MSI X99A Gaming Pro Carbon Broadwell-E Motherboard @ [H]

Discussion in 'Intel MoBos' started by FrgMstr, May 31, 2016.

  1. FrgMstr

    FrgMstr Just Plain Mean Staff Member

    Messages:
    47,987
    Joined:
    May 18, 1997
    MSI X99A Gaming Pro Carbon Broadwell-E Motherboard - MSI’s X99 refresh is here in time for Intel’s new Broadwell-E processors! We put the new processor through its paces and our MSI X99A Gaming Pro Carbon was the lucky winner that we chose to torture in order to determine it’s "Broadwell-E readiness." Our time with the X99A Gaming Pro Carbon was filled with many surprises.
     
  2. Murzilka

    Murzilka Gawd

    Messages:
    552
    Joined:
    Dec 11, 2011
    Thanks for the review. Did you try disabling the HT and overclock the real cores only?
     
  3. TheLAWNoob

    TheLAWNoob Limp Gawd

    Messages:
    330
    Joined:
    Jan 10, 2016
    What cooler did you use when testing Boardwell-E? I couldn't find it, sorry if it's already stated in the article.
     
  4. cageymaru

    cageymaru [H]ard|News

    Messages:
    19,245
    Joined:
    Apr 10, 2003
    Oh I get it. So if you OC you're going to have the turbo off anyways so whichever core the new tech is targeting doesn't matter. Makes sense. I guess it would be a feature for a person running the processor at stock with the Turbo enabled.
     
  5. FrgMstr

    FrgMstr Just Plain Mean Staff Member

    Messages:
    47,987
    Joined:
    May 18, 1997
    Pretty much. We could have been cherry picking threads for a long time now if that was the way we wanted to OC.

    Sorry I do not have a mora3. And the fact of the matter is that this CPU is a worse overclocking than Haswell-E. Plain and simple.
     
  6. cageymaru

    cageymaru [H]ard|News

    Messages:
    19,245
    Joined:
    Apr 10, 2003
    So you'll made me Google that Mo-RA3 thing. Looks like a car radiator to me. Off to research car radiators as PC watercooling loops now. Thanks a lot guys says my wallet in a sarcastic tone. ;)
     
  7. FrgMstr

    FrgMstr Just Plain Mean Staff Member

    Messages:
    47,987
    Joined:
    May 18, 1997
    The Broadwell-E is for those that need MORE cores. Nothing more, nothing less.
     
  8. Dan_D

    Dan_D [H]ardOCP Motherboard Editor

    Messages:
    53,177
    Joined:
    Feb 9, 2002
    That thought didn't even occur to me. To expand on what Kyle said, I don't see the point. The reason you'd buy a 10c/20t CPU is for incredible multithreaded and multitasking performance. These are capable in games as the benchmarks show, but again those benchmarks are CPU limited. In the real world you won't be able to tell the difference between a Core i7 6950X and a Core i7 5960X or even a 5820K in the same system provided you have similar CPU clocks. Or as we've shown, you can be slightly behind on Broadwell-E.

    It does say in the article on the subsystem testing page. I used a Koolance Exos 2.5 with a Koolance CPU370 water block.

    We used the same testing hardware we've used on all the other X99 motherboards we've tested. That seems like the best way to go for comparable results.

    Good luck getting to 5.0GHz on a 6700K. I've yet to see any break 4.7-4.8GHz. You are also correct in that X58 will out perform Z170 in terms of memory bandwidth. It has three memory channels. As for single core performance, Broadwell-E should be faster than Gulftown by a fair margin. Roughly around 30% give or take. The last point I'll make here is that you aren't really sacrificing anything dropping to 8x8 for SLI over 16x16. There are a few instances here and there where you might see a little more performance on 16x16, but those are few and far between and that difference isn't exactly earth shattering.
     
  9. Murzilka

    Murzilka Gawd

    Messages:
    552
    Joined:
    Dec 11, 2011
    Thanks for chiming in Dan :)
    Here is what I meant:

    My Geekbecnh results:
    [​IMG]

    Other resource' (with all due respect) results:

    [​IMG]

    Notice the Single Core Memory Performance i7990x vs i76950x. 3138 vs 2847 points. respectively. Overall single core performance of the i7990x vs i76950x is like... the same? 3514 vs 3550. I also had 10+ tabs opened in the IE while running that test.




    Here is the multithreaded performance for a better understanding of the single vs multithreaded difference:
    [​IMG]
     
    Last edited: May 31, 2016
  10. Peter2k

    Peter2k Limp Gawd

    Messages:
    309
    Joined:
    Apr 25, 2016
    mmm
    I know a shop that can guarantee a 4.8 GHz overclock for 6700k, and 6600k
    For a price though

    I'm pondering if I really need the HT in an i7 and take a risk of not getting very high in overclocking
    Vs
    Taking the i5 with a guaranteed overclock, remove the lid, put some extreme paste on it, and put a AiO or custom loop on it

    The guaranteed oc i5 is ~ same price as an i7
    I only do gaming, and not knowing if and when games under dx12 can make GOOD use of 8 cores/threads makes me think I'd rather have an really high OC than fake cores

    sigh
    If there would be a socketed i7; skylake or kaby lake, with eDRAM I'd make the purchase in a heartbeat (with new generations improvement towards the eDRAM as a cache)
     
  11. Dan_D

    Dan_D [H]ardOCP Motherboard Editor

    Messages:
    53,177
    Joined:
    Feb 9, 2002
    I'm not a benchmark whore. So I don't keep up with all the latest benchmarks, but those provide a number, and as far as I know its an arbitrary one like 3D Marks. I despise benchmarks that do that because you don't know what criteria determines the results. Broadwell-E has a lot more memory bandwidth than Gulftown does. You are talking about 3x channels of DDR3 vs. 4x channels of DDR4. Those tests you posted also don't tell me if we are talking about DDR3 @ 1866MHz, or DDR4 @ 2133MHz or 2666MHz+. Also, I'm going to assume that the clock speeds of those CPUs are at their defaults and your system is overclocked. When we talk about CPU and RAM speeds, there is potentially a big difference in synthetic benchmarks as those frequencies climb. Beyond that, DDR3 has less latency than DDR4 does. The latter makes up for it with more bandwidth provided it's clocked high enough. In some situations though, memory bandwidth would come in secondary to latency. Those tests also shows Skylake trouncing Broadwell-E in memory performance which isn't right. Even then, the multicore results show a much larger gap vs. your system which is expected. In single core tests, clock speed is almost always king which might explain some of those results if my assertions about the test system's clock speeds are all correct.

    I hesitate to bring this up because comparing data from different OSes and benchmark versions can be misleading, but the good thing is that Sandra provides a real bandwidth number and not an arbitrary score that's not comparable to anything. You seem hung up on the memory bandwidth, which has been proven to be of little consequence in games (and even most applications) aside from Skylake, which seems to gain more from it due to some magic in it's architecture. Our Rampage III Black Edition review with a Core i7 990X shows memory bandwidth of 20.53GB/s. That's about half what we saw out of Broadwell-E at 40.31GB/s. Skylake systems can achieve around 26-32GB/s in the same test. That actually trumps Gulftown's triple channel DDR3 setup in most if not all cases. I mentioned earlier that X58 may outperform Z170 on memory bandwidth, which isn't true. (I haven't looked at the numbers in a long time, but the HEDT side has been WAY out front of the mainstream parts for memory bandwidth for some time.) Don't kid yourself, Broadwell-E is a monster compared to Gulftown. Having said that, like Sandy Bridge vs. Skylake, the differences are enumerated better in some tests than others. Also do not discount the fact that you are also getting a substantially more modern platform with X99 over X58.
     
    Last edited: May 31, 2016
  12. Sir_David

    Sir_David n00b

    Messages:
    18
    Joined:
    May 10, 2016
    Kyle, I would love for you guys to review the MSI X99A Godlike Gaming Carbon motherboard. That looks like one of the best gaming motherboards on the market now and may give better results with Broadwell-E.
     
  13. Murzilka

    Murzilka Gawd

    Messages:
    552
    Joined:
    Dec 11, 2011
    That comparison was about stock 6950X with turbo boost 3.0 and i7990x@4.66ghz and ram at 1886mhz 9-9-9-24 2t. Uncore at 3600-something, don't remember now. I found the tool to perfectly reflect what the system is capable of in games, that's why I am referencing to it. The numbers provided are real world performance level. I start sounding like I am promoting it... I am not. It's just extremely useful tool that's all I'm saying that shows real world performance numbers. Quickly and effectively.
    Crazy numbers of memory bandwidth is non important for games, what is important is latency that's where Broadwell lags behind Skylake and even the nicely tuned i7 990x Gulftown.
    IMHO. I just know that program is spot on :)
     
  14. EchoWars

    EchoWars Limp Gawd

    Messages:
    497
    Joined:
    Jan 7, 2010
    X99 is and will continue to be a trainwreck until some real effort is put into the BIOS of these things (we'll ignore the dozen or so hardware issues for now). With the dominance of Z170, that just might never happen.
     
  15. CrazyElf

    CrazyElf n00b

    Messages:
    21
    Joined:
    Feb 23, 2016
    Looks like these early BIOS revisions are immature. Expected, but I don't think that Broadwell E is looking like a good clocker.

    If people are hitting 4.3 - 4.4, that's down from 4.4 - 4.6 on Haswell E, which all but negatives the 3% or so you'd expect in IPC improvements on Broadwell. The only advantage I guess would be the lower power consumption (not worth buying a new CPU over) and perhaps for those with the money, the 10 core version.


    They already reviewed the Godlike and the Carbon is basically the same, except Black.

    Right now it will be immature as well. Give it a few months for the BIOS to mature and we will likely see better results across the board.
     
  16. Dan_D

    Dan_D [H]ardOCP Motherboard Editor

    Messages:
    53,177
    Joined:
    Feb 9, 2002
    How does a tool that reports a 3138 for memory performance telling you anything that's real world? 3138 what? I'm sorry but that's not something that's a real world number that I can see. It's not a latency number, nor a bandwidth number. That's like saying 35,000 3D Marks translates into something real world when it doesn't. It doesn't mean you'll get 35,000FPS in Doom does it? Certainly not. If this program is spot on, then maybe you are seeing something I don't. All I see in your images are arbitrary numbers that don't by themselves easily translate into anything useful. The website gives a baseline for what a given configuration scores so you can figure out what yours is relative to that, but what those baseline numbers actually mean and what they actually are is a mystery to me. Here is what the website says:

    "Geekbench Score The Geekbench score is the weighted arithmetic mean of the three section scores. The Geekbench score provides a way to quickly compare performance across different computers and different platforms without getting bogged down in details"

    On their website they promote it using real world tests, and maybe it does behind the scenes. However it's scoring system seems to be the same numbers game that 3D Mark uses. At least when Sandra spits out a memory result it tells it to me in bandwidth numbers and gives me a breakdown of what those numbers mean in addition to showing me what that score is relative to other configurations. I downloaded that tool to try and figure out more, but that's getting off track. I'm probably not going to say too much more about that benchmark itself. One parting thought on this is that if anything, your making the case for Broadwell-E being much faster than your ancient Core i7 990X. and interpreting the results incorrectly. You state:

    Essentially your conclusion is that the performance between your Core i7 990X is basically the same as Broadwell-E, more specifically the Core i7 6950X. Yet, you have over 1,000MHz clock speed advantage on Broadwell-E (more without turbo mode) and it still keeps up. When you clock Broadwell-E at 4.3GHz, it isn't just going to rape your i7 990X. its going to fill your 990X's crevasse with sand, pull out a strap on and pound it like a jackhammer.
     
  17. Dan_D

    Dan_D [H]ardOCP Motherboard Editor

    Messages:
    53,177
    Joined:
    Feb 9, 2002
    I'd would love to look at the Carbon version, but I don't think it's different enough to warrant it. Certainly if Kyle sends me one I'll place it on the bench and go a few rounds with it. My guess is that the Carbon version is probably as good as the GODLIKE was, but costs too damn much money for most people to seriously consider it. Although I guess if your willing to pay for an i7 6950X, you might just go for it.
     
  18. Murzilka

    Murzilka Gawd

    Messages:
    552
    Joined:
    Dec 11, 2011
    Thanks, that's reassuring ))))) I meerly suggest this should be done by default with no need to overclock anything. There is 5 years difference between the cpus.

    Concerning the program - its numbers it provides reflect real world gaming performance. By that I mean - any change in single core department is reflected in games performance. Small increase in single core performance per geekvench translates into small increase in game performance which it you can clearly see as you play the game. That is the matter of practice, there is no one-size-fits-all equation obviously to translate it into fps.
    I certainly have troubles with expressing myself, but that point is pretty clear I suppose? :)

    In any case, I am not trying to prove you wrong or that my 990x beats the 6950x :). Just stating my opinion on this 1700-dollar cpu. That it is simply not good enough for its money.
     
  19. Dan_D

    Dan_D [H]ardOCP Motherboard Editor

    Messages:
    53,177
    Joined:
    Feb 9, 2002
    Your original assertion is that the performance between your Core i7 990X @ 4.46GHz was the same as the Core i7 6950X and it is similar. Your talking about a CPU that's got 1GHz+ on it. You weren't comparing apples to apples, but oranges to cherry's. IPC and general architectural improvements just can't close that gap. However, if you take Broadwell-E and overclock it your scores would be crushed outright by Broadwell-E.

    As for the CPU, we haven't done a formal review of that and normally that's Kyle's area. I haven't had near the time with the CPU he has but I find it lackluster to say the least. At 4.3GHz it achieves rough parity with the 5960X at 4.5GHz. There is some IPC improvement, but not enough to make up for the clock speed loss compared to Haswell-E. Getting it to 4.3GHz and stable using AVX instructions is far more difficult than it is with Haswell-E. It's definitely faster in multithreaded / multicore capable workloads that can scale well enough with an appropriate number of threads. There is no question of that, but the price tag for the i7 6950X is utterly ridiculous. It might be a win for content creators who have large budgets to work with, or may get them on the company's dime, but outside of that I see little to no value in this processor. Halo level products are never value propositions, but this one offers nothing for gamers and overclockers. You would be better served with the Core i7 5960X @ 4.5GHz or the 6900K@ 4.3GHz (it may clock better than the 10-core variant) due to more reasonable costs and similar performance. As far as I'm this processor launch is definitely a dud. It's like comparing Sandy Bridge and Ivy Bridge, but Ivy Bridge costing almost double the price of Sandy Bridge.

    That's retarded in my opinion. Intel's pulled crap like this with enthusiasts before. Skulltrail was arguably worse since the true intended configuration of the platform was dual QX9775's and the D5400XS motherboard. The processors alone were $1,550 and you needed two of them. So that's much more costly than an i7 6950X is. Having said that, Skulltrail offered a lot more because you got to double your core count and those CPUs used a motherboard that was infinitely better than most of what was available at the time in most areas. It also brought SLI to the Intel chipset so you didn't have to deal with the 680i SLI nonsense. So despite the stratospheric price tag Skulltrail was at least likable. The i7 6950X is barely a step forward thanks to having two additional (although useless for gaming) cores and yet, it's definitely a step back in clock speeds. I was hoping that the memory controller would be greatly improved over the one in Haswell-E, similar to what we saw with Skylake but that didn't happen either.

    I understand Intel driving efficiency with mobile and server markets having priority over desktop platforms, but the efficiency increases are so small that they are difficult to appreciate in light of what's lost to get achieve it. As far as desktop processors go, Broadwell-E just isn't all that compelling. Aside from motherboard chipsets getting feature updates every generation or so we've stagnated on the CPU side of things as far as the desktop market goes. With the right GPU, Sandy Bridge is still trucking along just fine as is Gulftown. Hell there are still people rocking Nahalem based machines and remain content.
     
  20. Murzilka

    Murzilka Gawd

    Messages:
    552
    Joined:
    Dec 11, 2011
    Well, yeah my original point was that at stock, 6950X Single Core memory performance is lower than that of my 5 years old 990x, granted it is heftly overclocked. Not something that one would expect from a top dog 2016 extreme cpu. Knowing also that a 5 times less expensive i7 6700 shreads my 990x to pieces in that same domain be it at stock or not. It could be at least faster...

    I have 2 titans x in my system and I notice they do not receive the full load in some situations. Overall games run smoothly on my system but there are some stuttering here and there due to the aging platform that can't provide full load to the gpus at 1440h 100+hz. Idk maybe I should wait til the next year and get the real big update instead of this heavy duty non-gaming oriented and also aging B-E.

    I fully agree with your post. I think I remember that Skulltrail thing :D And certainly the 680sli :D
     
  21. GothamsReckoning

    GothamsReckoning Limp Gawd

    Messages:
    243
    Joined:
    Dec 22, 2015
    Not to mention Dan that Skulltrail also brought out the first Nf100 appearance in an Intel chipset. Until the D5400XS came out the Nf did not generally appear if at all in Intel chipsets and Nvidia preferred to license/allow SLI on nForce chipsets only.
     
  22. Dan_D

    Dan_D [H]ardOCP Motherboard Editor

    Messages:
    53,177
    Joined:
    Feb 9, 2002
    Right, the point being that even though we are comparing two halo type products, the last stupid high priced thing they had was worth more than this one is to the gamer. Alternatively, you could grab a D5400XS and some Xeons to do the same thing at a lower price. Still expensive, but not necessarily as much as this CPU. It also constituted more of the total system components than this does when you were finished. All I'm saying, is that Broadwell-E, at least in regard to the 6950X has nothing to offer gamers at it's current price point.
     
  23. GothamsReckoning

    GothamsReckoning Limp Gawd

    Messages:
    243
    Joined:
    Dec 22, 2015
    I agree and from what I remember LGA771 Xeons were sought as alternatives to the QX9775 which ended up being much cheaper overall and cannibalizing Intel's hope for their Skulltrail platform (especially since the QX9775 was just the LGA711 equivalent of the QX9770 and offered no real advantages). Looks like you mentioned the Xeons as well.

    Also from your opinion and looking at the Broadwell-E lineup would you get a Broadwell-E CPU such as the 6900K or buy a used 5930K/5960X if you were to build a X99 system currently. That's the dilemma I am facing right now as the retail price for the V3/V4 Xeons is astronomical in Canadian dollars. I got the motherboard for a great price (it's a 1st gen Gigabyte motherboard for X99) but am looking to pair it with an older SLI/Tri-SLI setup and would prefer to get the 40 lanes if possible (unless there is a small price difference between the 5820K and the 5930K).
     
  24. Dan_D

    Dan_D [H]ardOCP Motherboard Editor

    Messages:
    53,177
    Joined:
    Feb 9, 2002
    I would probably try and find a good deal on a used 5960X or even a close out deal at Microcenter on a 5930K.
     
  25. heatlesssun

    heatlesssun [H]ard as it Gets

    Messages:
    44,157
    Joined:
    Nov 5, 2005
  26. Dan_D

    Dan_D [H]ardOCP Motherboard Editor

    Messages:
    53,177
    Joined:
    Feb 9, 2002
  27. heatlesssun

    heatlesssun [H]ard as it Gets

    Messages:
    44,157
    Joined:
    Nov 5, 2005
    Thanks, I got this confused with the Godlike version of this. I'm putting together a high-end Broadwell-E machine and the motherboard is the last thing I'm trying to decide on.
     
  28. Dan_D

    Dan_D [H]ardOCP Motherboard Editor

    Messages:
    53,177
    Joined:
    Feb 9, 2002
    Unfortunately, I don't have much to say on the subject. The GIGABYTE X99P-SLI and X99A Gaming Pro Carbon are the first X99 refresh boards I've looked at. I've only tested the X99A Gaming Pro Carbon with Broadwell-E, so I have little advice to give. On paper, the X99A GODLIKE Carbon will be similar to the original board as is the X99 Deluxe II to it's predecessor. Whether or not the changes are worth having or how well Broadwell-E behaves on older X99 motherboards is unknown to me at this time.
     
  29. heatlesssun

    heatlesssun [H]ard as it Gets

    Messages:
    44,157
    Joined:
    Nov 5, 2005
    Again, thanks for the feed back. I'm thinking I'll go with the Asus as it seems like a safe bet.
     
  30. Marcdaddy

    Marcdaddy 2[H]4U

    Messages:
    3,627
    Joined:
    Feb 21, 2003
    Oh man, I Just went with a full Build and ordered this Motherboard from Newegg and now I'm regretting it. Not sure if I should leave it closed and send it back and pick another one up locally ( Asus ) id hate to run into these problems next week when it gets here. I have the 2 1080s in sli in my 2600k setup right now. Here is the build, I have a INTEL 6850K on my desk and a 6900K OTW from Newegg too.
    [​IMG]
    • [​IMG]
    • [​IMG] Loading...
     
  31. heatlesssun

    heatlesssun [H]ard as it Gets

    Messages:
    44,157
    Joined:
    Nov 5, 2005
    Looks sweet! I'm building a new rig myself along similar lines, just got my Gigabyte FE 1080s yesterday, and wondering about motherboards myself. I think I'm going with the ASUS X99-DELUXE II.
     
    Marcdaddy likes this.
  32. Marcdaddy

    Marcdaddy 2[H]4U

    Messages:
    3,627
    Joined:
    Feb 21, 2003
    I'll report back, I'd hate to not try to get it to work since Kyle go it working. I live 25 minutes from a Microcenter and Frys so worst case scenario I might go grab the Asus if it acts up.
     
  33. Betauser

    Betauser 2[H]4U

    Messages:
    3,718
    Joined:
    Jul 14, 2003
    Can you manually OC this motherboard or do you have to use the MSI software OC feature? I read somewhere that the manual OC is not available at BIOS. Marcdaddy, keep us posted as I am also looking for X99 board where I can use M.2 and not hinder a 2 way SLI or 3 sli in the future.
     
    Marcdaddy likes this.
  34. Marcdaddy

    Marcdaddy 2[H]4U

    Messages:
    3,627
    Joined:
    Feb 21, 2003
    It will be here tommorrow but my 6850k says thursday/friday. I'll definitely be posting a review.
     
    FrgMstr likes this.
  35. Marcdaddy

    Marcdaddy 2[H]4U

    Messages:
    3,627
    Joined:
    Feb 21, 2003
    Guts it's alive!!!!! Installing windows now, so far zero problems. I'd like to report it has zero problems with my GSKILL 3400mhz ram, Xmp loaded it up. Messing around with the turbo dial I'm at 430o mhz installing windows. I have a big Pinball event downtown Chicago tonight ( i ordered a Stern Ghostbusters Premium pinball machine and I'm getting to go check one put tonight ) then I'm on vacation so I'll report back.
     
  36. Marcdaddy

    Marcdaddy 2[H]4U

    Messages:
    3,627
    Joined:
    Feb 21, 2003
    Everything is here and it's up and running, recognized my GSKILL memory at 3400 mhz no problems.
     
    Betauser and FrgMstr like this.
  37. Betauser

    Betauser 2[H]4U

    Messages:
    3,718
    Joined:
    Jul 14, 2003
    How's the Overclock mechanism? Can you do manual oclock in BIOS?
     
  38. Marcdaddy

    Marcdaddy 2[H]4U

    Messages:
    3,627
    Joined:
    Feb 21, 2003
    Yes you can, thete is a ton of options. Unfortunately I'm on my way to a previous engagement and will spend tommorrow doing a bunch of benchmarking and overclocking.
     
  39. Dan_D

    Dan_D [H]ardOCP Motherboard Editor

    Messages:
    53,177
    Joined:
    Feb 9, 2002
    I'm not sure where you would have heard that. I have never seen an X99 motherboard that didn't have options for overclocking in the BIOS.
     
  40. Betauser

    Betauser 2[H]4U

    Messages:
    3,718
    Joined:
    Jul 14, 2003
    I think I read it on one of the newegg reviews which made me think if that's possible, that's why I asked. Been out of the CPU overclocking so who knows what changed.