 in the shell of a CX600 with two 8-pin plugs removed. Like I said, the PSU has proven reliable, I wouldn't have done this if I didn't think the PSU couldn't handle the load.

    anort3 said:


    Nope. The CX 430-600 use the Channel Well DSA III platform. It was created specifically for Corsair and is unique to those models.

    http://www.orionpsudb.com/platforms

    Interesting.

    The board is black and has Corsair's logo on it. I'm gonna move it to a different chassis that has little windows so you can see the fan and replace the fan with an LED one.

    Rogue Leader said:
    Either way, they don't stamp that wattage rating on the side of a PSU for their health.

    I know that, but I'm fairly sure this PSU was borked somehow and that's how an 800W PCB made it into a 600W chassis. It's proven it's capable, and only time will tell.

    The fan only spins up when the interior of my case gets warm, like when I play GTA5 for 4 hours straight, and even then I can just take off the side panel.
Ask a new question

Read More

Google AMD Nvidia Gtx GPUs Graphics Cards