An interesting look at the Asus tuf card. Has really nice cooler on it. He also does a shunt resister mod... Spoiler: Spoiler alert Doesn't really change much, but he only does the mod on one of the 2 resisters for that 8pin, so might get more luck if done on both and got under water.
It's just a crap overclocker. Mostly due to the node. If Nvidia had left the gas in the tank it would have been no faster than a 2080Ti FE stock vs stock. So they have overclocked it themselves, hence the shameful power consumption. Agreed though, that TUF card is nice. The Dual I had looked very similar but had no alu on it. That said even in plastic it was in no way offensive like some of the supposed uber cards.
Just need and to turn up and put some pressure on. Why make something twice as fast when you're already twice as fast as the competition.
They're not though and I think they know it. AMD haven't even launched a serious Navi GPU yet to be fair. When they do? it's TSMC, so it won't be riddled with issues like Samsung. Plus why let your card get worse reviews at stock when you can just overclock the balls off it yourself and make it look better when reviewed? I await the "real" next gen from Nvidia on 5nm TSMC, because from what I am hearing it will bury AMD.
That's what I mean. They've done enough to for the ampere cards to be just enough faster than Turing that it looks good. They don't have to do any more than that as there is no competition that high in the product stack. When AMD do being something worthy to bring to the table, then we'll see what they can really do. Or maybe we'll find out that actually they're spent and have to rely on trick software to get more performance. A lot of people thought Intel were sandbagging because AMD had nothing to offer, but when push came to shove they had very little left in the tank and they're likely to get surpassed come zen3.
They're not sandbagging. I don't believe so any way. The fact is if Ampere could even overclock I think the TDP would be absolutely hideous and they know it. They've done whatever tricks they could do to make it look as good as humanly possible, so I will congratulate them on that at least. I can't see any air cooler, no matter how well designed, costing no matter how much, being able to deal with a 500w GPU. In fact, my Kingpin can only cope (with a dual rad !) with about 450w before you hit a heat brick wall and need LN2 to go any further. And it doesn't even have an air cooler. Hardly like you can slap a NHD-14 on a GPU really is it? They've held back what they needed to hold back to make this launch look good and reasonably priced. I bet the 20gb version won't be anywhere near as reasonable (the real one, that they should have launched now). I still don't care how people try and shake this launch, 10gb is not enough for what the card is and that is a pure 4k card. Drop it any lower and no CPU on earth can keep up and the 2080ti totally closes the gap, especially when overclocked (because it's a TSMC). So OK it's a pure 4k card? cool. So why are there games that already exceed that limitation? and I am not talking next gen games either, I am talking now. Steve @ GN made the only mistake I have ever seen his cynical ass make, and that was to say "10gb? pfft, it's not the amount we are talking about here it's the bandwidth". Know what? that is *EXACTLY* the same crap they said when Fury X launched and I believed it the first time around. So you can imagine my horror when within 6 months COD BLOPS II came along and ran my card out of VRAM and hung my rig. Activision tried to get around this by detecting it was a Fury X and removing all of the "extra" settings you could get on the 980Ti, but that was easily hacked out and the damage it caused plain to see. AMD then "fixed it" by allowing the card to texture stream from the paging file, and it tanked performance as you would expect. And that hasn't been improved much. Instead of going to your HDD it now goes to your RAM and tanks performance in exactly the same way, as in exhibit A. When a 2080 runs out of VRAM and gets its ass kicked by a 1080Ti. But the most delicious piece of irony about the whole launch was how they used VRAM to make the 3080 splat the 2080. That's ironic, because in four month's time when AMD have had their go you will see the *real* 3080 with 20gb VRAM and it won't cost no £649. I reckon more like 800 quid.
I cancelled my pre-order yesterday for the 3080 both because of the silly wait (with money in retailers bank account) and my belief that the 20gig version will be the 'real' 3080.
All of the third party apps (AB, Precision and Gigabyte's app) do it already. All I had to do was click it, bingo bongo 2150 2080Ti. Der8auer has shunt modded a 3080 now and there really is nothing of note left in the tank.
I had to use a top secert vbios with my 2080 Ti to get rid of all power limits to get my card to run at 2150MHz.
That thermal image looks weird. The guy claims to have treated the board in a conformal coating of known thermal emissivity, but you see those "cold" passives? Yeah, they don't look like that if you've coated the board properly. They look like this: See how all the passives are roughly at the same temperature as the board? That's what you'd expect to see. Having them "cold" suggests they're a different emissivity, which in turn suggests that the board hasn't been properly coated.
The coating he's talking about sounds like a waterproof conformal coating, and he specifically says it's transparent - so I'm thinking it's probably transparent to IR as well as visible light. In which case he's wasted his time: might as well not coat the board at all! EDIT: Or the opposite - maybe the coating's highly IR reflective, which would mean coating the board made his readings less accurate than if he'd left it alone... Either way, I wouldn't trust any of the figures he's sharing. Can't see anything about having verified with a contact probe, either.
Yeah, we don't use the Thermal stuff for thermal testing, we just tape heat probes in place and be done with it or I build thermal probes into the PCB designs! SWIR just isn't good for it.
We use thermal epoxy but our stuff goes though ambient temp swings (most abusive being -40 to 95 in < 10min then back again after a 1 hour soak)
https://www.tomshardware.com/news/n...witter&utm_medium=social&utm_campaign=dlvr.it https://www.reddit.com/r/nvidia/comments/ix9zcg/early_msi_geforce_rtx_3090_gaming_x_trio/ (UPDATED with Port Royal, VRmark Blue Room, and OpenVR benchmarks) Hey guys. I posted this stuff at ResetEra but I figured I'd do it here too since the 3090 is launching this week and I'm sure some people are deciding between that and the 3080. And here are some scores from benchmarks I've run, which you can compare to 3080 reviews: Time Spy (graphics): 20569 Time Spy Extreme (graphics): 10560 Port Royal: 13613 (63.03 fps) Blue Room: 6324 (137.87 fps) OpenVR: 84.15 fps (Valve Index) Assassin's Creed Odyssey (Maxed) 3840x2160: 71 2560x1440: 84 (this feels very CPU bottlenecked) Horizon Zero Dawn (Ultimate Quality) 3840x2160: 86 2560x1440: 123 Red Dead Redemption II (Maxed/No MSAA) 3840x2160: 66 2560x1440: 90 Shadow of the Tomb Raider (Maxed/DLSS/No RT) 3840x2160: 112 2560x1440: 133 Shadow of the Tomb Raider (Maxed/DLSS/RT) 3840x2160: 82 2560x1440: 108 I've also run Control through its paces. Maxed with DLSS rendering at 2560x1440, performance hovers around 55-70 depending on the room/action. With DLSS rendering at 1920x1080, numbers are closer to the 80-95 fps range. The rest of my system is due for an upgrade next year, but I have an i7-8086K @ 5.1 GHz with 32GB of 2666MHz RAM. I've managed a minor overclock of +75 Core, +200 Memory. In terms of temps, at full load, I'm seeing around 59°-63° in a fully air-cooled Corsair 1000D case. With fans fully ramped up in Afterburner, it's still quieter than my PS4 Pro So double the price for up to 20%. Bargain.
We understand the 2080Ti is decent but if you want the fastest card out of the box money be damned, the 2080Ti isn't it, it is that simple really.
I think you've been - entirely fairly - confused by @Vault-Tec copying and pasting someone else's post from Reddit without using quote tags: those benchmarks are from a 3090 (allegedly), not a 2080 Ti. They show about a 5-20 percent uplift over the 3080, but that's based on a dude running an older CPU with slower RAM than most test-bench systems professional reviewers have used so it ain't an apples-to-apples comparison.
