asus tuf rtx 3080 non oc version

[Mando772004]

does this card https://www.asus.com/Graphics-Cards/TUF-RTX3080-10G-GAMING/

have the manufacturing problems other 3080 cards do that cause crashes?

[SPARTAN VI]

I suspect not. The renders/product photos on etailers and Asus' official product website show the cheaper "SP-Caps" (left image below), but apparently they changed it to an all MLCC configuration for mass production (right image below).

 

[xg32]

9 minutes ago, InfernalClaw said:

does this card https://www.asus.com/Graphics-Cards/TUF-RTX3080-10G-GAMING/

have the manufacturing problems other 3080 cards do that cause crashes?

the tuf is the best card atm but the crashes are not all caused by power delivery, it's likely due to poor binning (some cards can't even be stable at the stock boost clocks)

[Jurrunio]

all of them do, you need luck to get a unit that doesnt

 

or you could just buy the FE

[i_build_nanosuits]

3 hours ago, Jurrunio said:

all of them do, you need luck to get a unit that doesnt

 

or you could just buy the FE

that is not correct, at this point it is understood what the problem is and what makes some models better than others.

[StarsMars]

3 hours ago, Jurrunio said:

all of them do, you need luck to get a unit that doesnt

 

or you could just buy the FE

Feels weird Nvidia didn’t use reference for fe. 
Doctor won’t take the medicine it prescribes. 

[Jurrunio]

1 hour ago, i_build_nanosuits said:

that is not correct, at this point it is understood what the problem is and what makes some models better than others.

stand corrected: They're crashing regardless of model like how 16 and 20 series cards with GDDR6 are artifacting during launch

 

59 minutes ago, StarsMars said:

Feels weird Nvidia didn’t use reference for fe. 
Doctor won’t take the medicine it prescribes. 

the "my_dog_ate_part_of_my_PCB" design isn't a good starting point for board vendors which keep their dogs out of the room I think.

[Exeon]

1 hour ago, StarsMars said:

Feels weird Nvidia didn’t use reference for fe. 
Doctor won’t take the medicine it prescribes. 

I thought Nvidia reference was 5 POPSCAPS and 1 MLCC, FE went with a 4/2 setup.

 

Honestly considering everything it's pretty clear this can't be the only issue, if it's even an issue at all.

Could just be a bad driver, which we've seen in the past.

[Gohardgrandpa]

I've seen so much non sense surrounding these capacitors on these cards. MLCC is prone to cracking, I'll gladly pass on that and take the big ugly caps any day. Seems to me they pushed these cards out of the box like ryzen cpu's. There isn't shit for OC headroom. I haven;t overclocked a video card in years as last time I did it I saw a marginal increase in performance. I would rather undervolt and leave clock speed alone to lower temps then to push the core and vram to gain a small 3-5% improvement in FPS but higher temps

[StarsMars]

4 hours ago, Jurrunio said:

the "my_dog_ate_part_of_my_PCB" design isn't a good starting point for board vendors which keep their dogs out of the room I think.

Not talking about the pcb shape but the upped components

[SPARTAN VI]

10 hours ago, Jurrunio said:

stand corrected: They're crashing regardless of model like how 16 and 20 series cards with GDDR6 are artifacting during launch

 

the "my_dog_ate_part_of_my_PCB" design isn't a good starting point for board vendors which keep their dogs out of the room I think.

I noticed that tweet too, but I hope HUB replicated the crashes on other boards and boot/OS drives, otherwise it's still only one data point. 

[i_build_nanosuits]

13 hours ago, Jurrunio said:

stand corrected: They're crashing regardless of model like how 16 and 20 series cards with GDDR6 are artifacting during launch

 

the "my_dog_ate_part_of_my_PCB" design isn't a good starting point for board vendors which keep their dogs out of the room I think.

pretty SURE i consider Buildzoid more reliable than hardware unboxed for this type of stuff.

https://www.igorslab.de/en/what-real-what-can-be-investigative-within-the-crashes-and-instabilities-of-the-force-rtx-3080-andrtx-3090/

https://videocardz.com/newz/manufacturers-respond-to-geforce-rtx-3080-3090-crash-to-desktop-issues

 

Look even MSI is aware of it they fixed the gaming x trio already:

 

https://videocardz.com/newz/msi-quietly-changes-geforce-rtx-3080-gaming-x-trio-design-amid-stability-concerns

 

[SPARTAN VI]

There's really room for both of you to be right with regards to the root-cause of the crashes. HUB's testing alludes to software/driver causing crashes while Igor/Buildzoid attribute the widespread crashing to be poor power filtering by SP-cap configs. The symptoms are similar, but doesn't mean the root-causes are mutually exclusive.

 

HUB's experience is that even MLCC configurations are crashing (inconsistently) so Igor's/Buildzoid's analysis alone doesn't reconcile the fact that MLCC configurations are still crashing. Without more information regarding HUB's testing, I suspect there's something environmental causing HUB's crashes rather than something more widespread like the power filtering issues Buildzoid and Igor's lab had identified. 

[i_build_nanosuits]

3 minutes ago, SPARTAN VI said:

There's really room for both of you to be right with regards to the root-cause of the crashes. HUB's testing alludes to software/driver causing crashes while Igor/Buildzoid attribute the widespread crashing to be poor power filtering by SP-cap configs. The symptoms are similar, but doesn't mean the root-causes are mutually exclusive.

 

HUB's experience is that even MLCC configurations are crashing (inconsistently) so Igor's/Buildzoid's analysis alone doesn't reconcile the fact that MLCC configurations are still crashing. Without more information regarding HUB's testing, I suspect there's something environmental causing HUB's crashes rather than something more widespread like the power filtering issues Buildzoid and Igor's lab had identified. 

HUB has been the only one to report crash on founders edition/TUF cards tho...what if he is mistaken and crashes only happened with other cards and on the mist of all the testing he just didn't noticed that only some cards would crash and not all of them...

Have you seen other outlets reporting crash on FE specs cards?

[SPARTAN VI]

I've no idea if HUB has done their due diligence, but "maybe they forgot" is a bit of a stretch. There's a sprinkling of individual reports on various forums/publications where TUF, XC3, and FE owners have reported crashing. 

[FaxedForward]

6 minutes ago, i_build_nanosuits said:

HUB has been the only one to report crash on founders edition/TUF cards tho...what if he is mistaken and crashes only happened with other cards and on the mist of all the testing he just didn't noticed that only some cards would crash and not all of them...

Have you seen other outlets reporting crash on FE specs cards?

Some owners on Reddit have reported crashes with FE and TUF cards but keep in mind there are very, very few FE cards in particular out there in the public so sample size is low.

 

It has become clear that the 3080 clocks are not stable above 2,000 MHz for ANY card as tested by Paul’s Hardware so for cards that allow you to boost that high out of the box (Zotac is one IIRC) that is definitely a known cause.

 

For other cards that do not boost over 2,000 MHz I would wager that the issue is some combination of poor silicon (Samsung 8N is VERY immature and we all know how rushed this launch was), poor power delivery (capacitor spec/layout), and early drivers (when has this NOT been an issue for new graphics chipsets?).

[-rascal-]

1 hour ago, FaxedForward said:

Some owners on Reddit have reported crashes with FE and TUF cards but keep in mind there are very, very few FE cards in particular out there in the public so sample size is low.

 

It has become clear that the 3080 clocks are not stable above 2,000 MHz for ANY card as tested by Paul’s Hardware so for cards that allow you to boost that high out of the box (Zotac is one IIRC) that is definitely a known cause.

 

For other cards that do not boost over 2,000 MHz I would wager that the issue is some combination of poor silicon (Samsung 8N is VERY immature and we all know how rushed this launch was), poor power delivery (capacitor spec/layout), and early drivers (when has this NOT been an issue for new graphics chipsets?).

 

Also note that the crashing can ALSO be caused by poor power delivery to the card via the PCI-E power connections.

I don't remember who / which reviewer it was, but they swapped out their standard test system PSU for a different one, and it fixed the crashing issue.

[Exeon]

18 minutes ago, -rascal- said:

 

Also note that the crashing can ALSO be caused by poor power delivery to the card via the PCI-E power connections.

I don't remember who / which reviewer it was, but they swapped out their standard test system PSU for a different one, and it fixed the crashing issue.

Some reviewers have issues reproducing the crashes.

 

The more I read the more I notice there are 2 seperate issues:

 

• "Instability" caused by low quality POPSCAPS, present on 1-2 brands like ZOTAC, other brands have a mix of MLCC's or if you believe Gigabyte use higher quality ones, this is confirmed by both EVGA's public statment of changing their capacitor layout before launch (but sending the bad ones to reviewers) The fact that you can see old renders on cards with the POPSCAP's but having a changed layout on release means this was a known issue for most board partners

 

• Crash to desktop issues when pushing the cards beyond 2000mhz, which is an issue all card have, maybe it can be fixed with a driver update maybe not, but in the end this is an overclocking issue the promised performance isn't affected, this isn't isn't as clearly present on the FE or Asus TUF models because the default settings don't push the cards beyond 2000mhz on boost clocks.

[Jurrunio]

2 hours ago, StarsMars said:

Not talking about the pcb shape but the upped components

Reference PCB does have the better components, but there's always room for partners to make it worse to cut cost and all of them did. All that matters is the design passes Nvidia's criteria, in this case Nvidia's criteria is actually too low

 

4 hours ago, FaxedForward said:

Samsung 8N is VERY immature

Is it not just 10nm+ for them?

[FaxedForward]

1 hour ago, Jurrunio said:

Is it not just 10nm+ for them?

I'm trying to find the article I read that went into detail, but it basically said that Nvidia wanted enough changes to Samsung's process that it's not the "regular" 8nm and as a result yields are trash and Nvidia is minimizing production until they improve, which is why the cards are so hard to come by in general.

[Yoshi Moshi]

I have a 3090 TUF non OC version that I picked up at MC on launch day. My MC had two, and thankfully I was one of the first two people in line, waited 36 hours in line. I can confirm on my card that it does not have the cheaper CAPS.

 

Also the TUF non OC card is the only non-reference card that you can pick up at MSRP (other than FE) that has a WB for it. If you can't trust known brands, than look at the brands making cards on the reference design, they are all manufacturers I have never seen for sale at physical stores.

 

Also fun point to note - The backplates on the cards do not cover up the CAPs! Very interesting. Why didn't they put a thermal pad over it and allow the backplate to go over it to spread out the heat? Clearly there's some stability issues with cards, at higher clocks the CAPs seem to have issues. Anywho, the backplates from EK cover the CAPs and you put a thermal pad over it. This might have some actual benefit, the EKWB backplate over the stock backplate. I'll try it out and see if it makes a difference.