1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

bit-tech GPU reviews - feedback please!

Discussion in 'Feedback & Suggestions' started by Dogbert666, 29 Apr 2019.

Page 1 of 2
  1. Dogbert666

    Dogbert666 *Fewer Lover of bit-tech Administrator

    Joined:
    17 Jan 2010
    Posts:
    1,678
    Likes Received:
    181
    Hello lovely people!

    It's that time where I can no longer ignore that our GPU benchmarks are long in the tooth and need an update to newer/more demanding titles. Before I go about refreshing them all, however, I wanted to put it to the community and ask what you would like to see from our GPU reviews. What do we do right, what do we do completely wrong, what do you want to see more/less of, etc.?

    The current approach is six carefully chosen games (different APIs, different engines), a bunch of synthetics (very useful for letting our readers compare their own performance), power/thermals, and overclocking. I believe this allows us to capture enough data as to whether a card is worth your cash or not.

    I would like to finally add some proper noise testing, which I'll start looking into. Repeatable, reliable results are more important to me than having numbers for the sake of it, though, and noise testing is highly variable, especially in the current work environment, so no promises.

    Crypto, rendering, and record/capture are other areas we could explore, but gaming will remain the focus in GPU reviews.

    Basically, what data is most important to you when considering a new GPU purchase, and how is it best presented? Feel free to reference competitors.

    I understand that the needs of someone buying 2080 Ti might be different to those of someone buying a 1660, but the core benchmarks need to be scalable from top to bottom. We can look into additional testing for high-end cards where it's worth it.

    Looking forward to your thoughts!
     
    Dogbert666, 29 Apr 2019
    #1
  2. Vault-Tec

    Vault-Tec Green Plastic Watering Can

    Joined:
    30 Aug 2015
    Posts:
    14,982
    Likes Received:
    3,743
    Far Cry 5 and Division 2 spring to mind. That's Duna? And Snowdrop? Covered. Metro Exodus would be nice also. RAGE 2 shall be upon us soon too, so please try to cover as many new titles as you can. I'll be honest I've found BT reviews waning something rotten and useless to me to compare my card in the latest games to new cards.

    Then obviously BF5 and heck you would even use that new RT Star Wars Demo aswell.
     
    Vault-Tec, 29 Apr 2019
    #2
  3. Anfield

    Anfield Multimodder

    Joined:
    15 Jan 2010
    Posts:
    7,062
    Likes Received:
    970
    If you want to stay with the 6 games format, the 6 I would use in no particular order:
    1: Shadow of the Tomb Raider
    2: BF5
    3: Either F1 2018 or Dirt Rally 2 (yes I know... F1 2019 is around the corner, so replace as appropriate based on timeline)
    4: Far Cry New Dawn
    5: Overwatch (yes it runs on a potato so not of much interest for those looking at the high end cards, but for low ends cards it would be nice to have it in the stable)
    6: Hitman 2

    As for the synthetic benchmarks I couldn't care less and would not mind if every single reviewer boycotted them, but others are likely to disagree so whatever floats your boat.

    Including rendering seems like a good idea to me, Blender CUDA / OpenCL seems like a safe bet?

    Power & thermals? Just keep them as you do them now.
     
    Anfield, 29 Apr 2019
    #3
  4. Omnislip

    Omnislip Minimodder

    Joined:
    31 May 2011
    Posts:
    640
    Likes Received:
    158
    Concerning noise, I often find that reviews do not account for fan curves that bias for different thermal targets. If something is hot and quiet, or cool but loud, it is hard to tell which is actually the better card after some trivial optimisation by the user.

    If you could measure a "noise at a constant 80 degrees" or some similar measure, I would find that very useful :)
     
    Omnislip, 29 Apr 2019
    #4
    wolfticket likes this.
  5. Gareth Halfacree

    Gareth Halfacree WIIGII! Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    17,133
    Likes Received:
    6,728
    Linux compatibility.

    There are dozens of us! Dozens!
     
    Gareth Halfacree, 29 Apr 2019
    #5
    adidan, perplekks45 and The_Crapman like this.
  6. Wakka

    Wakka Yo, eat this, ya?

    Joined:
    23 Feb 2017
    Posts:
    2,117
    Likes Received:
    673
    TPU-style performance-per-watt and performance-per-pound (£) charts would be nice.
     
    Wakka, 29 Apr 2019
    #6
    edzieba and MLyons like this.
  7. MLyons

    MLyons 70% Dev, 30% Doge. DevDoge. Software Dev @ Corsair Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    3 Mar 2017
    Posts:
    4,198
    Likes Received:
    2,781
    Year of Linux on the desktop!!!
     
    MLyons, 29 Apr 2019
    #7
    adidan, yuusou and Gareth Halfacree like this.
  8. RedFlames

    RedFlames ...is not a Belgian football team

    Joined:
    23 Apr 2009
    Posts:
    15,426
    Likes Received:
    3,013
    You might want to add forkknife, apex or pubg to the games roster if you can. Especially at the lower end of things that's what a not inconsiderable chunk will be playing.
     
    RedFlames, 29 Apr 2019
    #8
  9. edzieba

    edzieba Virtual Realist

    Joined:
    14 Jan 2009
    Posts:
    3,909
    Likes Received:
    591
    Frame times.

    While I'd be happy ditching FPS altogether, I understand that "bigger number means more better" is popular for mass-market clicks. However, generating FPS (both average and instantaneous) is possible from recorded frametimes, going the other way is not possible. 2011's Inside the Second is as relevant today as it was 8 years ago, even moreso if you're interested in VR (where frame latency is king).
     
    edzieba, 29 Apr 2019
    #9
    adidan and perplekks45 like this.
  10. yuusou

    yuusou Multimodder

    Joined:
    5 Nov 2006
    Posts:
    2,879
    Likes Received:
    955
    Percentiles (0.1% lows, etc) is always nicer to see than the absolute lowest frame rate.
     
    yuusou, 29 Apr 2019
    #10
  11. The_Crapman

    The_Crapman World's worst stuntman. Lover of bit-tech

    Joined:
    5 Dec 2011
    Posts:
    7,683
    Likes Received:
    3,942
    I think having an equivalent card from the previous 2-3 generations is helpful when trying to decide if the upgrade is worth it. Although 'equivalent card' has got a lot more complicated recently thanks to Nvidia almost doubling of their product stack and separation of RTX and GTX and even based on price would be a mine field.

    E.g. New 2070, how does it compare to 1070 and 970.
     
    Last edited: 29 Apr 2019
    The_Crapman, 29 Apr 2019
    #11
    CrapBag, adidan and Omnislip like this.
  12. perplekks45

    perplekks45 LIKE AN ANIMAL!

    Joined:
    9 May 2004
    Posts:
    7,553
    Likes Received:
    1,796
    Frame times, ray tracing settings, continue to include GTX 10xx, ... just off the top of my head.
     
    perplekks45, 29 Apr 2019
    #12
  13. MightyBenihana

    MightyBenihana Do or do not, there is no try

    Joined:
    8 Sep 2011
    Posts:
    1,484
    Likes Received:
    123
    Any chance of adding the 21:9 3440*1440 resolution in?
     
    MightyBenihana, 5 May 2019
    #13
  14. Guest-44638

    Guest-44638 Guest

    THIS... not sure what percentage of the market upgrades their GPU more than annually, but some kind of throwback to older GPU's would be interesting.

    Maybe... even use (this is where I get a black mark from the mod's) World of Warcraft to compare how much improvement the newer GPU's give for the other 'standard kit'...?
     
    Guest-44638, 5 May 2019
    #14
  15. perplekks45

    perplekks45 LIKE AN ANIMAL!

    Joined:
    9 May 2004
    Posts:
    7,553
    Likes Received:
    1,796
    I agree with the above. A comparison of (9xx,?) 10xx, 16xx, and 20xx would be very interesting. I understand why you remove older cards from the graphs (re-testing with new drivers is too much work, having sent them back, ...) but I also think there are still a lot of people on "old" hardware.
     
    perplekks45, 5 May 2019
    #15
  16. MLyons

    MLyons 70% Dev, 30% Doge. DevDoge. Software Dev @ Corsair Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    3 Mar 2017
    Posts:
    4,198
    Likes Received:
    2,781
    Is there is WoW benchamrk I can just click "benchmark" and walk away and it'll do? If so Then it's a maybe.
    We normally keep 1 of every card. It's hard due to benchmark and driver changes but I wonder if we could do it better.
     
    MLyons, 5 May 2019
    #16
  17. Guest-44638

    Guest-44638 Guest

    Not that I'm aware of... it's the only game I play, so knowing whether other ones do/don't (I'm guessing more do than don't) is not in my remit... sadly.
     
    Last edited by a moderator: 6 May 2019
    Guest-44638, 5 May 2019
    #17
  18. perplekks45

    perplekks45 LIKE AN ANIMAL!

    Joined:
    9 May 2004
    Posts:
    7,553
    Likes Received:
    1,796
    The main reason I'm saying that is that I know quite a few people on geforce 9 series cards thinking about upgrading to 10/16/20 series cards or Radeon vii/580/590, but it's hard to find an overview of three generations of cards, let alone four.

    Also, just throwing it out there, maybe test with a very mainstream CPU (2600x/9600k)? I know you're trying to avoid CPU bottlenecks, but an "average rig" graph might help some people to better estimate the performance of GPU X in their current system.
     
    perplekks45, 5 May 2019
    #18
  19. Guest-44638

    Guest-44638 Guest

    Do many games last long enough to cover 4 generations of GPU these days?
    Yeah, I don't play more than one title... but do the 'current six' titles even get acknowledged by cards as old as nVidia 9xx, for comparison's to be viable?

    Going back a while Crysis was a GPU killer for ages; is there anything like that these days?
     
    Last edited by a moderator: 6 May 2019
    Guest-44638, 6 May 2019
    #19
  20. TechnoMod87

    TechnoMod87 What's a Dremel?

    Joined:
    13 Nov 2012
    Posts:
    54
    Likes Received:
    5
    probably need the resolution 21:9 3440*1440
     
    TechnoMod87, 19 May 2019
    #20
Page 1 of 2

Share This Page