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

Reviews: GeForce GTX 1660 Ti testing Galore - MSI, ASUS and Palit

Discussion in 'Frontpage news' started by Hilbert Hagedoorn, Feb 22, 2019.

  1. HardwareCaps

    HardwareCaps Master Guru

    Messages:
    452
    Likes Received:
    154
    GPU:
    x
    non-sense, BF5 is highly optimized. the performance is solid and consistent unlike a lot of other games. BF5 is more accurate than Firestrike today.
    Firestrike fails to show the real world performance today. Paul's hardware does the exact test (Firestrike Ultra) only to show that the RX 590 is much slower than the 1660Ti in all other benchmarks and games

    Video removed, please stop inserting Videos as a reaction - this is a discussion forum.
     
  2. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,023
    Likes Received:
    1,832
    GPU:
    -NDA +AW@240Hz
    But that statement is incorrect because no benchmark can cover entire spectrum of HW.
    If you make game heavy on Shader/compute, you'll need to stress other parts of GPU. And it will favor likes of Vega 64.
    Make something what does little texturing and almost no shading, but has loads of polygons... and boom Vega 64 sux.
    What you see is effect of unbalanced GPUs. But who is to say that it is unbalanced. In 2 years best looking games may use techniques utilizing parts of GPUs which are underdeveloped today.

    No benchmark is accurate because ti does show you some composite result. Want to have accurate benchmark? Then get someone to create following:
    - separate execution of workloads for each part of image preparation, so user can clearly see capability of GPU in that department
    - > You'll see how good texturing power card has in comparison to others
    - > how many polygon checked for visibility per second
    - > how many pixels run through shader units based on shader code complexity per second
    ...
    Then you do same workloads, but in combination. Therefore users can see how can GPU handle 2/3/... stages at same time and where limitations come from.
    And finally you run it all at once to see how it does in composite.

    Reason why you see anomaly is because cards have unbalanced number of execution elements.
    And reason why BF5 is not matching much is because it has glued in experiment and had workarounds to allow that experiment survive.
     
    carnivore likes this.
  3. BlackZero

    BlackZero Ancient Guru

    Messages:
    8,868
    Likes Received:
    451
    GPU:
    RX Vega
    I want to see a review by this dude here

    [​IMG]
     
  4. HardwareCaps

    HardwareCaps Master Guru

    Messages:
    452
    Likes Received:
    154
    GPU:
    x
    but games lean towards certain rendering techniques and effects. this is why Tessellation is much more important today than it was 4 years ago.
    if you GPU does great in heavy shading workload but almost no game uses it then it's pointless.
    the question is what represents modern games today, it's not firestrike.
     

  5. alanm

    alanm Ancient Guru

    Messages:
    8,638
    Likes Received:
    1,081
    GPU:
    Asus 2080 Dual OC
    Before Firestrike there was... 3DMark 99, 2000, 2001, 03, 05, 06, Vantage, 11, then... 3dmark w/Firestrike (2013). It had been updated every couple years of so. FS 6 years now. Have a feeling something else will replace it soon.
     
  6. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,023
    Likes Received:
    1,832
    GPU:
    -NDA +AW@240Hz
    Proper question is: "Who is to define what is 'modern' rendering technique."

    Because that FS is using modern rendering techniques in my book. It does not reflect new game's performance for simple reason.
    Those games are implementing nVidia's technologies directly. Are wrapped around them and optimized to run well with them.

    FS is not nVidia centric benchmark which would go and implement same black-boxes game studios implement. And we can be happy for that, because nVidia's effects are made to fit their architecture.
    = = = =
    Take it this way (fictional scenario depicting how they do it):
    - New HW is able to execute INT4/8 and FP16 at same time
    - > Next effect will be based on both INT4/8 and FP16
    - - > If AMD/intel does not have same parallel execution capability, their performance is going down

    That's why nVidia introduces "new best visual improvement" together with making any small improvement in uArch. New uArch change makes night and day difference for new technique (which is nowhere to be seen at launch), but does almost nothing for older games. (That's best case scenario for nV as sites will change their game benchmarks to accommodate new HW.)

    Their approach does 2 things:
    - hastening obsolescence of older generations
    - constantly shifting paradigm so AMD/intel has to play catch

    Having claws this deep in gaming industry does a lot of harm once abused.
     
    carnivore likes this.
  7. HardwareCaps

    HardwareCaps Master Guru

    Messages:
    452
    Likes Received:
    154
    GPU:
    x
    What games? every game tested the 1660Ti was faster than the 590. Nvidia supported, AMD supported, DX11, DX12, Vulkan.
    I get that you hate nvidia but overdefending AMD in a clear situation when they lose is blindness.
    there are clear tendencies to what game developers favor. this is exactly what Futuremark is doing when building these synthetic benchmark they talk with developers and understand how they are implementing their games, what features are they pushing more heavily and build a benchmark around that.

    wait So FS is ok but Time spy is not? SuperPosition? both are a lot more modern and both show the 1660Ti leads by alot. your points are horrible at best

    I'm sure Navi will be super competitive but it seems to be pretty far off.
     
    Last edited: Feb 23, 2019
  8. sverek

    sverek Ancient Guru

    Messages:
    4,989
    Likes Received:
    1,709
    GPU:
    NOVIDIA -0.5GB
    I like this new card a lot. The age of GTX10xx series is finally over with it. (not considering GTX1050Ti).

    This is pretty much no brainier card for its price, performance and effectiveness.
     
  9. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,023
    Likes Received:
    1,832
    GPU:
    -NDA +AW@240Hz
    Think again. Or start to learn how to think. Read statements. Preferably your own.
    = = = =
    I never wrote that one is OK and another is not. I countered your sentiment that FS, which you do no longer like, should go away.

    So, do not put words into my mouth.
    = = = =
    As for that FS score. There is reason for it. Think for a moment about it. Why it performs about same as RX-580/590 under some workload?
    Benchmark reveals that card has some weak point. And instead of trying to find out what that is, you want to throw benchmark away?
     
    BlackZero likes this.
  10. HardwareCaps

    HardwareCaps Master Guru

    Messages:
    452
    Likes Received:
    154
    GPU:
    x
    FS is not meant "to find weakness" what none-sense, but to predict real world performance for consumer GPUs. they literally write it on their website.
    if FS finds a "weakness" in this GPU then Timespy find a "weakness" in the RX 590/ Fury X.

    your point makes no sense, it's like running a game from 8 years ago and claiming here the 1660Ti "has a weakness" or whatever that even means....
    AMD shilling
     
    Last edited: Feb 23, 2019

  11. Andy Watson

    Andy Watson Active Member

    Messages:
    58
    Likes Received:
    13
    GPU:
    960
    I'm not sure why HardwareCaps keeps going on about 590 when the competitor at current price structure is the Vega 56. Actually I know, an nvidia fanboy!

    I've been around since 3dfx v tnt so I can spot one a mile off. ;)

    Vega 56 is cheaper currently and faster and has more VRAM, plus some games which you may or may not like. On the downside it is more noisy. The power draw is not an issue as I assume anyone spending £250+ on a video card does not use a 100w PSU costing £15. Neither will they faint when seeing their next electricity bill.

    So, if you can live with the noise, at current prices AMD takes the crown, if you can live with the noise. If only all those pesky phone manufacturers were not stealing all the 7nm process time ...

    At last a price war .. or at least shots fired, not seen one of those since mining for money. Good news. Let it continue.
     
  12. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,023
    Likes Received:
    1,832
    GPU:
    -NDA +AW@240Hz
    I do agree. My explanation makes no sense... to you. I did, in simple enough way, described you that there are multiple different actions which have to be done to render frame.
    GPUs are not equal at architectural block count level. Therefore differently configured GPUs do finish each action at different rate.
    If GPU is weak at something, there is going to be few situations where this weakness is revealed.

    And try to be less single-task-oriented. FS may be benchmark, but if GPU sux at it, there is reason for it.
    Ignoring it because it is called "benchmark" and not "architectural unbalance detection tool" is same as saying: "Ignore sucky performance in FH4, because it is "Game" and not 'architectural unbalance detection tool'." And that it should not be benchmarked.
    Or even better, asking to stop using games for comparison when you have benchmarks.
     
    Last edited: Feb 23, 2019
  13. HardwareCaps

    HardwareCaps Master Guru

    Messages:
    452
    Likes Received:
    154
    GPU:
    x
    o_O
    The point of the product is to play games. if the benchmark fails to give reliable data about the product performance then it's not relevant.
    games are the best way to test a GPU that was meant just for that.... true there's variation between different engines but not as big as you describe them.
    I'm done talking... you're just an AMD fanboy I get it....
     
  14. Andy Watson

    Andy Watson Active Member

    Messages:
    58
    Likes Received:
    13
    GPU:
    960
    "I'm done talking... you're just an AMD fanboy I get it."

    The problem is that because you are not neutral yourself, you are fanboy for nvidia, you have problems with any comment that is not pro-nvidia, you see it as completely the opposite and then claim they are biased.

    You put a filter on everyone's comments from a biased viewpoint, so even one that is fairly neutral and accurate can be painted how you see it, which is heavily from one side of the spectrum. I have seen this countless times over the years. The problem is the biased person can never spot they are biased.

    Glad you decided to stop talking. Perhaps some thinking time can be done during the pause?
     
    JAMVA likes this.
  15. warlord

    warlord Ancient Guru

    Messages:
    2,137
    Likes Received:
    684
    GPU:
    Null
    The cheapest rtx2060 is still the better option in my opinion. Unless they can drop the price even further in gtx1660ti of course. It is good, but not good enough for the price.
     
    Robbo9999 and Fox2232 like this.

  16. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,023
    Likes Received:
    1,832
    GPU:
    -NDA +AW@240Hz
    You can go around and see all kind of workloads...
    - In Redshift, GTX 1070 does finish rendering 40% faster than GTX 1660Ti.
    - In luxmark 3.1, GTX 1070 is just 12% faster than GTX 1660Ti.
    - Or here, see how 1660Ti is great winner when compared with GTX 1070 in some tests. And then it is perfectly equal in other. And finally loses badly in rest. And same is true in comparison of GTX 1660Ti and RX-580/590. Yes, GTX 1660Ti loses badly in some aspects even against RX-580.

    There is absolutely no reason to blame benchmark for showing that something in card is way out of balance. Looks like nV did cut bit too deep.
    It is still good card, but out of balance. Cards like GTX 1070 & RTX 2060 are well balanced.

    And if I am to be "AMD fanboy" for disagreeing with your idea of discarding benchmarks you do not like, you are going to call a lot of people who never even had AMD's card: "AMD fanboys".
     
  17. HardwareCaps

    HardwareCaps Master Guru

    Messages:
    452
    Likes Received:
    154
    GPU:
    x
    can't see a single point of biased here. all I said is that the 1660Ti is superior to the RX 590. which is a fact.
    now luxmark is great but that's not what the card is meant for. AMD always have more compute power but fails to deliver because of other issues like driver overhead.
    the point was made.
     
  18. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,023
    Likes Received:
    1,832
    GPU:
    -NDA +AW@240Hz
    I linked you data showing that RX-590 has higher texture fillrate as one of examples from where similar or even worse performance of GTX 1660Ti may come.
    If you compare abilities of those GPUs in detailed enough level, you can find why even RX-590 can perform better.

    But RX-590 is not goal here. Goal is, that FS which reveals it is valid.
     
    warlord likes this.
  19. warlord

    warlord Ancient Guru

    Messages:
    2,137
    Likes Received:
    684
    GPU:
    Null
    @Fox2232 You are absolutely right. Even at a couple of instances, if gtx1660ti loses from weaker gpus, it means, it is weak itself. It has some optimization problems or hardware limitations, that they will prevail at some time, or even if they won't, it is a risk to take. When you buy a product, you want stability and reliable performance all over. You never said it is a bad gpu.

    GTX 1660ti for me is the cheap version of Radeon VII debut. The same situation, even if it is strong, it loses some times by weaker gpus or it rarely equals a better one. That's why it should be cheaper too. For me it is exactly the same problem. Experimental releases to buy for a crowd never aimed to before, and the owner will be an early tester for future notice.
     
  20. airbud7

    airbud7 Ancient Guru

    Messages:
    6,879
    Likes Received:
    3,327
    GPU:
    pny gtx 1060 xlr8
    I think you just defined yourself.
     

Share This Page