MSI GeForce RTX 2080 Ti Lightning Z PCB Pictured, Overclocked to 2450 MHz

[btarunr]

MSI is ready with a Lighting Z branded flagship graphics card based on the GeForce RTX 2080 Ti. While the company is rumored to launch the card at CES 2019, two pieces made their way to overclockers "Gunslinger" and "Littleboy." While there are no pictures of the card in one piece yet (with its cooling solution in place), there are plenty of its PCB. The large, spread-out PCB draws power from three 8-pin PCIe power connectors. A gargantuan 19-phase VRM conditions power for the GPU and memory, split between two portions of the PCB, on either side of the GPU.

The heatsink cooling the larger portion of the VRM takes up most of the vacant space over the rear end of the PCB. In addition to some finnage, this heatsink uses a flattened heat-pipe to spread heat pulled from the area over the DrMOS (indirect contact), to the tail-end of the heatsink. "Gunslinger" and "Littleboy" have each managed to overclock this card to Godlike GPU clock-speeds in excess of 2450 MHz, likely using exotic cooling, such as liquid nitrogen.



View at TechPowerUp Main Site

 

[btarunr]

Inb4 memory bandwidth in that GPU-Z screenshot. I've asked W1zzard to take a look.

 

[Vayra86]

2450mhz... Idle. That just about sums up the added value of this card.

 

[btarunr]

2450mhz... Idle. That just about sums up the added value of this card.

Apparently it's Catzilla-stable.

 

[Robcostyle]

likely using exotic cooling, such as liquid nitrogen.

And that’s the reason I don’t understand why they’re pushing these extreme models on the market. Since maxwell, nvidia gpu oc was strickted heavily, even more with pascal. You just can’t simply use voltmod, overclocking it further exact point (it’s 2050-2150 Mhz for Ti’s, afaik), impossible without tampering with hardware - thus, you’ll stick your own PCB, so msi vrm the card comes with becomes useless.
Nowadays, lucky golden gpu makes 99% of high overclock - and vrm is almost insignificant here. Especially considering nvidia “recommendation” regarding vrm design, you won’t be having huge issues with phase temps either on any card, if it was designed just “well” - FE can oc better than strix or !amp.

So here is the question - are they gonna make exclusive override? for lightning, allowing soft voltmod? Because, otherwise, it is really useless stuff then - so much power, just in order to get stuck on 2020 MHz...

 

[the54thvoid]

likely using exotic cooling, such as liquid nitrogen.

And that’s the reason I don’t understand why they’re pushing these extreme models on the market. Since maxwell, nvidia gpu oc was strickted heavily, even more with pascal. You just can’t simply use voltmod, overclocking it further exact point (it’s 2050-2150 Mhz for Ti’s, afaik), impossible without tampering with hardware - thus, you’ll stick your own PCB, so msi vrm the card comes with becomes useless.
Nowadays, lucky golden gpu makes 99% of high overclock - and vrm is almost insignificant here. Especially considering nvidia “recommendation” regarding vrm design, you won’t be having huge issues with phase temps either on any card, if it was designed just “well” - FE can oc better than strix or !amp.

So here is the question - are they gonna make exclusive override? for lightning, allowing soft voltmod? Because, otherwise, it is really useless stuff then - so much power, just in order to get stuck on 2020 MHz...

Some people make money from professional benchmarking and OC events, these cards go hand in hand with that but it's more about PR than a reasonable card for everyday gamers. And even for those that won't use LN to cool it, its a desirable object to some.

 

[dj-electric]

Lightning versions stop exciting the moment NVIDIA decided to hard-cap their GPUs for OC
Many simple versions of RTX 2080 Ti can already handle a lot more than the core can do but blocked

 

[T4C Fantasy]

Inb4 memory bandwidth in that GPU-Z screenshot. I've asked W1zzard to take a look.

They are using ancient version xD its fixed now

 

[hat]

nVidia doesn't want to deal with warranty issues arising from users blowing cards by pushing more than what the board can handle. Just because the core can take it doesn't mean the VRMs can. Basically it's like overclocking a 9900k is a budget board. Locking the cards down with "turbo boost" is one way to remedy this issue. It's the same reason you can't edit and flash your own BIOS anymore. That said, it really shafts users who know what they're doing and go out of their way to purchase a premium card like this (with a premium price tag, no doubt)...

 

[TheinsanegamerN]

nVidia doesn't want to deal with warranty issues arising from users blowing cards by pushing more than what the board can handle. Just because the core can take it doesn't mean the VRMs can. Basically it's like overclocking a 9900k is a budget board. Locking the cards down with "turbo boost" is one way to remedy this issue. It's the same reason you can't edit and flash your own BIOS anymore. That said, it really shafts users who know what they're doing and go out of their way to purchase a premium card like this (with a premium price tag, no doubt)...

They could easily solve this by putting in a software fuse that blows if voltages are screwed with, allowing an easy flag for if someone had modified a GPU when it comes to warranty repairs. Samsung knox could detect when a user was screwing with the OS, even after a full factory device restore the fuse would remain blown. Putting one in to detect voltage being adjusted would be trivial.

But nvidia desires control over their consumers, hence why they tried to lock driver updates to a geforce now account some time back. Why these OCers even bother with nvidia anymore is a mystery. They have AMD, why not do insane tweaking to vega GPUs instead?

 

[Vayra86]

Why these OCers even bother with nvidia anymore is a mystery. They have AMD, why not do insane tweaking to vega GPUs instead?

Because they have already been tweaked beyond the edge by AMD out of the box. Which is why most ppl undervolt them instead.

On Nvidia GPUs, their GPU Boost is so advanced it becomes pretty difficult to surpass with manual OC. You may get higher clocks out of it, but it may adversely affect performance. Seen it many times. These GPUs are delivered out of the box with minor OC headroom to play around with and all clock within a tight margin. What's not to like? You know what you get and duds are rare. 98% of the userbase wants that, which is why even AMD is moving to provide a more streamlined experience, revamping their driver GUI etc.

Real OC'ers hard mod anyway so AMD or Nvidia is irrelevant.

You're right, this was never about reducing the amount of RMAs. Its easy to deny a claim. Its about control - control over the performance delivered in each price bracket. The only reason AMD doesn't lock everything down is because its cheaper for them not to. Which is why you occasionally still find unlockable shaders. But it also means the number of 'muh card is bricked' or 'shit it doesn't work anymore' topics is that much higher in the red camp. And that eats away at the AMD mindshare, like it or not.

 

[dj-electric]

nVidia doesn't want to deal with warranty issues arising from users blowing cards by pushing more than what the board can handle. Just because the core can take it doesn't mean the VRMs can

Have you seen how overly built the reference 2080 Ti PCB is? you have to be joking. This is like giving an Aorus Z390 Master to a 8350K

 

[Basard]

Have you seen how overly built the reference 2080 Ti PCB is? you have to be joking. This is like giving an Aorus Z390 Master to a 8350K

Yeah, that's what sells though.... A lot of people go for the beefier version so it will run silent.... skimp and u end up with coil whine or loud fans.
Even the X470 chipsets are over built for AMD's current offerings--maybe they were all just prepping extra early for Zen 2?

 

[Robcostyle]

Yeah, that's what sells though.... A lot of people go for the beefier version so it will run silent.... skimp and u end up with coil whine or loud fans.
Even the X470 chipsets are over built for AMD's current offerings--maybe they were all just prepping extra early for Zen 2?

I have strix 1080 ti, and it whines reaaaaly loud. And I have 144hz screen.....
That's more about 3rd party pricks like ASUS, rather than skimping.

 

[DR4G00N]

They have AMD, why not do insane tweaking to vega GPUs instead?

Because Vega is even more dull to OC, they don't scale well with low temps at all, maybe like 50-100MHz on LN2 vs Water. That and AMD's cards are a lot slower than what nVidia has out now.

 

[Jism]

The problem is not really the voltage, but the allowed current the GPU could use. This is simply limited by Nvidia for a good reason. Perhaps the internals of the chip could not sustain a higher maximum then it could do now. The difference is when going LN2 is that you could archieve a far higher clock and lower power usage by cooling it down like crazy.

The VRM design is at least properly done; but it adds no extra's at all if the chip's current is limited. Look at Vega 56 > Gamernexus made it a hybrid one and bypassing AMD's powerlimit all the way up to 400Watts. When that thing could be LN2'd i'm pretty sure it should easily archieve 2Ghz and above or so. You have so much headroom to play with.

 

[NicklasAPJ]

nVidia doesn't want to deal with warranty issues arising from users blowing cards by pushing more than what the board can handle. Just because the core can take it doesn't mean the VRMs can. Basically it's like overclocking a 9900k is a budget board. Locking the cards down with "turbo boost" is one way to remedy this issue. It's the same reason you can't edit and flash your own BIOS anymore. That said, it really shafts users who know what they're doing and go out of their way to purchase a premium card like this (with a premium price tag, no doubt)...

You Can flash any ref 2080 TI with a custom Vbios wich give 380watt max.

 

[Blueberries]

MSI does a great job with their Lightning cards and this one looks just as great as their previous models, but as others have pointed out...

as great as this card is-- it really doesn't matter.

Typically, power limits are what handicap the performance of NVIDIA's silicon, but as many reviewers have shown: it does not matter with Turing. With the power target adjusted, shunt mods, or other methods, Turing hits a maximum clock speed before power target becomes an issue or visa versa.

This card, with Afterburner, and without modifications, will perform EXACTLY the same as their Duke or Gaming X Trio.


Is the hardware better? Yes. There are three 8 pin power connections, three 60A input filters, as well as an improved VRM setup across the board. This may mean less coil-whine, improved thermals, and more stable over clocks, but as others have pointed out, and will continue to point out, with Turing it doesn't matter. These high performance cards only benefit overclockers that are pushing their silicon to the limits with subzero cooling. If you're not going to modify this card and overclock the living hell out of it, you won't see any benefit.

 

[cadaveca]

Have you seen how overly built the reference 2080 Ti PCB is? you have to be joking. This is like giving an Aorus Z390 Master to a 8350K

OK, but we know that nvidia designs PCB for worst-case scenario, then bins GUPs extra-strict so they look god-like. You cant' blame a manufacturer for offer more so that you can ensure that what hey provided to you isn't ever going to be the limit in your overclocking adventures.... while the reference PCB does.