HWUB/Techspot adds more games, 5700XT falls further behind 2070 Super.

Discussion in 'Video Cards' started by Snowdog, Aug 27, 2019.

  1. Snowdog

    Snowdog [H]ardForum Junkie

    Messages:
    9,216
    Joined:
    Apr 22, 2006
    The only reason you wouldn't compare it to the card closest in performance, is because you are purposefully trying to mislead.

    I guess you somehow got the 5070 for $250, which is a fantastic deal. At $250, I would certainly recommend the 5700 over the 1660 Ti which is the only NVidia card in that price range, though I doubt many could replicate that pricing.
     
  2. ManofGod

    ManofGod [H]ardForum Junkie

    Messages:
    10,791
    Joined:
    Oct 4, 2007
    No, I said $250 less than a 2070S, try to keep up. :)
     
  3. GoldenTiger

    GoldenTiger [H]ard as it Gets

    Messages:
    18,758
    Joined:
    Dec 2, 2004
    2070s is $500. You claimed to get a 5700, a much slower card, for 250 less. That would mean you supposedly paid 250. Basic math. Do try to keep up!
     
    5150Joker, Snowdog and Dayaks like this.
  4. ManofGod

    ManofGod [H]ardForum Junkie

    Messages:
    10,791
    Joined:
    Oct 4, 2007
    2070S is typically around $530. I paid $284 for my barely different 5700 so you do the math. Oh, and I took a look at the Guru3d review and as I thought, very little difference, especially at the 1440p resolution I play at. Keep trying though and nice play. At most, the difference between the 5700 and the 2070S was about 7fps on average at 1440p, which is what I game at. Therefore, it is not a much slower card. I did not claim anything, I stated a fact jack.
     
  5. GoldenTiger

    GoldenTiger [H]ard as it Gets

    Messages:
    18,758
    Joined:
    Dec 2, 2004
    Cool and all but the 2070s is $500. Nice try spinning. P. S. The 1 percent frame times are awful on your 5700 too.
     
    5150Joker likes this.
  6. ManofGod

    ManofGod [H]ardForum Junkie

    Messages:
    10,791
    Joined:
    Oct 4, 2007
    Not on average: https://www.newegg.com/p/pl?d=2070+super
     
  7. Snowdog

    Snowdog [H]ardForum Junkie

    Messages:
    9,216
    Joined:
    Apr 22, 2006
    Even $284 is a great deal for the 5700.

    You are comparing a special deal you got (open box?), vs the average price, not for the actual competing product, but one tier up on that...

    Have you been taking lessons from GamerX on how to make tilted comparison?
     
    Last edited: Aug 31, 2019
    Ready4Dis, Araxie, amenx and 5 others like this.
  8. GoldenTiger

    GoldenTiger [H]ard as it Gets

    Messages:
    18,758
    Joined:
    Dec 2, 2004
    Who cares about an average when you can just choose to get a $499 reference which has a nice cooler on nVidia nowadays? Your one off special doesn't make any sense to debate since it isn't replicable. And the right product to compare to is the 2060s, not the 2070s.
     
    Ready4Dis and Snowdog like this.
  9. ManofGod

    ManofGod [H]ardForum Junkie

    Messages:
    10,791
    Joined:
    Oct 4, 2007
    Then why are you debating something and this thread is about the 2070S but hey......LOL! :D $250 pocketed, I call that a plus in my book. :) Oh, and the 5700 is typically faster than teh 2060S so there is that, if you really need to go there. I care and most other folks do as well.
     
    Gamer X likes this.
  10. GoldenTiger

    GoldenTiger [H]ard as it Gets

    Messages:
    18,758
    Joined:
    Dec 2, 2004
    Wrong. You're doing a great gamer x imitation though!
     
    Araxie likes this.
  11. 5150Joker

    5150Joker 2[H]4U

    Messages:
    3,139
    Joined:
    Aug 1, 2005
    They go to the same school.
     
  12. ManofGod

    ManofGod [H]ardForum Junkie

    Messages:
    10,791
    Joined:
    Oct 4, 2007
    So, I see an AMD postulation thread has turned into finger pointing now from the envy team. Oh well, at least I paid for my hardware from my own hard earned money. Oh, that and my 5700 Reference model is only a few frames per second off from a 2070S for a lot less money.
     
    Gamer X and blackmomba like this.
  13. Snowdog

    Snowdog [H]ardForum Junkie

    Messages:
    9,216
    Joined:
    Apr 22, 2006
    It's even fewer frames per second away from a 2060S which is the fair comparison to make.

    Your unwillingness to make the fair comparison, and comparing your open box deal, to the average prices no less, is obviously a tilted biased comparison.

    Tilted comparisons don't make your case, they just make you look biased.
     
    Ready4Dis, GoldenTiger and Araxie like this.
  14. ManofGod

    ManofGod [H]ardForum Junkie

    Messages:
    10,791
    Joined:
    Oct 4, 2007
    Actually, it is a few frames per second ahead of the 2060S but then again, you might want to look at the video you posted in the original link, it says 2070S.
     
    Ranulfo and Gamer X like this.
  15. Snowdog

    Snowdog [H]ardForum Junkie

    Messages:
    9,216
    Joined:
    Apr 22, 2006
    Sure it's ahead if you cherry pick the games. Average many games, the 5700 is behind the 2060S.
    https://tpucdn.com/review/amd-radeon-rx-5700/images/relative-performance_2560-1440.png

    5700 100%
    2060S 108% (8% faster)
    2070S 127% (27% faster)


    Sure the video says 2070S, It also says 5700XT. If you step down one step on one side, it makes sense to do the same on the other.
     
    GoldenTiger likes this.
  16. MangoSeed

    MangoSeed Limp Gawd

    Messages:
    487
    Joined:
    Oct 15, 2014
    The 5700 is great for the money especially the aftermarket ones. It's not faster than the 2060S though. For a few dollars less you get a few fps less.

    Cheaper cards are usually better value than more expensive ones so no surprise there. Enjoy your new card.
     
    Ready4Dis and cybereality like this.
  17. ManofGod

    ManofGod [H]ardForum Junkie

    Messages:
    10,791
    Joined:
    Oct 4, 2007
    I checked, the games I play ard faster than a 2060s and only a few was framed per second off from a 2070s. Then factor in the cost and a really good freesync monitor and it is golden.
     
  18. Snowdog

    Snowdog [H]ardForum Junkie

    Messages:
    9,216
    Joined:
    Apr 22, 2006
    No one is saying you made the wrong choice.

    If sticking strictly to price points. Even the regular priced $350 5700 wins it's price range for me. Obviously if you get it for less than that, I consider it an even bigger win.

    I just don't consider it realistic competition for 2070S.
     
  19. Gamer X

    Gamer X Limp Gawd

    Messages:
    347
    Joined:
    Jul 5, 2019
    FFS I know this.

    I was using that "wafer yield illustration" to illuminate how GPU dies are priced. And exactly how a GPU size, is factored into pricing, and thus, how AMD can use the advantage of 7nm and smaller chips to saturate the market & tilt the economic scale in their favor. Exactly like they are doing to Intel, overwhelming them with tons and tons of small-inexpensive chips.

    And yes more chips per wafer, is the exact definition of Economy of Scale: a proportionate saving in costs gained by an increased level of production.




    I never said anyone cared about chip size.

    I had to explain how die size determines cost and educate a bunch of people who seemed to not know (but they do now). SO we all are in the better and do not have to have that discussion anymore. RDNA is more efficient at Games than GCN and Turing. (It takes less and does more with it). Navi10 was never meant to compete with the RTX2070 Super... it was a Polaris replacement, but after taping out RDNA was so good, AMD saw it punching above it's intended weight... past the Vega64 and it's own R7. It was no wonder AMD said RDNA is the future of gaming, because GCN in Games is horrible compared to RDNA.


    So the question is... is the SUPER worth $100 more?
    Or.. is it worth holding on to that extra $100 until Black Friday and get a 5800 Series..
     
  20. MangoSeed

    MangoSeed Limp Gawd

    Messages:
    487
    Joined:
    Oct 15, 2014
    No. You really should stop repeating incorrect claims about things you don’t understand. Chips per wafer affects marginal cost of production. It has nothing to do with the fixed costs that are amortized by economies of scale.

    In addition to that the marginal cost of production for 7nm is not currently an advantage given the high cost of 7nm wafers. All of this has already been repeated to you as nauseam so feel free to continue on your crusade of misinformation.

    Yes, this is the correct definition of economies of scale. Now bonus points for this next question.

    Which companies currently produce more CPU and GPU chips than AMD and therefore enjoy greater economies of scale?
     
    GoldenTiger, 5150Joker and Dayaks like this.
  21. Snowdog

    Snowdog [H]ardForum Junkie

    Messages:
    9,216
    Joined:
    Apr 22, 2006
    Still Totally ignoring the increased wafer cost at 7nm that everyone pointed out to you repeatedly.
    https://hardforum.com/threads/hwub-...ind-2070-super.1985908/page-2#post-1044315797

    Note the writing on the slide. "Cost per yielded mm^2 for a 250mm"

    Yielded chip cost essentially doubled at 7nm. So even if the die were half the size, you aren't ahead on cost/chip. And news flash, Navi 10, isn't half the size of TU106.

    Your argument is completely groundless. Yet you continually repeat it, after it has been thoroughly debunked, multiple times, by multiple people.

    Just NO. It's nothing to do with chips/wafer. It's everything to do, with producing many more products than your competition, to better amortize fixed costs.
     
    Ready4Dis, GoldenTiger and Dan_D like this.
  22. ManofGod

    ManofGod [H]ardForum Junkie

    Messages:
    10,791
    Joined:
    Oct 4, 2007
    The last part has all of nothing to do with water and 7nm manufacturing costs. It is also logical that you will get more out of 7nm wafer than a 14nm wafer. Otherwise, by the math mentioned here, the cost would be 4 x that of 14 rnm chips. Basic math would indicate practically no increase in cost for the same transistor count per wafer.
     
  23. Dayaks

    Dayaks [H]ardness Supreme

    Messages:
    7,113
    Joined:
    Feb 22, 2012
    Snowdog is specifically addressing Gamer X’s incorrect statements about economies of scale. Economies of scales is basically your non-reoccuring costs and capital being spread out over a vast amount of sales. Those costs are usually somewhat fixed regardless of the qty of product. (Unless you hit breakpoints where you need an additional production line, ect.). What Gamer X is talking about is a reoccuring cost reduction, which has nearly nothing to do with economies of scale. He’s also over exaggerating the yield hit of the die size differences.

    You are right though, it’s about the same per transistor and double per mm^2 of wafer for cost. Gamer X ignores the double per mm^2 part for 7nm or the same per transistor. He pretends it’s the same cost per mm^2.

    For years I was technical lead that would input most of the variables into a pricing for a product that would range from 5k to hundreds of thousands of units per a year, so I have thought about these things before...

    A general vent - A lot of the posts here remind me of a manager I had that used all the key catch phrases but had no idea what they meant. It was annoying.
     
    Last edited: Sep 1, 2019
    Ready4Dis, GoldenTiger and Snowdog like this.
  24. Gamer X

    Gamer X Limp Gawd

    Messages:
    347
    Joined:
    Jul 5, 2019



    You are not doing your due diligence and are the one spreading mythical absurdities. You do not realize the "fixed cost" of 7nm node process, has diminished substantially by every tape out. TSMC's 7nm node is not new anymore.

    7nm++ and 7nm EUVL is coming next year, with 5nm coming in 24 months time. AMD has quite a bit of 7nm in production (What is it 9 chips now @ 7nm?) and is the Worlds leader in this technology. Rome, Zen, Radeon, etc... are all on 7nm.


    Nvidia..? Don't confuse experts with 7mm like AMD is, with experts at 12nm... which Nvidia is. Nvidia has not even taped out a 7nm node yet...!

    It will be EXPENSIVE for them, it is no longer expensive for AMD. Matter of fact, Nvidia just signed a 7nm EUV deal with Samsung... almost 4 years after Dr Su signed her exclusive 7nm contract with TSMC. So again, TSMC's 7nm process is well up and running and yields are great and the cost of running the process has gone down significantly. Just like the cost of producing 14nm/12nm declined after the process matured after the first years. This is expected and often noted in ROI and discussed on Wall St. Nothing new.

    So, do understand this FACT: that TSMC's 7mn process is no longer expensive to run, as the 7nm node is TSMC's new base and they have plenty of capacity. 7nm Wafers are TSMC's norm. AMD already ate the premium in the billions, for moving to 7nm. That again is the cost and expense of a new node, not that the 7nm node process is twice as expensive, like you keep falsely claiming.

    The fixed cost = the cost of transitioning, while the process gets cheaper and cheaper and cheaper... don't argue with me, argue with history & facts. Also, as the node process matures, so do the yields. (Also, what AMD learns with Navi10, will be applied to Navi20)




    Lastly, both TSMC and AMD have made public their great strides in 7nm. And have both mentioned their continued partnership and are aggressively moving forward to 7nm++, 7 EUVL and 5nm node processes. TSMC has plenty of 7nm capacity and a contract with AMD.

    Dr Su is going to be relentless. And using AMD's advantage of small dies, will use Economy of Scale to undercut the GPU market.






    * ED: Going to start a new thread... to continue this scaling discussion, since many here don't understand why smaller chips cost less than bigger chips.

     
    Last edited: Sep 1, 2019
  25. Dayaks

    Dayaks [H]ardness Supreme

    Messages:
    7,113
    Joined:
    Feb 22, 2012
    You need links to back that up. You’ve been shown slides from AMD themselves showing it is more costly. The vast manority of what you say isn’t how any of this actually works.

    Lets pretend any of what you say is correct. Why isn’t the 5700xt 1/2 the price of the 2070S and AMD wiping the market out at 80% gross margin?
     
  26. Gamer X

    Gamer X Limp Gawd

    Messages:
    347
    Joined:
    Jul 5, 2019

    I am quoting this^ for posterity.
    You are tripping over yourself and hypocritical in your own logic. Arguing for the sake of it, or you misstated what you wanted to say...?

    Again, you are overly obsessed with that illustration I posted, but still don't even understand why I posted it. I should've just PS all the words out of it... and maybe you would've got the bigger picture. Smaller chips have an doubly increase in yield, than do larger chips.
    • 1) From being smaller, there are more per wafer, thus naturally yield more numbers of chips, sometimes 3x more. (see previous illustration)
    • 2) Since chips are smaller, they have less defects than larger chips, thus making their ratio of good -to- bad, much better than that of bigger chips. (see previous illustration)

    That^ right there is why Companies are forever seeking smaller and smaller node Processes, because they want to leverage economy of scale, in the market.



    To all.… because I can't stop laughing anymore at biased cheerleaders and want to clarify for all:

    Economies of Scale in Semiconductor Manufacturing
    How to Achieve and to Destroy
    by Michael Leitner


    2 Economy of Scale in Semiconductor Manufacturing
    2.1 Economy of Scale - General Principles
    2.2 Breakdown of Manufacturing Costs
    2.3 Product CostModels and Their Importance
    2.4 Capital Costs - Depreciation
    2.4.1 Investments
    2.4.2 Calculating Capital Costs - The Costing Model
    2.4.3 Some Definitions about Tool Availability and Processing Speed

    • 2.4.4 Operating Curves
    • 2.4.5 Capacity Planning
    • 2.4.6 Granularity of Investments
    • 2.4.7 Process Complexity Factors
    • 2.4.8 Tool Complexity Factors

    2.4.9 Single Technology FAB’s versus Multiple Technology FAB’s
    2.5 Personal Costs
    2.5.1 Operators (ProductionWorkers)
    2.5.2 Maintenance
    2.5.3 Process Engineering
    2.6 Material Costs and Payed Services
     
    Last edited: Sep 1, 2019
  27. Gamer X

    Gamer X Limp Gawd

    Messages:
    347
    Joined:
    Jul 5, 2019
    Navi10 doesn't cost AMD as much as Vega, (if they were the same size it might, but Navi10 is much smaller than Vega10). And AMD are making mad killing on the 5700 series, because the 251mm^2 margins are so great. You will see evidence of this soon after the first of the year, when the sales data and revenues come in. TSMC has plenty of capacity and AMD will be bringing out it's 3rd 7nm GPU soon.

    Also, why should Dr Su sell her Navi10 chip for $199, when she can easily get $399 for it... (and it is still outperforming/competitive with cards a $100 more...). The fact is, Nvidia can't compete any lower, or they would be loosing money. Super is big and Dr Su knows they have nvidia on the ropes. I wouldn't be surprised if AMD isn't letting the AIB's have a little taste of those profits either.


    Subsequently, those slides will back me up, because you guys didn't watch Dr Su's keynote and why She put that slide up there... and how AMD is overcoming the obstacles in those slides.
     
  28. Dayaks

    Dayaks [H]ardness Supreme

    Messages:
    7,113
    Joined:
    Feb 22, 2012
    Because AMD only has 30% market share and nVidia is the premium brand with more features and clout. If what you are saying is true they can sell it at $250 and have 60-80% gross margin, more importantly win mind share... and actual “economies of scale.”

    If facts are on your side, provide some links and/or charts.
     
    Last edited: Sep 1, 2019
    Ready4Dis likes this.
  29. Snowdog

    Snowdog [H]ardForum Junkie

    Messages:
    9,216
    Joined:
    Apr 22, 2006
    Troll repeats same ignorant falsehood (based on pretending that AMD slide doesn't exist). Just in a different context.

    In reality, Vega and Navi dies are about the same cost.

    Navi is about half the size (1/2 X), but is on a process which costs about twice as much (2 X, see AMD slide). 1/2 x 2 = 1
    Or
    Vega is about twice (2 X) as big, but it is on a process that costs about half the price (1/2X). 2 x 1/2 = 1
     
  30. ManofGod

    ManofGod [H]ardForum Junkie

    Messages:
    10,791
    Joined:
    Oct 4, 2007
    RTG will never win mind share by severely undercutting their profits for R&D. Doing so has never caused the AMD / RTG market share to increase because then the Green Team folks just keep buying the Green stuff, just because. AMD's CEO knows what she is doing and that is why AMD / RTG is way ahead of where they were just 2.5 or so years ago.
     
    Maddness likes this.
  31. Dayaks

    Dayaks [H]ardness Supreme

    Messages:
    7,113
    Joined:
    Feb 22, 2012
    Really, they’ve tried half the price like Gamer X said they are capable of with 60-80% GM? I don’t remember that. And according to his logic they’d still be making double the margin they are now. I am pretty sure most people would buy a $250 5700XT over a $500 2070S.
     
    Last edited: Sep 1, 2019
  32. Gamer X

    Gamer X Limp Gawd

    Messages:
    347
    Joined:
    Jul 5, 2019
    You are just white noise and I am putting you on ignore because you are unable to comprehend, nor willing to read people's full post. Please go back and read my last sentence in that post. You can't possible be this needy, can you?

    You didn't listen to the reasons Dr Su showed that chart. Because nodes are getting more expensive to produce, She is going to use Economy of Scale like Dr Su did in the CPU world, to gain market share in the GPU world. We will see 2 more 7nm GPU before Nvidia releases Ampere in 14 months time.



    Welcome to my ignore list, you can't even bother to read people's posts and argue with a mutual respect.

     
  33. ManofGod

    ManofGod [H]ardForum Junkie

    Messages:
    10,791
    Joined:
    Oct 4, 2007
    Navi and Vega dies are about the same cost? Have a proof of that, actual verifiable proof and not your opinion? Yeah, thought not.
     
  34. cybereality

    cybereality [H]ardness Supreme

    Messages:
    4,592
    Joined:
    Mar 22, 2008
    The key to all this is if AMD can figure out something clever for ray tracing.

    If, big IF here, they can find a way to make RT work on lower end cards (or mid/high end at reasonable prices), they will have a huge win.

    Let us be honest, Nvidia's mid-range is kind of useless for RT. I tried that Star Wars demo on a 2060 and it was slide show. Even with a 2080 Ti, you're talking about just passing performance at 1080p.

    So AMD has a huge opening here, but they only have maybe less than a year to pull it off. Nvidia is not going to sit still, but they are also invested in what they have built and might not want to throw it away.
     
    MangoSeed and Maddness like this.
  35. cybereality

    cybereality [H]ardness Supreme

    Messages:
    4,592
    Joined:
    Mar 22, 2008
    I don't know about the dies, but the overall cost of Navi must be cheaper if you look at the prices.

    Navi starts at $350, while Vega took at least a year to fall below $400 (inventory and mining could be the cause, but the prices were inflated for a long time).
     
    ManofGod likes this.
  36. MangoSeed

    MangoSeed Limp Gawd

    Messages:
    487
    Joined:
    Oct 15, 2014
    Yeah this is just the beginning. We don't know what tax RT will add to Navi if any and we don't know what performance will look like once AMD rolls out DXR support. It's strange that they're so quiet on the issue given all the hype.

    I don't see why Nvidia would be stuck with Turing-like RT cores. Games don't care about the hardware implementation as it's all sitting behind the "standard" DXR API. It's quite possible we will see architectures in the near future achieve decent RT performance without dedicated hardware.
     
    cybereality likes this.
  37. Snowdog

    Snowdog [H]ardForum Junkie

    Messages:
    9,216
    Joined:
    Apr 22, 2006
    The explanation is in the post if you can handle the math. Though it seems to challenge both you and GamerX.
     
    Dayaks likes this.
  38. ManofGod

    ManofGod [H]ardForum Junkie

    Messages:
    10,791
    Joined:
    Oct 4, 2007
    Your own personal math is not proof, I mean specifically something from AMD or TSMC?
     
  39. Snowdog

    Snowdog [H]ardForum Junkie

    Messages:
    9,216
    Joined:
    Apr 22, 2006
    The reason she posted the slide, doesn't alter the factual data of the slide. Factual data you keep ignoring in your small die rant.

    7nm chips cost double what 14nm class chips cost. So there is no financial advantage to a die half the size, when it costs twice as much to produce.


    Math isn't personal. It's universal.

    If you can't handle 2 x 1/2 = 1.

    You really can't contribute much meaningful to any discussion here.
     
  40. Ready4Dis

    Ready4Dis Limp Gawd

    Messages:
    501
    Joined:
    Nov 4, 2015
    It's the anti gamer-x, lol. Going on about how much Intel 10nm can pump out is a really funny line as th y even said they are having so many issues they're basically going to skip it for most things. First feed the troll, then give him this kind of crap to work with? Seriously, not quite, but almost as bad as gamer x just in the opposite direction. At least you had a few valid points intermixed.