Zoraptor

I tend to assume that most professional pundits are either literally or effectively paid to hold whatever view they hold, since there's a vast difference between being a pundit and an expert. Maybe not directly paid; but you're a lot more likely to be asked for your opinion and get a position in a think tank, panel show, exposure for your website etc if the person asking the questions can be guaranteed they're going to like what you say, or like the clicks you bring in. What is really needed is a specific 'no confidence' option rather than generic ballot spoiling or write in votes for novelty 'candidates'. 2017 French Presidential election; 1 out of 8 people (4 million in absolute terms) who bothered to turn up and choose between oleaginous corporate lickspittle Macron or odious crypto-ish fascist Le Pen chose to spoil their ballots instead, and that's in addition to the extra 4 million people who didn't vote at all compared to 2012's contest. That certainly didn't stop media trumpeting Macron's 'landslide' victory as a triumph of democracy though.

Went on sale at 3am here, sold out by (earlier than) breakfast time. That's with no 5900/50X though as they're coming in next week. ~20USD markup over US prices, which is way less than for most products. I do have to admit that despite deciding not to get a 5000 due to the pricing and dead socket AMD's marketing has got me reconsidering, so I won't be altogether disappointed if there isn't any Ryzen/ Radeon stock through to beyond Christmas.

Sky News Australia == Uncle Rupert's Australian mouthpiece, much as Faux News is his US one.

Killjoys got two (?) seasons to wrap up. That was the deal Dark Matter would have had if they'd cancelled Killjoys instead, one suspects the deciding factor was that Killjoys was incredibly cheap to make.

Supposedly the 20GB 3080Ti is now reinstated as a forthcoming product. Rumoured, rumoured cancelled, now rumoured reinstated; it's Schrodinger's GPU and you'll only know whether it exists when it turns up. (Still say that a 12GB version with the full bus makes more sense, but I guess having more VRAM than RX6_00 is more important as a marketing point)

Trump 'wins' on the night, Biden win after counting finishes for me. Which may be an interesting aftermath, to say the least. After being repeatedly told that 2020 was not 2016 and the same mistakes wouldn't be made again... sheesh, the Democrat Party is not a learning animal. I think they'll get away with it- just- this time though. One thing is for sure, Donald Trump may have his limitations as a human being, leader, husband everything else; but he's got the most absolute talent for effective campaigning I've seen from anyone, win or lose.

It's not surprising, they use previous patterns etc for predictions which is why they predict some states almost instantly. Say small rural booths report first, as they have less counting to do, and such booths usually vote R. Result therefore skews R early. The big urban booths from, dunno, Richmond or the Washington outskirts are bigger and report late, and they historically skew heavily D. If they're historically 75/25 in favour of D and are projected to stay that way Biden wins, easily.

The zipline incident was almost certainly deliberate. He may go out of his way to appear goofy to appeal to the common man, but Alexander Boris de Feffel Johnson (yes, really) is actually so competitive he'll steamroll Japanese children to win.

If your choices are 2020 model Biden or Donald Trump disappointment of one sort or another is pretty much inevitable. New Zealand and Australia have travel advisories out for the US for election related violence. I suspect there may be some passive aggressiveness at play there though, since the US issued covid related advisories for both of us despite the rates here being magnitudanally lower than those in the states.

I wouldn't be overly concerned about VRAM, AMD (and their partnered games) will always imply that you need lots of it because having more is an AMD point of difference vis-a-vis nVidia; especially now when perhaps the main criticism of the 3070/80 (apart from general availability) is low VRAM. I'd suspect the raytracing on the 2080Ti is going to become obsolescent first, which is always a risk for a 1st gen tech.

I'd expect the Intel GPU line up to be 'weird' for at least the first two generations. They will be miles behind in terms of pretty much every value/ power/ performance proposition so will target some very odd seeming segments where they think they can be competitive despite that. Assuming they have any appreciation of reality at all they must know that the first years at very least are going to be loss making for the gpu division. Integrated graphics has a natural home, the very top end professional/ compute/ AI/ ASICs etc have (generally at least) massive margins or 'simple' demands (ASIC) and some scope for innovative novel solutions to find niches, plus the ability to try leveraging Xeon. Consumer... there is some theoretical space with the price inflation of the last few years but Intel cards will be hot, weak and unstable/ not feature complete. Their only real advantage is that they will (or should at least) not be expected to make a profit off them yet so can price aggressively.

Wouldn't say autocrats get that sort of chant much- if they do, it's usually forced so not genuine- but it's certainly the sort of thing you get from cults. Unfortunately the sort of unquestioning obedience/ prosetylising/ mania you get from cults and which is mostly how they 'work' is also very appealing to politicians too.

If it really is the 6800 nonXT I'd say it's better than not bad. That would have been very top end expectation for the 6900XT's RT performance, a week ago. I will be very highly amused if AMD has been sandbagging on the upscaling front as well.

The flip side of being well regarded by fans is that journos will look for anything they can to drag them down.

That Linux tech has existed for a couple of years, iirc, and is theoretically vendor agnostic. In theory the MS/ Windows implementation is vendor agnostic too, but it's fair to say that AMD is likely to milk SAM for all it's worth anyway.

If I had to guess, 3070Ti (GA102) sometime early next year, 3080Ti... not sure we'll get one. A price drop on 3090 is more likely, imo, if there is a 3080Ti I'd suspect it would have the full bus and 12GB VRAM as its main selling point. Assuming, from the leasked slides, that Ti is the preferred naming scheme this gen. 3070Ti as a rescue bin for failed GA102 makes sense, especially if the rumours of yield problems are true. An extra 2GB of RAM and some speed over the 3070 give sales points, and that approach adds stock to the obviously depleted GA102 pool. A 3080Ti however would just spread that pool around even more shallowly and while there is a big gap in price from 3080-90 the performance gap is small. My suspicion is that SUPER cards, and if they arrive it would be around a years time, will be reserved for a node switch to TSMC for the GA102 based product line (with the GA102 Samsung line probably being retired). That's if they don't simply decide to rename the whole GA102 suite to 4000 or 3XX0 series if they switch nodes.

Attacking other people's patriotism is a simple yet effective rhetorical device. "If you like ___ so much, why don't you move there?" is perhaps the staple reflex question when suggesting somewhere else does something better, and has been forever. It's a worthless rhetorical device designed to attack the arguer rather than the argument, and shift the argument's subject. Should be noted though, that type of argument is hardly coming only from the R side of the 'debate'. 99% of accusations of people shilling for Russia are exactly the same, attempts to attack the person making an argument rather than the argument itself. A solid number of anti Trumpers are 100% convinced that anyone who supports him has to be a 'traitor' and think the- utterly unsubstantiated- peepee tapes etc are gospel.

The cache is meant to be used in concert with very aggressive on chip culling to reduce data to/ from the memory. It's also a pretty obvious interim step towards chiplet/ mcm videocards, and has the side effect of reducing heat. The 16GB of memory is definite overkill, but it isn't that expensive and gives marketing advantages over nVidia. I'd happily have a gentleman's bet about AMD 'winning' the raytracing battle if you want. RTX will no doubt linger on as nVidia specific branding since they've invested a lot into it, but it's 'dead' as physX long term, for the same reasons. The only games that will have nVidia specific solutions will be where nVidia pays them specifically to block AMD- or at the moment Vulkan games*, since (ironically, after who made Mantle) the only full release raytrace module there is the nV one. Everything else will be plain old DXR (and Vulkan_RT when released). If the rumours about raytracing performance on AMD being ~2070/80 level were true I'd be more bullish on RTX's future but AMD has solidly leapfrogged Turing with their first try, and isn't that far behind Ampere. *So at the moment the only (?) RTX game that fundamentally cannot run on AMD's solution is Youngblood, since it's Vulkan, the rest use DX12U/ DXR. And since I complain about nVidia co-opting others' technologies with their branding per RTX I/O etc, should be pointed out that SAM is actually yet another MS tech this time with AMD branding, and should work on more than just Zen 3 CPUs including (theoretically) the nextbox.

For years the only non miniature red pumpkins available here that were (occasionally) sold came with the disclaimer that you should feed them to livestock when finished with them, which was not encouraging. I kind of presumed things were different in the US since everywhere seemed to have pumpkin pie for sale as a dessert when I was last there- I'm not sure I've ever seen it offered here. Since I've been prompted I checked my seed catalogues and we do now have a good tasting carving pumpkin available (variety name "Jack o' Lantern", appropriately enough; though the seed supplier is pretty niche and it isn't available retail) but it's a bit late to try planting any now plus I don't have the space. Something to try for next year though.

Yeah, fundamentally you cannot ever get around weight as the primary factor. But reusability meant it was very complicated and had to be overengineered- partly why it was so heavy of course- and especially once it started having problems the economies of the reusability equation reversed due to the amount of maintenance and safety checks required. If you're at the stage of almost having to completely disassemble a complicated vehicle for safety checks you're likely to be better off using a simpler single use vehicle instead. Ultimately, If you sat down in 2010 to decide on a Shuttle replacement that took its mistakes into account you'd inevitably end up with something less complicated, more reliable and lighter rather than doubling down on a space plane- and you tend to get those things but making stuff single use. The counterpoint being the Shuttle was late 70s/ early 80s tech and there was 30 (40 now) years of advances since then- and it certainly is one of the great ironies that it was failures with the boosters rather than the reusable shuttle itself which caused the fatal problems. So yeah, arguments for/ against both approaches.

Yeah, it's a fair bit behind a 3080 whichever card it is, but If you said that any AMD card would have ~20% better raytracing performance than a 2080Ti 6 months ago most people would have thought you were licking toads or something. Best case scenario is probably that it's 6800XT, and if the AIBs can really get the clocks up to 2.4Ghz as some of the leaked benchmarks suggest that would make up a lot of the deficit to 3080 in raytracing (assuming clockspeed is the limiting factor, of course).

I don't think there's anything more behind the 6800 pricing than it being a rescue bin for failed chips and not wanting to cut down good chips for it without extra margin- similar to Radeon 7/ Instinct. One of the things that certainly did surprise me was the 6800 being 60CU and otherwise identical to both cards above it though, I definitely thought there would be a ~60CU 'native' chip. The same cache amount seems odd too, does a single error in the cache instantly result in a failed chip? Can't see it being a great overclocker, presumably some of the failed chips will be due to not being able to maintain clocks in general rather than outright failed hardware sections. I suspect exclusivity agreements are the reason for some features barely being mentioned. On paper, that raytracing is... better than decent, and they've said nothing about specialist hardware based raytracing being present. Technically, they just implied it wasn't by talking about the work being done by CUs instead of dedicated hardware, but it's a weird implication to make since it implies poor performance. OTOH, SAM and RAGE get mentioned because they are AMD techs and consoles use Zen2 which aren't compatible with them.

There is a good precedent for that view though, in the Shuttle program. It was meant to be cheap because it was reusable, but ended up incredibly expensive, in part because it was reusable. (I don't necessarily agree with that view personally, but I'd say it's certainly not just due to lobbying that that opinion is held)

haha, I'm a muppet Since I managed to delete my own post, tldr version. 6800 is Navi21 with the same memory etc. AMD will only want to sell genuinely defective chips as 6800, if they can. They want to drive the value proposition towards the higher margins. If they have to choose recycling a chip or a 6800 they'll chose the 6800, if they have to choose between a 6800 or a 6800XT+ they'll want to choose the higher margin card. It should be a bit better than a 3070, and if AMD's claims are true brutally murder and dismember its corpse when it comes to memory bandwidth and quantity.

Not really, it's a proprietary implementation/ extension. It still uses the raytracing built in to DX12U/ Vulkan. It's clear that CDPR could have at least implemented console raytracing. That they didn't, along with the massive amounts of nVidia branding, is more than suggestive that the 'problem' is an agreement to use nVidia tech exclusively at launch rather than anything else. As a personal note, this sort of thing is exactly why I would not consider buying an nVidia card. My tolerance for co-opting standards is not high. Note, since it is easy to miss: there is added specialist raytracing hardware on the 6000 series (and one benchmark for the RT elements); "New to the AMD RDNA 2 compute unit is the implementation of a high-performance ray tracing acceleration architecture known as the Ray Accelerator. The Ray Accelerator is specialized hardware that handles the intersection of rays providing an order of magnitude increase in intersection performance compared to a software implementation [..] Measured by AMD engineering labs 8/17/2020 on an AMD RDNA 2 based graphics card, using the Procedural Geometry sample application from Microsoft’s DXR SDK, the AMD RDNA 2 based graphics card gets up to 13.8x speedup (471 FPS) using HW based raytracing vs using the Software DXR fallback layer (34 FPS) at the same clocks. Performance may vary." It's just not physically separate hardware like nVidia's RT/ Tensor cores. They're not going to be using the 'GTX solution' unless deliberately nerfed. As Keyrock said it was mentioned briefly, but is not ready for launch.