MSI GTX 970 mini review with rFactor 2 performance

Discussion in 'General Discussion' started by DrR1pper, Sep 25, 2014.

  1. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,294
    Likes Received:
    36
    UPDATE: PLEASE READ

    rFactor 2 performance loss issue with my GTX 970 has now been sold. Please read the start of this thread to see why and how my problem may also be affecting you and if so, how you can fix it: http://isiforums.net/f/showthread.php/21983-Live-Performance-Benchmarking-Comparison-for-rFactor-2

    [HR][/HR]

    The MSI GTX 970 Gaming 4G card arrived in the post today and i've spent around 6 hours testing it and comparing it to my Palit GTX 770 Jetstream.

    [​IMG]

    I think they could have packaged it better tbh, just a little bubble rap around the box would have been nice.

    [​IMG]

    [​IMG]

    [​IMG]

    [​IMG]

    [​IMG]

    and apologies for the dust assault on your visual cortex in this next one..

    [​IMG]

    I bought this specific GTX 970 for 3 reasons.

    1) It was the second fastest and cheapest of the top tier GTX 970's on the market

    2) I would have bought the Gigabyte GTX 970 which is a tiny bit better but it was sold out everywhere.

    3) I would have bought the Gigabyte GTX 970 if only it would fit in my friggin case. It's 312mm in length, about 1cm too long. If i had realised this fact sooner, i would have stopped waiting for gigabyte stock and gone straight for the MSI a week earlier.


    SPECS

    Factory GTX 970 Specs:

    Core - 1050Mhz
    Boost Clock - 1178Mhz
    Mem - 7000Mhz

    MSI GTX 970 (factory overclocked) Specs:



    Core - 1140Mhz (+90Mhz)
    Boost Clock - 1279Mhz (+101Mhz)
    Mem - 7010Mhz (+10Mhz)

    First thing first, overclocking! It's a pretty decent overclocker. I couldn't get the memory above an extra +230Mhz before i noticed the slightest of artifacting in MSI Kombustor (and decided to fall back on +194Mhz), whilst of the 2 reviews i found for the same card they managed a stable +500Mhz. I was however able to match their core overclock of +185Mhz (very stable, was actually able to go to +225Mhz before fail). The core clock in load was 1501Mhz stable and memory 7388Mhz.

    MSI GTX 970 (additionally overclocked) Specs:

    Core - 1325Mhz (+185Mhz)
    Boost Clock - 1501Mhz (+222Mhz)
    Mem - 7388Mhz (+388mhz)

    Compared to the factory gtx 970, that's a 27% core clock increase and 5% memory overclock (in load).

    Lastly, the GTX 970's core clock is 21% higher than my GTX 770 whilst in game/benchmarking.


    BENCHMARKS - taken at overclocked settings

    System Spec:
    i5-2500k @ 4.3Ghz
    8GB DDR3 1600Mhz

    Sorry for the mess, i saved the scores for some whilst only noting the difference for others.

    3D Mark Vantage (gpu score only)

    GTX 770: 38,193
    GTX 970: 42,105 (+10% increase)

    3D Mark 11 (gpu score only)

    GTX 770: 11,451
    GTX 970: 15,443 (+35% increase)

    3D Mark 2014 (gpu score only)

    Firestrike - +50% increase over gtx 770

    Sky Diver - +55% increase over gtx 770

    Cloud Gate - +62% increase over gtx 770

    Ice Storm - +7% increase over gtx 770 (this is not a mistake. I tested many times and seems to apply to others too.)

    Unigine Heaven 4.0

    Settings: 1920x1080, 8xAA, Quality: High, Tessellation: Normal

    GTX 770: 48.7 fps average
    GTX 970: 72.0 fps average (+48% increase)


    Tomb Raider

    Settings: 1920x1080, mediums and highs

    GTX 770: 205.5 fps average
    GTX 970: 273.6 fps average (+33% increase)


    Now if i ignore the 3D Mark Vantage and Ice Storm test, it's a pretty decent improvement over my GTX 770. However, the next test was with rfactor 2....

    Using John's NEW Benchmark comparison between AMD & Nvidia benchmark....

    rFactor 2 replay benchmark

    GTX 770: 57.86 fps average
    GTX 970: 66.22 fps average (+15% increase)

    A measly 15% increase.

    I've tried to find something causing this result but can't seem to find anything at fault. A little depressing since when you look at all the reviews out there and you'll see the following sort of performance gains from a gtx 770 to gtx 970:

    BF4: +37%
    Bioshock Infinite: +35%
    Crysis 3: +24%
    Skyrim: +27%
    Unigen Valley: +27%

    On the outside, looking at mainstream games, this seems like a completely worthwhile upgrade. However when it comes to rfactor 2, i can't help but find the complete opposite is true.


    CONCLUSION

    It's pretty annoying and confusing at the same time since when i look at the spec difference between the two cards, i would have thought the GTX 970 would have sailed passed my GTX 770 in rfactor 2. Taking just the core clock and stream processor numbers into consideration, the GTX 970 is clocked 21% higher and has 8.3% more cores. When you combine them together, that's effectively 31% more gpu power in the GTX 970 vs my GTX 770 and yet it only delivers 15% more fps performance in rfactor 2, falling short of matching GTX 780Ti (stock clock) scores in rfactor 2 and yet matching them in other games and benchmarks.

    I'm starting to wonder if it has something to do with the other spec differences between the two cards that has something to do with this.

    Here are the differences in specs between the cards:

    Boost Clock: +20%
    Stream Processors: + 8.3%
    Tessellation units: +100% (i guess this on helps explain the stellar Unigine Heaven benchmark +48% increase results)
    Texture units: -19% (wondering if this could be the culprit)
    ROP's: +100%
    Transistors: +47%

    ...and that's the end of the mini review.

    [HR][/HR]

    UPDATE

    I've done a few different benchmark tests to compare the GTX 770 to the GTX 970, in both their nvidia rated stock clocks and the overclocks i was able to achieve on these cards. Each graph holds a different card as it's constant, representing the 100% performance reference point.

    New graphs including rf2 performance with the live benchmark method:

    GRAPHS DELETED - RF2 PERFORMANCE SHOWN IS WRONG. I WAS USING PCI-E 2.0 X16 AND NOT PCI-E 3.0 X16 WHICH YIELDS MUCH BETTER PERFORMANCE.

    UPDATED PERFORAMNCE GRAPHS WITH CORRECT RFACTOR 2 PERFORMANCE AT THE END


    The second one is what interests me most ofc and based on Sentri's stock GTX 970 performance, with my MSI overclock, i should be able to score about 84.2 fps in the civic + silverstone replay benchmark (a 46% increase over my Palit GTX 770 Jetstream).

    Now, regarding price-to-perofrmance's, for the £275 MSI GTX 970 card (with the easy overclocks), you're looking at about 60% more performance than a factory stock clocked GTX 770. To match it in price-to-perofrmance, you'd have to find a stock GTX 770 for £170 or my Palit GTX 770 Jetstream for £190. The cheapest i could find was an new MSI GTX 770 for £200 which performs roughly the same as my Palit. Ebay second hand GTX 770 cards however are selling at the same price-to-performance ratio.

    [HR][/HR]

    Corrrected performance graph with the GTX 970 in PCI-e 3.0 x16 mode should yield the following rf2 results:

    [​IMG]

    [​IMG]

    [​IMG]
     
    Last edited by a moderator: Oct 1, 2014
  2. matf1

    matf1 Registered

    Joined:
    Dec 27, 2011
    Messages:
    671
    Likes Received:
    2
    I have the same cpu and have often wondered if it's holding the sim back. What resolution are you running?
     
  3. ViSo

    ViSo Registered

    Joined:
    Feb 25, 2013
    Messages:
    316
    Likes Received:
    0
    What I think:
    GPU demand games and benchs gets high improvements.
    CPU demand games and benchs gets poor improvements.

    I am right? Is what Im supposing reading your post.
    However, I think that the GTX970 is only capable to match a GTX780 and not a GTX780ti, and that if you are lucky, I can be wrong ofcourse, is what I think, I didn't investigated so much.

    To see if the CPU is a bottleneck, check the GPU % load while playing, if it is 99% / 100% all the time, its ok, if it is changing all the times, then, there is a bottleneck, you must OC your CPU.
     
  4. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,294
    Likes Received:
    36
  5. Satangoss

    Satangoss Registered

    Joined:
    Jun 2, 2011
    Messages:
    1,123
    Likes Received:
    7
    I don't think 15% increase is measly. In other hand I would assume if you have a 6 core 3960X (for example) your gain in rF2 would increase than 15%.

    A got an i7 @ 4.8 Ghz and I fell it's suffering when I run rF2 with lots of car on the track.
     
  6. adaptable1

    adaptable1 Registered

    Joined:
    Feb 14, 2013
    Messages:
    112
    Likes Received:
    0
    Resolution is key... the new NVidia GPUs (970 and 980) supposedly yield higher frame rates at resolutions up to 2560 x 1440. At 4k (3840 x 2160) the 780/780ti have higher frame rates. I have yet to see a review where someone tests the new GPUs with NVidia surround (5780 x 1080) and sim racing titles. They do however consume less power.
     
    Last edited by a moderator: Sep 25, 2014
  7. Golanv

    Golanv Registered

    Joined:
    Nov 7, 2012
    Messages:
    1,041
    Likes Received:
    9
    I would like to see the difference between these cards with triples.
     
  8. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,294
    Likes Received:
    36
    It's measly compared to the 30-40% increase i can expect to observe in every other game as vetted by the other benchmarks i did.

    Sure, but i was benchmarking a replay. It should take the cpu out the equation when measuring performance.
     
  9. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,294
    Likes Received:
    36
    Based on the reviews, the MSI GTX 970 i bought is around 95% of the GTX 780Ti in games and lies between the 780 and 780Ti. With the additional overclock (extra 11% performance), it would surpass the stock GTX 780Ti to around 105% performance.

    edit: looking at more reviews, looks like i may be mistaken when adding AA into the equation. Results don't seem as consistent then. It's strange.
     
    Last edited by a moderator: Sep 25, 2014
  10. alpha-bravo

    alpha-bravo Registered

    Joined:
    Dec 22, 2012
    Messages:
    1,374
    Likes Received:
    31

    How many cpu cores ? Hyperthreading is on or off (not sure if your cpu support hv. It's to late here for a further research :p )
    25 % sounds like one core is under heavy workload.
    Check the workload per core with the ressourcmonitor maybe there is a tiny little chance for a cpu bottleneck.
    For rF2 IMO hyperthreading should be set to off.


    try aida64 maybe it can help you with some detailed info's or some special benchmarks to see how your system compare to others with similar components.

    http://www.aida64.com

    thanks for the review :) I changed from two gtx 670 to two r290x for a few months after I realized that the release of the gtx9xx series is to far away.
     
  11. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,294
    Likes Received:
    36
    i5 has no hyper-threading. i5-2500k has 4 cores, each o/c'd to 4.3ghz.

    25% utilisation average was also for each core. None of the core's were remotely close to 100%.
     
  12. Marek Lesniak

    Marek Lesniak Car Team Staff Member

    Joined:
    Oct 6, 2010
    Messages:
    1,585
    Likes Received:
    101
    Weird score but I got the same when I switched from 580 to 770 (in fact, the latter was performing a bit worse...).

    What is the most demanding track and car from ISI? Silverstone and Formula 2 or Marussia? Indy Car?
    Anyway, let's take such a combination on a one lap around the track solo (AI can drive) and then compare scores. CPU might have more significant effect in this case but at least we can check, if your gain is still so low vs 770. I will do the same with my overclocked 780.

    EDIT:
    - what build (860?),
    - what gfx drivers (and their settings),
    - what rF 2 settings (also rF Config).
     
    Last edited by a moderator: Sep 25, 2014
  13. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,294
    Likes Received:
    36
    Sorry, what do you mean LesiU. What score are you talking about and on which gpu and in rf2?

    If you look at this graph, 31% increase.

    [​IMG]

    edit: Quick google search shows around 50-60% increase in other games. WTH is going on?!?
     
  14. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,294
    Likes Received:
    36
    Oh wait, i'm such a numty.

    and i scored:

    Avg: 66.224 - Min: 47 - Max: 97

    So it would appear that the performance percentage ratio's seen more consistently higher in other games do not translate to rfactor 2. In fact the gtx 970 see's a 30-40% increase in fps over my gtx 770 in other games but only a 16% increase in rf2.

    edit: scratch that, this is getting really confusing. Someone posted another stock GTX 780 Ti score and it's much higher:

    The only real difference is the cpu speed which would otherwise suggest cpu bottleneck but going to look over this again.
     
    Last edited by a moderator: Sep 25, 2014
  15. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,294
    Likes Received:
    36
    I'm going to do some more testing today with different cpu clock speeds on both cards to see if it really is a cpu bottleneck issue or not (despite task manager showing only around 15-25% utilisation per core.)

    In the mean time, at least in crysis 3, demonstration that a GTX 970 slightly bests a GTX 780 Ti:



    and beating a R9 290X in BF4:



    http://www.eurogamer.net/articles/digitalfoundry-2014-nvidia-geforce-gtx-970-review

    I am perplexed as to why the performance gain ratio's are not remotely transfering onto rf2.
     
    Last edited by a moderator: Sep 25, 2014
  16. Marek Lesniak

    Marek Lesniak Car Team Staff Member

    Joined:
    Oct 6, 2010
    Messages:
    1,585
    Likes Received:
    101
    You still didn't say, what build, drivers and settings, so it's really hard to comment.

    You pointed to a newer benchmark. If you look at my thread, I made my oen comparison before our official benchmark and got different results.
    That's why I mention that - your case sounds familiar. We need to make more tests to check, if 970 is really that slow in rf2.
     
  17. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,294
    Likes Received:
    36
    I don't know of the thread you're talking about.

    Build: 860
    Benchmark with rf2 gfx settings and instructions are all in here: http://www.mediafire.com/download/iiawknhifwlwb1a/rF2_unofficial_benchmark.7z

    I have to go for a few hours now. Be back later. I really appreciate you lending a hand. ;)
     
  18. Ari Antero

    Ari Antero Registered

    Joined:
    Jul 27, 2012
    Messages:
    1,882
    Likes Received:
    829
    Rely, you tube is you tube and internet is internet You can read much information and watch videos that is not true, GTX 980 beats GTX 780Ti with few fps not GTX 970.

    GpuBoss claims here that GTX780 beats GTX870 but is it true?
    http://gpuboss.com/gpus/GeForce-GTX-970-vs-GeForce-GTX-780

    My advice is that if you need a new card go for GTX 970 but if you have GTX 760, 770 or 780 don`t do it, gain is not that great, with% yes but not in experience in the game.
     
  19. Marek Lesniak

    Marek Lesniak Car Team Staff Member

    Joined:
    Oct 6, 2010
    Messages:
    1,585
    Likes Received:
    101

    Was writing from my mobile phone...
    Here's a link to that thread (post #15 with my tests):
    http://isiforums.net/f/showthread.p...0-Titan-owners?p=186288&viewfull=1#post186288

    Notice, the difference was higher and sometimes it was lower, depending on tests.


    OK, you did your tests on B860. The benchmark you used, initially was done on B660 and on different drivers, so I wouldn't compare scores between them.
    Tell me, what gfx driver version you use for 770 and 970? Default settings?
     
  20. John.Persson

    John.Persson Registered

    Joined:
    Oct 17, 2011
    Messages:
    944
    Likes Received:
    13
    Well, this title is not very gpu scaling friendly.

    We known that for a long time! I don't see an i5 2500k OC to be a bottleneck, to be fair I don't even think it will be a bottleneck at stock either.

    We should get someone that has the insight on the engine to talk about the performance maybe.

    Tim, why don't the game scale as others?

    Is it becaused it's tide to the physics engine much more?
     

Share This Page