Nvidia Gives Lackluster Forecast on Inventory, Shares Plunge | Hacker News Reader

Nvidia Gives Lackluster Forecast on Inventory, Shares Plunge

bloomberg.com

54 pointsthrowaway575234 comments

I'm just excited that we might be able to buy graphics cards again at a decent price, soon I hope. The cryptoshit folks really inflated the price of graphics cards over the last few years..

There will be lots of used ones dumped on the market soon, presumably. From what I've read there's really no difference between a working used card and a working new one - no matter how heavily used they were. So there are bound to be some bargains to be had in the used department.

Having spent some time in the silicon manufacturing business side of things, I would have a hard time buying a chip that was run at 100% for 3 years with an expected life span (which is something chip manufacturers calculate) of 5 years (which is about average). The 'how long will this thing last' calculation almost never assumes 100% utilization.

My understanding is miners don't keep it at 100% utilization, because that is not energy efficient.

It takes a certain level of technical sophistication to go the extra mile and set that up.

Additionally, I've spoken with miners who rent with utilities included and simply let their mining rigs mine nonstop.

Oh this sounds good for people who play videogames.

They've been milking consumers for over 2 years and they try to continue doing it with the latest GPU iteration.

Hopefully this means that people is no that stupid and they are not to purchase whatever they release at those prices.

It's clear they were underestimating how much demand crypto was driving. They deliberately tried to limit how many gaming cards crypto people could buy. A predictable consequence of that is that crypto people would start to hide how much they were buying from Nvidia and Nvidia would lose their ability to tell who was buying their cards.

A coworker and I recently talked about building gaming rigs. Then we went to some benchmarks and all of the cards had problems rendering games with good settings at 4k - unless you’re willing to drop several k on multiple 2080s/Tis. Maybe we’re still in an odd place with expectations and costs.

Honestly I don't see a point to 4k gaming. High frame rate, on the other hand, is really great. Skip 4k and go for 144 Hz instead.

Don't really have much choice in a home theater gaming PC. My X900F does 4Kp60 with HDR but nothing faster. Since I sit at THX recommenced distance I really need a proper 4K render to make it look good.

Nvidia has an effort to promote production of high frame rate 65" gaming TVs for the living room. They call it BFGD, Big F[ormat] Gaming Display. Unfortunately they keep getting delayed.

I posted about this a few weeks ago; it's a catch 22 situation. Gamers won't upgrade until games and hardware support high frame rate 4K gaming. Hardware manufacturers won't provide hardware until there is demand. Average businesses will stick to 1080p until 4K monitors become cheap enough, and scale well enough (which they don't on anything other than Mac).

4k displays are cheap now, at least flat 60Hz ones.

On Windows 10, most of the apps I use scale just fine. Hi-dpi displays became common in laptops several years ago already, e.g. for 13" FullHD you do need scaling. By now, the majority of Windows software was updated accordingly.

I use a curved 40" 4k monitor. It's the same dpi as a 1440p 27" display and has the same real estate for productivity as 4 1080p monitors.

Then when I play games it melts my face off and is fucking incredible.

I really want a good 4k gpu

Current generation GPUs are too weak for 4K especially if you also expect high framerate for high refresh rate monitors. And practically no games are using recent multi-GPU features like in Vulkan. So you can either use 4K at lower framerates, or stick to lower resolutions for now.

4k high refresh rate monitors are prohibitively expensive anyway. From my experience a lot of games hit decent framerates at 4k already using last generation cards (GTX 1080/Vega 64).

I believe high framerate monitors are a niche market with a lot of hype but not a lot of buyers.

I'd say go with higher framerate over 4K for now.

I think it depends on what games you are trying to play, and how much graphics combined with high fps matter to you. I play Rocket League on a 4k screen with a 1060, and it holds up at 60fps without any problem. I also play GTA V, racing games, and the occasional first-person shooter, with no problems whatsoever. It's not the best for top tier competitive play in highly demanding games, but I think it depends on what you want out of it.

I agree, I think NVidia really stumbled with this RTX nonsense. I'd much rather have enough GPU cores to render 4k 60fps comfortably over some extra eye candy.

I was really looking forward to the 2080 release to upgrade my 1080, only to be disappointed by the benchmarks.

Hopefully AMD will take advantage.

I mean this in the most humanist way possible. God bless gamers. They funded a fabulous amount of research. A modern card compared to a voodoo2 is, well, incomprehensible.

Thank you for the future. You paid for it. I get the reward.

I’m still bullish on Nvidia because they essentially have the entire deep learning market. I think attempts from e.g. Google to develop their own chips will fail, just as they fail in every non-software venture I can think of, and even a lot of software ventures.

Interesting. Do you think this is more interpolability between packages, i.e. Pytorch not being written or having extensive testing for TPUs, or most deep learning problems don’t fit neatly into the exact problem domain TPU is specialized for?

> I think attempts from e.g. Google to develop their own chips will fail

I'm curious what you think failure will look like. In a lot of ways, Google's TPU program has already been extremely successful.

I've also been seeing lots of gpu databases kind of hitting the mainstream( Mapd, brytlyt , blazingdb, Kinetica ). Not sure how big of a market that is.

Nvidia is shooting itself in the foot with their Gsync strategy. They are relying on an extra $200 (for the Gsync module) from what - 1% of their customer base?

Instead of adopting the open standards (adaptive sync or HDMI 2.1 VRR) and adding support to all their cards.

1440p gaming really does well with adaptive sync, since you can less reliably hit 60 fps. A 50-75 FPS window that even the cheap Freesync monitors offers works pretty well.

Otherwise without adaptive sync, it doesn't make sense going to 1440p, and without 1440p, it doesn't make sense upgrade the GPU. Even a very cheap old 980 will run 1080p very well.

Xbox One supports Freesync (open standards adaptive sync), too. I'm wondering how long Nvidia can go against the current, but I guess 70+% market share is a tough nut to crack.

https://www.microsoft.com/en-us/p/xbox-one-x-1tb-console/8nq...

The real problem is that there are only 2 significant players in this sector: AMD and Nvidia.

Real estate of modern GPU chip has become astonishingly large and expensive. It will make absolutely no economical reason to have GPU separately from CPU. The future looks like desktop CPU and GPU will be placed on one die as it is done in every modern mobile System on Chip.

In the near future AMD has a better chance on desktop because of x86 cores used by desktop gamers, but Intel may start to compete here more seriously. Nvidia might be more successful on mobile/robotics/automotive markets with their ARM+GPU System on Chips, but Nvidia will start loosing desktop gaming.

Now data centers and AI. It is clear that CPU+GPU SoC is not a solution here because right now one modern compute node has 2 multi-core Xeons and 8 Voltas. Even if AMD will make very powerful SoC like EPIC+Vega right now, it cannot be sufficient for many compute-bound problems. This looks like Nvidia will still dominate in this sector for a while. But tendency will be to have 2-4 socket mainboard of powerful SoCs because it's more efficient in terms of money and energy.

I’m a bit skeptical about this report. Share prices drops after hours don’t necessarily mean drop in prices, if anything prices should be looked over a weeks time.

Another point, crypto was never a cash cow for Nvidia which makes their point quite silly. Spreads on power efficient cards were really small.

https://www.nasdaq.com/symbol/nvda/after-hours

7.5M NVDA shares traded after hours today. That's more than half an average day's volume.

Next quarter's guidance was really bad.

The shortage of high end parts due to crypto demand definitely drove prices up, inflating their income.

RIP my portfolio it’s at like 50% compared with 6 weeks ago.

Well, I'm no financial advisor, but keeping the bulk of your portfolio in individual stocks is probably a bad idea. Consider funds that track markets.

I still have a good ROI tho.

Just like every Bloomberg article, this is clickbait. Going down 16% in after hours trading is not unusual and words like "Plunge" and "Lackluster" are clearly designed to invoke an emotional response.

I don't understand how articles from this garbage site keep making it to the front page.

Reply on news.ycombinator.com