Blade Stealth + Core + GTX 1080 benchmarks, first impressions

Discussion in 'Systems' started by vitorious, May 27, 2016.

Thread Status:
Not open for further replies.
  1. I'm curious why the Firestrike 1.1 scores are so low on the Core + GTX 1080. My MSI GT72 Dominator Pro w/ GTX980m with a +135 core and +500 memory overclock pull 9595

    http://www.3dmark.com/fs/7233475

    So seems like it possibly is limiting in some regard, maybe I should run Firestrike Ultra on my MSI GT72 to see what it pulls.
     
  2. ididntmemeto

    ididntmemeto Member

    If performance in games is really that bad I feel really ripped off.
     
  3. Most likely the tests that rely more on the GPU are comparable to a desktop (Firstrike Ultra) but others that rely more on CPU are lower in Firestrike 1.1 considering it's still running on a U series 15w chip on the stealth. The blade 2016 and MSI GT72 are using quad core HQ 45w chip in comparison, and desktop even higher again. The core can't overcome the CPU limits. It's still an awesome boost I think.
     
  4. KillerFry

    KillerFry New Member

    On my thread using benches from my desktop, the Blade 14 and the Stealth, and a 980 Ti that was exactly my conclusion.

    As the game requires more time to render in the GPU, both the CPU and TB3 bandwidth become less of a bottleneck.

    So, oddly enough, the higher in-game setting the less the performance gap is.

    Sent from my Nexus 6P using Tapatalk
     
    Lost Dreamer and Chrono like this.
  5. vitorious

    vitorious New Member

    With an external display, it's not that disparate, my 9388 to your 9595. Here's the comparison: http://www.3dmark.com/compare/fs/8602226/fs/7233475

    The GTX 1080 is getting way better graphics scores, but you're dominating on the CPU-heavy physics portion, maybe because you have more cores?

    The 7k Fire Strike score is using the laptop display. If others missed the revised benchmarks, they're:
    • Fire Strike Ultra (external monitor): 4310
    • Fire Strike Ultra (laptop display): 3936
    • Fire Strike (external monitor): 9388
    • Fire Strike (laptop display): 7623
    Details and links: http://vitor.io/razer-blade-stealth-core-gtx-1080-12h#3dmark

    As others here have said, anything that's GPU heavy should be much closer to desktop performance.
     
    Shr3ddermill3R and Lost Dreamer like this.
  6. Diblegs

    Diblegs Member

  7. vitorious

    vitorious New Member

    No idea, but there are practical reasons why you probably wouldn't want to do that:
    • If you plugged this into the Stealth the way it shows being plugged into a Macbook, you'd lose access to a USB 3 port, and the headphone jack on the Stealth.
      • Since USB-C ports are reversible, maybe you could plug this in backwards and not lose access to those other ports?
    • The USB-C port on the Stealth is also a Thunderbolt 3 port, and this says nothing about Thunderbolt 3 support.
    • The USB-C port on the Stealth charges at 45w, and this says nothing about supporting that much power (more power than the Macbook takes, and a Macbook charger can't charge the Stealth).
    • If you plugged this into the back of the Core, you'd have all that weight torquing that tiny little USB-C connector, and I'd bet it'd break something. Plus, again, it says nothing about Thunderbolt 3 support.
    I wouldn't personally spend my money on it, but if you try it, post a thread about it!
     
    Diblegs likes this.
  8. Diblegs

    Diblegs Member


    Yeah you are probably right concerning all your points especially the lack of Razer Core support concerns me.
    I will probably not buy this, maybe if Razer comes out with something I'll get it.
     
  9. vitorious

    vitorious New Member

    I own Doom now, and updated the page with the results, using default quality settings and vsync off.

    Using the Stealth's QHD LCD, I'm getting FPS in the 50s through the 70s. It never breaks 80 FPS in the fight I was testing it in, the bit after you get the yellow keycard and have to cycle the airlock. Quiet interiors were in the 90s.

    Using an external 2560x1440 LCD, I'm getting FPS in the 70s through the 90s. It breaks 100 FPS frequently, and quiet interiors were getting upwards of 140 FPS.
     
    BuckRogers, Lost Dreamer and Chrono like this.
  10. BuckRogers_no_id

    BuckRogers_no_id New Member

    Thanks for sharing this.

    What people don't know about Thunderbolt GPU cases is that the bandwidth is not a problem. Its been proven that the 40Gb/sec (bidirectional) bandwidth that it has doesn't bottleneck any GPU on the market and probably never will.

    The bandwidth, which is equivalent to PCIE 1.0 x16 or PCIE 3.0 x4 does do is limit your max FPS only. It doesn't affect minimum frames or frame latency. As long as you're above your chosen refresh rate, there's nothing to worry about. Running a GPU on PCIE 3.0 x4 in worst-case scenarios reduces your max FPS by 5% absolute tops. That's in old/lighter titles that go above ~200FPS. In a modern AAA title that you'd use a 1080 for, like GTA5 or VR HMD, it's a non-issue completely.

    If you're going to face any bottleneck that's worth worrying about, it's the CPU in the Stealth, not Thunderbolt. Pairing a Razer Core with Intel Skull Canyon NUC will remove any CPU bottleneck (and more since it has a 128MB CPU cache that few have).
     
    Last edited: Jul 20, 2016
    Eason85 likes this.
  11. Firebat246

    Firebat246 The One

    I have been trying to tell people that thunderbolt isn't a bottleneck. Thank you for posting this!
     
  12. ididntmemeto

    ididntmemeto Member

    With a Blade Stealth + GTX 980 I can barely get 60fps on medium to low settings with a 1440p external monitor on GTA V and Battlefield 4. I'm extremely extremely disappointed. I'm thinking about getting a Blade since it is quad core but at this point idk what i'm going to do.
     
  13. Buck if that's the case explain to me why my 1080 in a desktop configuration gets me 22000+ GPU score in 3dmark firestrike and 16900-17000 in the Core GPU score. Something is causing a bottleneck here, this is GPU dependent only scores not factoring CPU in the equations. Even with CPU's the i5-4690k stock to the i7-6700hq isn't a huge discrepancy.

    I feel like there is some overhead in Thunderbolt 3 and I'm not alone, until I can see a smaller gap I'm sticking to the notion that something is a bottleneck to the GPU over Thunderbolt 3. Heck look at the benchmarks for a GTX 970 (not a top-tier card) in the Core vs users putting the same card in a desktop netting much higher performance out of it.

    I can tell a noticeable difference in many of my VR games on Rift / Vive on the Core setup vs the desktop setup with a medium range i5. Even GPU specific things still are slower, someone help get to the bottom of why we are losing 10%+ performance with a Thunderbolt 3 eGPU if "bandwidth" limits are not the case.
     
    Last edited: Jul 20, 2016
  14. ididntmemeto

    ididntmemeto Member

    It could be drivers as well. I'd like to see some more comparisons of AMD cards (core vs desktop)
     
  15. I'm curious too haven't seen any Fury X / High-end AMD cards benched in these yet, I know AMD XConnect is the technology that eGPU over Thunderbolt 3 is built on so I'm curious if the loss is much lower on those AMD cards.
     
  16. ididntmemeto

    ididntmemeto Member

    I have a 290x I can try, I know it isn't the latest gen but I'm curious to see how it will perform vs the 980
     
  17. Curious if you have it in a desktop now to test something like Firestrike 1.1 and Ultra, then do the same test in the Core same drivers and see the differences in GPU scores. That's a quick and easy way to compare performance loss in the Core.
     
  18. ididntmemeto

    ididntmemeto Member

    yeah I can do that after work
     
    Shr3ddermill3R likes this.
  19. Ran 3dmark Timespy on my Blade 14 + Core + GTX 1080 FE and got a GPU score of 6047 http://www.3dmark.com/spy/87120

    Looking at various articles a 1070 clocks around 5630 in GPU score and a 1080 scores 6880 so we are 415~ish points ahead of a desktop 1070 setup but over 800 below a desktop 1080 setup.

    If there isn't some form of bottleneck call me crazy, 800 points is a pretty large gap. That's a 12% difference between the same card in a desktop, some form of bandwidth bottleneck is possibly happening. I don't know what kind of overhead protocol wise you have on Thunderbolt 3 / USB-C but if it's anything like iSCSI traffic over TCP/IP I assume loss will occur.
     
    Last edited: Jul 20, 2016
  20. Deathalo

    Deathalo Active Member

    Um, well there's the difference in CPU.... probably affects more than the TB3 throughput TBH
     
    BuckRogers likes this.
Thread Status:
Not open for further replies.
Sign In with Razer ID >


Don't have a Razer ID yet?
Get Razer ID >