EVGA

Igors Lab: Less MLCC, more crashing and lower clocks?

Page: 12345.. > >> Showing page 1 of 6
Author
Lokator
Superclocked Member
  • Total Posts : 147
  • Reward points : 0
  • Joined: 2018/03/13 18:25:57
  • Status: offline
  • Ribbons : 1
2020/09/25 06:38:03 (permalink)
https://www.igorslab.de/en/what-real-what-can-be-investigative-within-the-crashes-and-instabilities-of-the-force-rtx-3080-andrtx-3090/

So Igor means, Zotac cards crashing have 6x POSCAP and 0x MLCC.
He means less MLCC and more POSCAP behind chip ---> more crashing

FE has 4x POSCAP and 2x MLCC.

Asus Strix has 6x MLCC.
I saw pics with Tuf and 6x POSCAP, but Igor says even Tuf has 6x MLCC now.

He says that bord partners had no drivers to test so most of them are slapping more MLCC in now.

Looked at Pic of XC3 and 1x MLCC, 5x POSCAP.

At pic of FTW3 Ultra it has 0x MLCC and 6x POSCAP.

Igor means 6x MLCC brings potential higher stable Mhz on the chip.

If FTW3 really has 6x POSCAP and Asus Strix 6x MLCC, I really consider waiting for the Strix.

Any Info if FTW3 really has 0x MLCC, 2x less than FE and 6x less than Strix?

Will EVGA change it if true?
post edited by Lokator - 2020/09/25 06:44:05

Attached Image(s)

#1

179 Replies Related Threads

    Lokator
    Superclocked Member
    • Total Posts : 147
    • Reward points : 0
    • Joined: 2018/03/13 18:25:57
    • Status: offline
    • Ribbons : 1
    Re: Igors Lab: Less MLCC, less crashing? 2020/09/25 06:40:16 (permalink)
    https://reddit.com/r/nvidia/comments/izhpvs/the_possible_reason_for_crashes_and_instabilities/

    now one Strix spotted with 2 MLCC and one with 6. Seems like a lottery now?

    Thanks Nvidia!
    post edited by Lokator - 2020/09/25 06:49:09
    #2
    HawkOculus
    iCX Member
    • Total Posts : 456
    • Reward points : 0
    • Joined: 2019/04/10 10:50:51
    • Status: offline
    • Ribbons : 1
    Re: Igors Lab: Less MLCC, less crashing? 2020/09/25 07:16:09 (permalink)
    I also posted a thread about this. Seems like everyone is keeping quiet about it. It’s a big issue especially if an “OC” card has cheaper components on it that will just make it crash when you try to overclock it.

    I’ve been trying to get folks on this forum reporting problems to post images of the back of their cards to see what the deal is.

    All I can say otherwise is that I’m glad I didn’t get a day one card now...
    #3
    Lokator
    Superclocked Member
    • Total Posts : 147
    • Reward points : 0
    • Joined: 2018/03/13 18:25:57
    • Status: offline
    • Ribbons : 1
    Re: Igors Lab: Less MLCC, less crashing? 2020/09/25 07:32:43 (permalink)
    I am mad I just sold my 2080 Ti on ebay.... with this knowledge a week ago I would have sit that out and enjoy my 2080 ti... if I get the 3080 FTW3 via amazon soon and it has low amount or no mlcc, I might consider sending it back after 30 days
    #4
    vgerik1234
    iCX Member
    • Total Posts : 301
    • Reward points : 0
    • Joined: 2007/02/03 11:59:20
    • Status: offline
    • Ribbons : 1
    Re: Igors Lab: Less MLCC, less crashing? 2020/09/25 07:34:30 (permalink)
    First off, that was a really educating read, thank you for the igor's page. 2nd, someone mentioned (whether on forums or GN), Nvidia supplies some of the components as well as the chip itself. So we know for a fact NVIDIA supplies the chip + 6X memory modules. Its possible they also are the suppliers of the MLCC/POSCAP. If this is the case, the theory that at some point NVIDIA recognized a problem and began supplying some MLCC but not all doesn't really make sense. Why would some cards get 6 and some get 2 from the same line? Maybe there was supply line issues for acquiring MLCC and they just had to deal with what was given? I think to meet manufacturing demands people placed orders for MLCC and just rationed it out accordingly. When wave 2 hits, we definitely will see the full picture.

    Either way it is 100% something to look out for. Luckily some of the cards have an open back design where you can just check right away with what you were given. 



     
    #5
    jpotter
    New Member
    • Total Posts : 74
    • Reward points : 0
    • Joined: 2008/07/01 19:15:56
    • Status: offline
    • Ribbons : 0
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 07:38:18 (permalink)
    Lokator

    At pic of FTW3 Ultra it has 0x MLCC and 6x POSCAP.

    If FTW3 really has 6x POSCAP and Asus Strix 6x MLCC, I really consider waiting for the Strix.

    Any Info if FTW3 really has 0x MLCC, 2x less than FE and 6x less than Strix?

    Will EVGA change it if true?



    This needs do be answered. I love EVGA, but I'll be pissed if I pay $810+ for a card that can't hold stable overclocks. Even if just a little. 
     
    Which phrase sounds better to you guys, CapacitorGate or just MemoryGate? We have to get it right the first time and stick with it.
    #6
    XhanXaijin
    New Member
    • Total Posts : 31
    • Reward points : 0
    • Joined: 2009/07/24 17:34:52
    • Status: offline
    • Ribbons : 0
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 07:42:01 (permalink)
    jpotter
    Lokator

    At pic of FTW3 Ultra it has 0x MLCC and 6x POSCAP.

    If FTW3 really has 6x POSCAP and Asus Strix 6x MLCC, I really consider waiting for the Strix.

    Any Info if FTW3 really has 0x MLCC, 2x less than FE and 6x less than Strix?

    Will EVGA change it if true?



    This needs do be answered. I love EVGA, but I'll be pissed if I pay $810+ for a card that can't hold stable overclocks. Even if just a little. 
     
    Which phrase sounds better to you guys, CapacitorGate or just MemoryGate? We have to get it right the first time and stick with it.





    That would be capgate.
    Memgate will be when nvidia starts using 23gb/s mem modules in december to replace 21gb/s and will start charging an extra 100 per card while they're paying the same price for either module.
    #7
    gravedigger78
    SSC Member
    • Total Posts : 964
    • Reward points : 0
    • Joined: 2010/05/21 13:26:03
    • Location: Michigan
    • Status: offline
    • Ribbons : 2
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 07:42:05 (permalink)
    jpotter
    Lokator

    At pic of FTW3 Ultra it has 0x MLCC and 6x POSCAP.

    If FTW3 really has 6x POSCAP and Asus Strix 6x MLCC, I really consider waiting for the Strix.

    Any Info if FTW3 really has 0x MLCC, 2x less than FE and 6x less than Strix?

    Will EVGA change it if true?



    This needs do be answered. I love EVGA, but I'll be pissed if I pay $810+ for a card that can't hold stable overclocks. Even if just a little. 
     
    Which phrase sounds better to you guys, CapacitorGate or just MemoryGate? We have to get it right the first time and stick with it.


    How about the I successfully checked out and still didn't get one... gate
     
    If there is any glaring defect EVGA will take care of you. They always have gone above and beyond.
    #8
    Lokator
    Superclocked Member
    • Total Posts : 147
    • Reward points : 0
    • Joined: 2018/03/13 18:25:57
    • Status: offline
    • Ribbons : 1
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 07:50:18 (permalink)
    yes but if I would get a higher OC with more mlcc, I will choose, if available, the strix. Has already insane out of the box clock speeds.

    also to the memory: 384 bit 12 gb gddr6 with way less power consumption and similar memory bandwidth might have been not the worst idea for the 3080.
    #9
    roccale
    New Member
    • Total Posts : 19
    • Reward points : 0
    • Joined: 2017/06/21 06:42:30
    • Status: offline
    • Ribbons : 0
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 07:54:51 (permalink)
    EVGA mount 6* popcaps so if the problem is true evga have a problem... low q.
     

    #10
    Frammish
    Superclocked Member
    • Total Posts : 236
    • Reward points : 0
    • Joined: 2014/10/18 15:24:44
    • Status: offline
    • Ribbons : 0
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 08:18:27 (permalink)
    @roccale, that product image may be accurate, or it may be old. Early samples would be used for photos. Maybe even non-working cards. We unfortunately don’t know until we see photos of the cards people are getting. Obviously that photo was representative at some point in time but may not be now. EVGA needs to post updated product photos if it’s not accurate.
     
    @Lokator, unfortunately this is only a guess, but Igor indicates that manufacturers started lines before this behavior was known or apparently well tested since Nvidia didn’t have a driver until recently. Could be there are unintentionally early and late versions of these cards. It’s also worth noting, though could be totally bogus since it seems to never happened, but Microcenter managers reportedly thought they had to send all their 3090 XC3 stock back and were telling people in line last Wednesday that EVGA cards would not be available. Need to emphasize that all I saw were scattered reports of this without much else to back it up. But it would also fit with some kind of production issue that required hands-on with a card to fix.
     
    @vgerik1234, supplying the capacitors themselves is unlikely but who knows these days? Caps are commodity items and would probably be something AIB companies would source themselves with the rest of the parts. The memory is funky and a new design that uses intermediate voltage levels to represent the equivalent of two bits with just one. It effectively makes the memory bus twice as wide give or take (due to speed differences). The memory is much more specialized and apparently designed alongside the GPU so fits with Nvidia supplying it (though it’s made by Micron, IIRC).
     
    It will be interesting to see how this all plays out... sounds like worst case affected cards don’t clock quite as high is all.
    #11
    pointofdeparture
    New Member
    • Total Posts : 12
    • Reward points : 0
    • Joined: 2019/11/28 07:44:32
    • Status: offline
    • Ribbons : 0
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 08:21:54 (permalink)
    I tweeted @EVGA_JacobF about this, I imagine he will address it one way or another with how much attention news this is getting today.

    ~Desmond
    Current build: AMD Ryzen 3600, ASUS PRIME X570-Pro, EVGA GTX 1080 Ti SC2, G.Skill 2x16GB 3600C16 DDR4, Samsung 850 Evo 250GB SATA (boot), Sabrent Rocket 1TB NVMe (games/work), Corsair RM750x, Scythe Mugen 5 Rev. B, Phanteks Enthoo Pro M, LG 27GL83A-B
    #12
    roccale
    New Member
    • Total Posts : 19
    • Reward points : 0
    • Joined: 2017/06/21 06:42:30
    • Status: offline
    • Ribbons : 0
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 08:31:13 (permalink)
    Steve sample...6 poscaps 
     

    #13
    blazemore
    New Member
    • Total Posts : 30
    • Reward points : 0
    • Joined: 2008/05/03 14:44:40
    • Location: NC
    • Status: offline
    • Ribbons : 0
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 08:33:47 (permalink)
    Nothing to see here...my mistake. 
    post edited by blazemore - 2020/09/25 08:41:41

    ASUS Strix 570-E l 5600x with XSPC Raystorm l 64GB Ripjaws l EVGA 3080 FTW Ultra with EK WB l 1300w G2 
    #14
    Lokator
    Superclocked Member
    • Total Posts : 147
    • Reward points : 0
    • Joined: 2018/03/13 18:25:57
    • Status: offline
    • Ribbons : 1
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 08:38:10 (permalink)
    blazemore
    roccale
    EVGA mount 6* popcaps so if the problem is true evga have a problem... low q.
     





    Changes have been made. 



    thats a 3090.
    #15
    MiRai
    Superclocked Member
    • Total Posts : 112
    • Reward points : 0
    • Joined: 2010/12/15 22:52:06
    • Status: offline
    • Ribbons : 0
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 08:40:38 (permalink)
    There's a review up of a 3090 FTW3 Ultra that shows 2x MLCCs and 4x POSCAPs, which differs from the original stock photo shown on sites.
     
    Image (Source)
    EVGA Store Stock Image
    Amazon Stock Image
     
    EDIT: The nVidia sub-reddit is putting together a table for PCB designs that can be currently found around The Internet.
    post edited by MiRai - 2020/09/25 08:52:03
    #16
    HawkOculus
    iCX Member
    • Total Posts : 456
    • Reward points : 0
    • Joined: 2019/04/10 10:50:51
    • Status: offline
    • Ribbons : 1
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 08:42:06 (permalink)
    MiRai
    There's a review up of a 3090 FTW3 Ultra that shows 2x MLCCs and 4x POSCAPs, which differs from the original stock photo shown on sites.
     
    Image (Source)
    EVGA Store Stock Image
    Amazon Stock Image


    Great detective work. It seems likely that this is why stock is so low in general. They probably had to scramble at the last minute to change the designs on most of their cards.
    #17
    pointofdeparture
    New Member
    • Total Posts : 12
    • Reward points : 0
    • Joined: 2019/11/28 07:44:32
    • Status: offline
    • Ribbons : 0
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 08:42:29 (permalink)
    JayzTwoCents' 3080 XC3 review sample looks like 1xMLCC and 5xPOSCAP.
     


    ~Desmond
    Current build: AMD Ryzen 3600, ASUS PRIME X570-Pro, EVGA GTX 1080 Ti SC2, G.Skill 2x16GB 3600C16 DDR4, Samsung 850 Evo 250GB SATA (boot), Sabrent Rocket 1TB NVMe (games/work), Corsair RM750x, Scythe Mugen 5 Rev. B, Phanteks Enthoo Pro M, LG 27GL83A-B
    #18
    Lokator
    Superclocked Member
    • Total Posts : 147
    • Reward points : 0
    • Joined: 2018/03/13 18:25:57
    • Status: offline
    • Ribbons : 1
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 08:49:52 (permalink)
    Nvidia had most time optomising the cards, so 2x mlcc from FE should be fine. Still Asus' 6x mlcc very impressive. This case needs more investigation but for me this launch is a disaster. Controverse Samsung 8n, powerhungry board with gddr6x, memory size discussion, paper launch, broken cards with cheap capacitors, reviews same time with launch, bots buying all cards and selling them on ebay, gg
    #19
    MiRai
    Superclocked Member
    • Total Posts : 112
    • Reward points : 0
    • Joined: 2010/12/15 22:52:06
    • Status: offline
    • Ribbons : 0
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 08:50:18 (permalink)
    I added this to my post up above, but the nVidia sub-reddit is putting together a table for PCB designs that can be currently found around The Internet.
    #20
    Dakutagawa
    New Member
    • Total Posts : 87
    • Reward points : 0
    • Joined: 2014/07/16 15:16:00
    • Location: Nor-Cal
    • Status: offline
    • Ribbons : 1
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 08:50:27 (permalink)
    Techpowerup.com review of the Asus tuf



    https://www.techpowerup.c...0-tuf-gaming-oc/3.html
    post edited by Dakutagawa - 2020/09/25 08:54:26

    Attached Image(s)


    i7-4790k, 4.8GHz @1.28v  |  ASRock Z87 Extreme 4
    2x8GB Mushkin Blackline DDR3-2133MHz C9 @1.65v
    1TB Samsung 970 Evo NVME  |  2TB WD Caviar Black
    EVGA RTX3080 FTW3 Ultra Hybrid

    Acer Predator X34 3440x1440@120hz
    Seasonic Prime TX-850  |  Corsair 5000D Airflow

     

    #21
    Frammish
    Superclocked Member
    • Total Posts : 236
    • Reward points : 0
    • Joined: 2014/10/18 15:24:44
    • Status: offline
    • Ribbons : 0
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 08:57:46 (permalink)
    All I know is I won’t be buying any 3000 series card until this is settled.
     
    Let the scalpers and their bots eat the inventory they so greedily scarfed up. Paybacks. 😁
    #22
    TheGuz4L
    Superclocked Member
    • Total Posts : 130
    • Reward points : 0
    • Joined: 2016/06/09 09:24:29
    • Status: offline
    • Ribbons : 0
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 09:04:32 (permalink)
    This is crazy.  How can we be sure this is definitely causing the crashing?
    #23
    SprayingMango
    iCX Member
    • Total Posts : 273
    • Reward points : 0
    • Joined: 2006/03/10 21:45:38
    • Status: offline
    • Ribbons : 0
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 09:07:04 (permalink)
    Wow this is nuts!! Just knowing you have a time bomb of a card on your hands is super stressful. If it dies, you most likely would not get a replacement for a long time. I'm actually incredibly relieved I did not get a card now!! LOL

    Affiliate Code = I6E6SHELAX
    12900K
    ASUS Z690 Strix E
    32GB DDR5 Corsair Dominator 5600
    Samsung 970 Evo Plus 2TB
    WD Black 2TB PCIe 4 NVME
    EVGA 3090Ti FTW3 Ultra
    Corsair AX1600i PSU
    Corsair 1000D 
    Corsair QL120 fans x 15
    Corsair H150i Pro XT
    Asus PG27UQ w/ PG279Q x2
    #24
    Lokator
    Superclocked Member
    • Total Posts : 147
    • Reward points : 0
    • Joined: 2018/03/13 18:25:57
    • Status: offline
    • Ribbons : 1
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 09:08:20 (permalink)
    need more investigation, but I feel a correlation being possible, what Igor says and writes does make sense.
    #25
    Lokator
    Superclocked Member
    • Total Posts : 147
    • Reward points : 0
    • Joined: 2018/03/13 18:25:57
    • Status: offline
    • Ribbons : 1
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 09:09:30 (permalink)
    Usually they have enough RMA cards stored for quick exchange.
    #26
    ehabash1
    iCX Member
    • Total Posts : 463
    • Reward points : 0
    • Joined: 2019/01/03 12:02:48
    • Status: offline
    • Ribbons : 0
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 09:09:32 (permalink)
    MiRai
    There's a review up of a 3090 FTW3 Ultra that shows 2x MLCCs and 4x POSCAPs, which differs from the original stock photo shown on sites.
     
    Image (Source)
    EVGA Store Stock Image
    Amazon Stock Image
     
    EDIT: The nVidia sub-reddit is putting together a table for PCB designs that can be currently found around The Internet.


    Thanks for posting this!
    Does the 107% power slider mean 470.9 TDP?
    If so thats pretty good and very comparable to the strix
    #27
    Frammish
    Superclocked Member
    • Total Posts : 236
    • Reward points : 0
    • Joined: 2014/10/18 15:24:44
    • Status: offline
    • Ribbons : 0
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 09:27:02 (permalink)
    @SprayingMango, the cards aren’t time bombs. Just less reliable overclockers. Apparently Zotac is having to reduce  power limits (aka clocks indirectly) but not sure how many others will need to do that. EVGA apparently caught this somewhat early and at least some shipping cards differ from their product photos and do indeed have MLCCs in one or more of the 6 locations.
     
    I certainly don’t want a card that missed being revved to have more MLCCs, but EVGA may be good. They seem to have delayed a lot in the run up to releases, and the FTW3 3080 may even had to be reworked since it’s delayed so long. Hard to know. But EVGA stands behind their cards. In light of all that is coming out, I’d bet they just swap “bad“ cards for “good“ - if there even are any bad EVGA cards in distribution. (Cannot speak for them) It is something to be aware of though.
    #28
    cneuhauser
    FTW Member
    • Total Posts : 1334
    • Reward points : 0
    • Joined: 2007/12/18 12:49:31
    • Status: offline
    • Ribbons : 5
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 09:27:43 (permalink)
    blazemore
    Nothing to see here...my mistake. 




    OUCH... and this is why I don't buy the first round of cards, especially when they didn't get tested or run through their paces like usual. Not necessarily blaming EVGA, nor Nvidia, because this is part of product development. However, this is maybe a lesson learned between the two and other mfgs in the process, during unusual times maybe Nvidia should have more direct communication, and our outline potential weaknesses during development etc. Or maybe the mfgs should build a few hundred first, send them out to enthusiasts who benchmark and TUBE, and wait a few weeks.
     
    I dunno, this was just a funky launch all the way around. However I don't think this is the end of the turmoil with these cards...I'm waiting for more details.
     
    **Wow great article, the more I read about this, the more I'm leaning towards Strix, I'll take the high quality MLCC please.
    post edited by cneuhauser - 2020/09/25 09:52:38

    Entho Evolv Anthracite Grey
    ASUS ROG E-Gaming X299
    i7 9800x 4.8ghz on water (Cuplex Vision block) 
    EVGA RTX 2080 XC Ultra / Water / EK Block
    GSkill Trident Z - 32gig
    2x Samsung 1TB SSD Evo
    Intel SSD 730 Series 240GB 
    Windows 10 Ultimate  
    EVGA 1600 SuperNova T2 
    Samsung G7 Faker 32" 240hz monitor
    Asus ROG Swift PG329Q 175hz monitor
    SteelSeries Arctis Pro+DAC
    B&W 801s, Driven by Sumo Amp via DAC
    #29
    cneuhauser
    FTW Member
    • Total Posts : 1334
    • Reward points : 0
    • Joined: 2007/12/18 12:49:31
    • Status: offline
    • Ribbons : 5
    Re: Igors Lab: Less MLCC, more crashing and lower clocks? 2020/09/25 09:34:28 (permalink)
    pointofdeparture
    JayzTwoCents' 3080 XC3 review sample looks like 1xMLCC and 5xPOSCAP.
     





    Agreed... same dam thing happened with the 2017 Focus RS< they put the wrong darn head gaskets in the vehicles. However, this also concerns me that they were going back and "fixing" cards. Any time you adjust/alter a product like that after it's out of the mfg process, there are ALWAYS problems; Porsche did this with intake and exhaust manifolds etc.
     
    My point is... I'm glad I've been waiting, hopefully all of these 1st gen bugs will be worked out; one last caveate'.... if you already own one of these cards, I'd register that card, and be making a phone call to EVGA about a lifetime warranty!!!

    Entho Evolv Anthracite Grey
    ASUS ROG E-Gaming X299
    i7 9800x 4.8ghz on water (Cuplex Vision block) 
    EVGA RTX 2080 XC Ultra / Water / EK Block
    GSkill Trident Z - 32gig
    2x Samsung 1TB SSD Evo
    Intel SSD 730 Series 240GB 
    Windows 10 Ultimate  
    EVGA 1600 SuperNova T2 
    Samsung G7 Faker 32" 240hz monitor
    Asus ROG Swift PG329Q 175hz monitor
    SteelSeries Arctis Pro+DAC
    B&W 801s, Driven by Sumo Amp via DAC
    #30
    Page: 12345.. > >> Showing page 1 of 6
    Jump to:
  • Back to Mobile