Start Saving for your 48GB 4090Ti 800watt "The Beast"
From Toms Hardware
Nvidia is projected to release three high-end GeForce RTX 40-series graphics cards based on the Ada Lovelace architecture later this year targeting demanding gamers. But apparently Team Green is also prepping a rather monstrous graphics board based on its top-of-the-range AD102 GPU that will carry 48GB of memory with a typical board power of 800W, clearly gunning for the top of the GPU benchmarks hierarchy and the fastest of the best graphics cards.
Hardware leaker @Kopite7kimi dubbed Nvidia's flagship GeForce RTX 40-series graphics card 'The Beast' without attributing a model name to it, but normally Nvidia would call it GeForce RTX 4090 Ti or Titan. This board is said to carry a GPU with 18,176 CUDA cores (up from 16,384 CUDA cores in case of the GeForce RTX 4090) as well as 48GB of GDDR6X memory featuring a 24 GT/s data transfer rate (up from 24GB of memory at 21 GT/s in case of the RTX 4090).
Price is expected to be "Above" $2000
Comments
The cost of that thing, plus a more powerful PSU to upgrade the already-expensive-computer I just got recently (how much would the electric bill go up also?), would probably be outrageous. I'll have to get by with the 3090 for however long it lasts me since I'll be paying on this PC for a while.
It is going to be interesting when I finally upgrade. I am still using a Quadro M6000. It has stood the test of time well. Getting a card that is a few generations newer and many times more powerful will be fun ... oh, and quite an expense which is why it has not happened yet.
At the same time, I wonder how long it will take to reach the limits of a 4090Ti ... with increasing map qualities and demanding animation, I have reached the limits of 64 Gb Ram and 24 Gb Vram multiple times.
...if I had the funds, I would stick with an A6000, still 48 GB but only 300w and dual slot so I wouldn't need to have my apartment rewired, a new PSU, and a new case.to boot.
I think we'd all "like to have" one, but realistically, it probably won't be in the cards for me. I'll be upgrading from my Ryzen 3 3900x in the next cycle, and the DDR5, New power supply, and Mother board are going to all add up. The 4080, with the rumored 16GB VRAM may be the sweet spot for me... depending on how close in price it is to the 'regular' 4090 with 24GB ram...
...right now wouldn't mind an M6000 as it has double the VRAM of my Titan-X and still has a TDP of 250w.
It'll probably become semi-affordable by the time the 6000 series comes around. Of course, by then you'll be barely able to render 2 Victoria10s with clothing and hair on it.
i cant even afford a 3050
I would have to save 18 months or more to buy such a card and only after I buy & add the beadboard to all the rooms in my house. Funny thing, I got the ideal to put up beadboard from DAZ 3D itself, because there is a architecture model of rooms that Sloshwerks did and also from an old Norman Rockwell painting but thought I could never afford that but looked it up at the national hardware store chain and whaddaya know, each 4'x8' panel can be had for only $27. Unfortunately, they raised the price by $3 in the sometime past 4 months, as it was $24 a panel, but still, I can afford to do approximate one room every month if I do the work myself. I'm very glad because the drywall all over the house is kind of worn and beaten and short of ripping it all out and putting new drywall up, which is difficult, more expensive, and more mistake-prone then just panelling over it with beadboard.
I had the good fortune to be given a RTX 3060 12GB from AgitatedRiot and have done a few tests with nVidia's Omniverse and DAZ Studio. With DAZ Studio, the 3060 is great but with apps from the Omniverse it is very slow, slow enough to question that if I saved 18 months and bought an RTX 4090 TI would that 4090 TI be so fast as to negate all the slowness of those Omniverse apps in the 3060? I have my doubts but by the time I get ready to seriously consider it I will have to try and find some time trials from the difference RTX cards when running nVidia Omniverse apps. I've not even checked to see if such data exists yet though.
For some time now I've been noticing that manufacturers seem to be generating desire for machines with FAR more capability that even most tech people need. Much less the common man. Yet we keep falling for it. The whole industry is geared to making it harder and harder to be satisfied with still functioning machines.
I understand the logic, progress is a good thing, gotta keep corporate profits up, investors happy, employees paid and retained. But at what point do customers need to buy a locomotive when a Buick will do?
'bout 6 minutes.
I was never interested in the 3090ti as it looked like a rip-off to me.
The 4090ti, however, well that could just be worth it.
I've driven locomotives.. much more fun (albeit more work at times) than a Buick..
...I don't see why 24 GB GDDR5 of VRAM would be less able to hold a large poly/texture load than 24 GB of GDDR6 on an A5000 or 6x on 3090 could. Speed would be the major difference due to slower clock speeds and fewer cores (as well as the lack of RT cores), but 24 GB should be able to cover 24 GB matter which generation it is. 32 GB of DDR3 can hold just as much data as 32 GB of DDR4, again speed is the differentiating factor.
I still primarily work with G3 which is a bit "lighter" on the resources. Probably never get to G10 as Daz will likely dump W7 support before then. Surprised they held on this long particularly since Nvida ceased W7 driver support last year and.everyone else, from Adobe to Pixelogic has already done so as well. Just a matter of time.
...yeah the old workstation I built over a decade ago is still running just fine. Granted I upgraded it as far as it can go (24,GB System Memory, 6 core Xeon CPU, 12 GB Titan X, and new SATA SSDs) and yes it isn't a super powerhouse like I've seen some people have. but it i still does what I need admirably. Yeah I don't bother very much with dForce or volumetrics. but I'm fine with that.
I remember when I first built it and mentioned it had 12 GB of memory and 8 CPU threads, some people referred it as a "render beast" (of course that's when we only had 3DL with UberEnvironment to work with, Iray turned it into a "kitten" until I got the Titan-X).
Beginning to get cold feet about W11 as it sounds like MS is just going to throw fluff updates at users whenever they feel like it. W7's worked just fine for me all these years without so much as a rare BSOD (and that was due to buggy Nvidia drivers). So now considering looking back into slightly older school stuff, like the dual 8 core Xeon with 128 GB of DDR4 build that I originally considered. As long as it has PCIe 3.0, and a reasonably updated BIOS I should be able to run the 3060 on it and use the Titan to drive the displays and augment the core count.
Also really want to get into Wowie's AweShader system for 3DL more which is why the need for lots of CPU cores and memory. Seen some really great stuff in the Show us more of your 3Delight renders thread over on the Art Studio Forum that actual comes extremely close to what Iray can produce.
...the first car I learned to drive in was a 1969 Buick LeSabre, a total land yacht. Later I inherited my Granddad's old 1964 Buick Special.
I find myself amused when I see people in a Prius or Subaru give up parallel parking (after several bad attempts) in a spot I could easily get that LeSabre into. They don't teach parallel parking anymore (and in my state no longer even test for it). That and backing up in a straight line were usually the two major fail points in the road test.
Not sure about operating a steam or even diesel locomotive even though it's on a track. Performing on the pipe organ was enough, dealing with a myriad of different controls for me.
It sounds like the sweet spot for me, too. As in I'm-never-going-to-buy-another-GPU-ever. I don't know about Iray, but with two 3090s in Cycles X, it is close enough to realtime to get the benefits of realtime, without technically being realtime. It's a crazy world when you have to disable a 2080ti because it's dragging down the other GPUs... "Faster" is no longer necessary. 24 gigs meand "Bigger" is just about no longer necesary, and 48 certainly would be. So if a single 4090ti is faster than 2 3090s, then I don't see anything that would drive me to want a 5090 at all.
But it'd be naive to think that NVidia is not purposely cooking up something as we speak to drive sales of these cards...
That comment was at least partly tongue in cheek.... you may have noticed that hardware improvements are usually followed by higher software demands?
...that's why I'm avoiding 8.1 and only have a few G8 characters (mainly the "interesting" ones I wish would have been released for G3, don't need yet another Vicky. Mike, Aiko, Josie, etc).
steam is more complicated, but at least its a two man job.. although it requires a good bit pf proper team work. Modern AC diesels are to some degree easier than the older DC ones. - more computer assist (wheelslip limiting, etc..)
Way off topic, but as long as we're talking about trains... When I was a kid in the 1950s, the trains would come through this town several times a day, in long, long, slow(because of the curve just west of town), assemblages of freight, tank, & coal cars with several locomotives and even a caboose now and then (you didn't want to be on the wrong side of the tracks if you were in a hurry). Back then there were even "Erie RR" passenger trains that still stopped in Jamestown, NY. But in my parents day (WWII era) the trains (freight & passenger) would even stop at this little village. However, these days we get mostly just one or two a day, usually short and mostly tank cars, just passing through, and this village's train station has moved to the center of town and is now the postoffice. The days of waiting for a half hour at the crossing are over. And the days of passenger trains in this area are over too. Gotta go to Buffalo or Erie or Pittsburgh to catch a train.
I've been recently interested in the construction of the new BrightLine passenger service between Miami and Orlando, Florida. This last week finally saw delivery of two more of the new train sets to the Orlando depot. The new highspeed tracks to Miami aren't completely ready yet, but the website below has a whole series of excellent video documentaries of the construction of the tracks and new stations. Service between Miami and Ft. Lauderdale(25 miles) has been available for several years, but the extension to Orlando (235 miles, 378 km) will supposedly open late this year or early next year.
I lived in Melbourne and Orlando area for a total of about 30 years so I know the area well and watch the documentaries to see what's happening to that area that I've been away from for 15 years.
The passenger station in Orlando is at the Orlando Airport and is just as impressive as the airport buildings themselves are and is connected to the airport terminal by local rail shuttle(horizontal elevator). Eventually there will be a further BrightLine extension to the Orlando amusement resort areas (20 miles), and then on across the state to Tampa (another 60 miles).
Check out the delivery of the new BrightLine trains to the Brightline maintenance depot in Orlando. America is finally getting new long distance trains. It's about time!
I'd be happy with the RTX 4070 since I don't do big scenes. But I can also go for the 4080 too I have the PSU to do so.
Freight trains can be pretty interesting. Me & a few of my cousins used to have a blast in 5th grade climbing on the caboose of freight trains that were stopped to allow other train traffic to pass elsewhere. Or some reason we just know they were extremely long freight trains that would be frequently stopped for even over a day and the caboose at the end of it would always be near where we lived by the overpass. It might be my imagination but one hardly even sees or hears freight trains nearly so frequently and certainly not nearly as long as those anymore. We used to get some of the flares off the cabooses and ignite them. I guess we would of gotten in big trouble had we got caught. Not many people know but it is tresspassing and illegal to even walk along railroad tracks in the US.
I will certainly be buying a 4090 TI eventually I think.
Oh, I'm sure they'll come up with something.
It's annoying really, now I have a 3090, I'm inclined to ignore AMD, even though I use Blender.
Yes, imagine how much a 980ti slows the render down. Then again, in Blender, my 1950x was faster than the 980ti - but slower in Iray, go figure!
I'm not falling for it, feel bad enough already (environmentally) when rendering just for fun with my little RTX 2060 mobile. And for me, it does the job very well. I'm not into huge scenes with large groups of G8, and since my upgrade from an external 1050 Ti wasn't that long ago, I'm still totally in awe at the speed of the 2060 ;-)
Yep, My IRAY displayport isn't real time yet ;p And the videos I've seen of VRChat.. look dated compared to what can be done in DazStudio and/or Blender. Pretty sure those 4xxx cards will be pushed to thier limits sooner than later.
...well tried to post the image of the old station in my former hometown but the site software seems to have derailed the attempt. the image is not very large, and not directly from, a Net site (it has been on my computer for a number of months now) but for some unexplained mysterious Daz site voodoo, the forum software just hangs in an infinte loop when I try to upload it. There was no errorr message or anything. I let it go for over 15 minutes and when I came back to it the status was still indicated as "Uploading...".
I even moved it to the desktop and resized it but, the same issue persiested.
I had no trouble uploading other images to another post, just this one.
Yes, the sad thing is that even I could build a superior setup with AMD parts, performance is no longer the driving factor. I'm watching compatibility with the software that I use now and may use in the future, e.g. Omniverse, which I have not even looked at yet.
Yeah, if they keep going in this direction, desktop/worstations employing "5000", "6000", etc., series cards will be built into special motherboards or have some kind of external solution using specialized port(s)(something like Thunderbolt?). The cards keep getting more power hungry, bigger, and heavier & it reaches a point where you can't plug such massive cards into your average PCIe slots that we currently use for our graphics cards.
I would expect the price to be above $3000, and with inflation in mind, I would venture up to $4000. Make no mistake, Nvidia is going to make you pay for that extra "oomph".
Or the GPU will become the motherboard with a PCIe slot for a "daughter" board or have the RAM, SSD, and CPU mounted directly on the GPU.