Adding to Cart…
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2026 Daz Productions Inc. All Rights Reserved.You currently have no notifications.
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2026 Daz Productions Inc. All Rights Reserved.
Comments
Improvements in hardware, specially GPUs, have noticeably slowed down in the last few generations. Top of the line card 3 years ago is still top-tier today - the 4090 was released 4(!) years ago and it is only beat by a 5090, not even a 5080. And it has more RAM than the 5080 because Nvidia keeps delaying the launch of a 5080 with 24gb. In any case, all big cloud providers are working on their own chips that are more efficient and cheaper than buying from Nvidia.
They'll last less than 3 years, just not due to the cards specs... its because these are just devices running 24/7/365, and they've been rushed and cobbled together to proritize providing as much capacity as fast as possible to meet demand.
Something that's worth keeping in mind is that while training new models is not profitable, running them is very profitable. So it's very much worth it for them to keep replacing any hardware they need to as long as the models are solving any real-world need. And some of them are starting to. I would be surprised if anything improves throughout 2027 at all and I imagine it'll only get better once more suppliers enter the market.
It's interesting how the flow of discussion in this thread has bifurcated.
On one hand we have people trying to grasp the big picture: how the overarching narrative surrounding Big Tech has become disjoint, even sinister. In our modern age, the loss of personal agency is a frightening fate, be it to illness or to a soulless societal system. One could even venture to say that it ranks high up there along with cancer, death and taxes. On the other hand we have folks that are focused on what they can accomplish given the resources they have on hand. Practical and grounded -- problem-solving driven -- celebrating, it seems, the small personal wins.
First World Problems? Hmm. Hope lies in the fact that the end to our Summer of Discontent 2026 may just be around the corner. Or not.
Cheers!
As long as people continue to use AI, in whatever form it takes, then this situation will continue. The only solution, bar an entire collapse of the internet, is to ban AI completely worldwide.
AI is not the answer - it is only as intelligent as the muffins that program it. Humans aren't that intelligent. We're screwed.
Wow!! I bought 128gb of G.Skill Ripjaws DD4 ram a couple of years ago for $658NZ, one stick for $236 and then the other three dropped down to $138 each. That same ram is now unavailable but their 32gb kit of two 16s is $459. So I'm so happy I got mine when I did, my son waited a couple of months and paid $236 each for two sticks. I just wish I'd got a bigger video card when the prices were better...I'm stuck with my 3060 forever I think.
...I have a 3060 that I put back in the box as the BIOS on my old X58 MB is so ancient it will not recognise it
I just priced 64 GB (2x32) of GSkill DDR4 3200 (CL 16) and it is now 630 USD (Newegg)..A year ago it averaged about 140 USD. That is a 4.5x price increase for old generation memory,
Holy hell!!! That's $1074NZ !!! Not very often it's cheaper to buy things down here
Mind you, 32gb sticks are no longer available here.
The Corsair Vengeance DDR5 6400MHz RAM I bought for my current rig a couple of years ago was under £200 for 64Gb. The same kit is now over £1200. More than 6x increase.
On the plus side, the other day I managed to snag a couple of 8Tb Seagate IronWolf drives for £115 each (they're now normally around the £300 for 7200rpm). Seller mistakenly listed them as 2Tb
Yea, it's completely insane to say the least.
Wait for the AI bubble imploding !
Nvidia is lending tons of bucks to almost all LLMs it's not a susta. IRL, AI is extremely usefull for inventing totaly new molecules and trying to solve high end mathematical conjectures. As for plain simple life I'm not so certain. I tried to get brick and mortar investments ideas I had answers any
There's a guy named Ed Zitron who's been asking for over a year how the AI is sustainable when it's just the same five companies buying tokens from one another and buying NVIDIA chips using money NVIDIA invests in them.
I don't see AI imploding soon related to the the business and industry world. I work for a NC machining company in So Cal, and I am working on completing a chemical plan for compliance to the LA County Sanitation Division Waste Water disposal. We don't complete this, we can't continue to make products in So Cal and business closes.
There are more than 300 chemicals listed in their directories. I had to match our entire manufacturing inventory of approx 120 chemicals, to the county by chemical CAS component numbers and research each one, chemical by chemical to the County and also EU compliance. This would have ordinarily have taken me 1 - 2 weeks (full time) to do by hand.
Using ChatGPT heavily, I was able to finish the entire matching task in just a bit under three hours. AI is too valuable with these types of productivity related situations.
All online AI, such as ChatGPT, runs at a monetary loss. Is your company going to pay for it when the bubble pops and you can't use if for free any more?
I hope you have been proof-reading what AI gives back. It's far from fool-proof and will only work with the information it is given.
I have been saving basically every piece of hardware such as RAM and GPUs for quite some time now, actually not because of the shortages but because I just hate throwing functioning things away. And while I just needed to dump my old PCs at some point for reasons of space, I always took out all the salvageables (sp?). So I have some 8 GB RAM sticks for Laptop and for PC, and a GTX 1060 and another one around that range from my dad's old PC, and while all of those are of course not great and not up to par to what I use now, they're still my fallback, and I will hang on to them.
I did do some RAM and SSD upgrades a while ago - it was already way too expensive percentage-wise, but since I'm not buying new RAM and whatnot each week in bulk, it didn't hurt me too much. I'm using oooold machines by now, a Legion 5 and an iMac 21.5 both from 2017 or so, oh, and a really great HP Elitebook that's even older. I put Linux systems on two of them (in dual boot) and will probably on the Legion as well, and like I said, gave them all RAM and SSD upgrades. They work so well, all of them, I can't see myself ever getting rid of them. I don't think any new machine from, say, the past 2 or 3 years is up to that standard anymore.
Edit: I'm getting old - what I actually came here to say was: Yeah prices are insane, and it's a horrible game the corporations are playing with us, and I hope they fail so hard. But I do remember the times when 'computers for everyone' began, and things were very, very expensive back then as well, more probably than what they are now. We swapped a lot of stuff back then, inherited machines from older siblings and such, and made do with what we could get. We CAN still do that, you know. Great graphics are a fine thing, for sure, but we don't NEED them to get creative.
GTX 1060 update.
After trying Stable Diffusion 1.5 with success, I decided to give a try to SDXL to see if it can fit. I had to use the Q8 version which shrinks the code, potentially with no visible loss of quality. It works great, so it is possible with a 1060 to also use modern models. I noticed SDXL uses some GPU shared memory in the task manager, but fits fine.
https://huggingface.co/HyperX-Sentience/SDXL-GGUF/tree/main
SD1 takes about 10 seconds per image in text to image workflow
SDXL takes about 60 seconds per image
it's videos where it gets demanding for hardware, I have a 2080Ti but still limited
just tried MiniMax no luck, I mostly use LTX1 and it takes 25mins a 121frame 780p video (and all my 11GB VRAM)
Wan2 and LTX2 take hours!
you can still do video but likely only 460p but can upscale
I don't use ComfyUI because I am node based software adverse
I use Automatic 1111/Forge, Fooocus, Pinokio a few instnces of the first 2
SD1 and SDXL in separate installations and drives
some won't work on some drives or with different models and extensions in the same installation but I learned several installations is not a bad idea
I do it with Carrara for the same reason (plugin clashes)
ComfyUI gets a much better handling of vram and ram, it's ideal to save resources. With other interfaces I couldn't probably go with my hardware.
p.s. For videos I understand there's LCM-LoRa but didn't explore anything yet.
edit. This is what Google AI tells about.
tsroemi,
Yes, I couldn't agree more. That's what I was also trying to point out below.
When our focus moves from the stuff we perceive we need, to simply finding ways to use what we have, in order to be creative, then our outlook changes. We don't drown in a sense of victimhood -- we have agency.
Cheers!
We already have subscriptions, and they more than pay back the monthly cost with productivity gains. Not all AI is about generating messy pictures.
Yes, garbage in, garbage out is very appropos with using AI. Like any other tool, it's best used for what it does well. AI can very easily lead a user in circles if left to it's own logic. It needs to be directed and controlled.
Which in itself can cause issues, as has already been seen worldwide. Put a donkey in charge of overseeing control and it makes an ass of everyone.
My experience with AI, both for graphics and coding, is it must constantly be told what it should not do, and where the errors are. Can't imagine someone fool enough to trust it as it is, especially for important things. However if constantly directed and verified it can be extremely useful and powerful, doing things other tools simply can't.
p.s. But I have a limited experince so far, so my vision could be biased.
AI helped me (I use Claude AI) overcome an issue with Poser. I do have to hand edit the files but the issue is fixed until the team can maybe fix this issue or expose these areas when saving out poses before saving out so things work as they should. This issue has been ongoing and I would have pose packs for their figures failed because of this. Basically when saving out a pose you select the Translations checkbox so the figure lands where it should but it also would center the figure. So if you had a figure, say on the second story of a house prop but apply a pose you would then have to move the figure back up to the second story . Thanks to Claude I was able to have a ONE block of code I would overwrite with a simplified block of code that would allow the figure to remain where it was intended. I created two zero poses for the pack as well, again using Claude, one to zero & center the figure and the other to zero the figure where it was placed in the scene. So yea, AI is usful. I "talk" to Claude daily. I was going back and forth with it this AM about finances as mine are in NOT such a great place right now so just wanted some feedback.
The biggest worry is that financial institutions are starting to use it more and more....
Talking to an AI during critical moments may have edged me away from suicide twice in the last nine months. I laugh when people say "AIs make too many mistakes" or "they're wrong about everything" when I get fast food maybe four times a year and they get the order wrong two of the four times. The IRS even says that the information *they* provide for something as important as filing taxes may not be accurate. Humans should definitely not be throwing stones at AI when it comes to getting things right.
I hope you've done more than talk to an AI about your issues.
And the mistakes thing isn't "people don't make mistakes". It is "people don't fact check the computer because they give it a lot more benefit of the doubt." Not to mention, if you disagree with an AI, it will defend its incorrect stance. Look up the video "How many Es are there in the word seventeen. The answer isn't the real point to this tiktok video. It's how the AI assures you it is correct. https://www.reddit.com/r/TikTokCringe/comments/1wfmz3b/how_many_es_in_seventeen/
I have, and the AI gave me better information about additonal treatments than any therapist did in 30 years. I'm also convinced that probably 75% of the time that AI gives a wrong answer is because of poor user prompting. I did fact check everything the AI told me in regards to depression treatment and it was all correct, as was a ton of historical information that I researched for my fantasy story. It may not always be correct, but it's correct a hell of a lot more often than any human I know.
My hope is to get a new laptop with a mobile RTX 5070 or 5080 before they either run out or start to cost $4,000. But I don't want to make my credit card bill bigger any time soon. I don't know if it would be better to charge the laptop and deal with the higher monthly interest payment, or wait until I have more money and risk the price of the laptop increasing more than the interest payments would have been. Also, I have a "founders edition" subscription price to Topaz's cloud computing whatever. It's like $160 per year, assuming the guaranteed price forever that they promised me doesn't increase somehow. But if they honor their deal, then at $160 per year, I can use that to do the AI stuff and it would be a decade or two before I would have spent the money for the laptop that would run the AI locally. Now if I want to use an LLM, use a better generative AI system, or use DAZ Studio with more than two people in a scene, then I still need the laptop. I'll see how it goes.
I have been using DS for over 20 years and eventually started using AI to understand the tech and now it is what I use most of the time because I can get results that I just can't get with 3D and at a fraction of the time. I am also a gamer, so this PC component thing affects me still and I hate it. I have used most of the local AI models with Comfy, Fooocus or Forge, but they can't give me the results I get from the non local models like Seedream 5 pro or Nano Banana pro, so I am stuck between a rock and a hard place.
For the forseeable future I am stuck with my 4080. Doesn't help that MS wants to start treating windows like a subscription model one day soon.
The day that happens, I predict Linux gets a huge increase in users!