A major Danish supplier Proshop, who also supplies other nearby countries, has released their supply numbers for RTX 3080, 3090 and upcoming 3070.
RTX 3080: Proshop has ordered 9 015 units. Out of these 489 have been delivered and further 372 are incoming: 5.4% of Proshop's order has been delivered and 4.1% are incoming.
RTX 3090: Prohop has ordered 1 972 units. Out of these 158 have been delivered and further 121 are incoming: 8% of Proshop's order has been delivered and 6.1% are incoming.
RTX 3070: Proshop has ordered 4 630 units. Out of these 151 units have been delivered and further 451 are incoming: Assuming the incoming units arrive by launch day, Proshop will have 13% of the inventory they'd have wanted.
The statistics are available here:
https://www.proshop.de/RTX-30series-overview
Comments
The 3070 numbers... well it's concerning, but seeing as how they haven't officially shipped yet, I don't know that it means a whole lot yet. If they can't deliver on 3070's either, it probably means something is up at Samsung, as we can't blame that release on a new immature memory standard.
Also Newegg tweeted that they have more 3070 inventory than they had previous 3000 series launches, but they expect to sell out in minutes.
EDIT: Also update on other 3000 GPU numbers: During last 8 days Proshop has received 162 more RTX 3080 and 27 more RTX 3090.
Assuming that the delivery speed stays constant, it would take manufacturers until December 2021 to deliver Proshop's current order for RTX 3080 and until May 2022 to deliver the RTX 3090.
"Be water my friend" - Bruce Lee
it does appear there is wider availability on the 3070, I've heard of many more people actually getting an order through than on the 3080 or 3090, but that doesn't mean it's good by any means - the 3080/3090 is a very low bar to clear.
It's Day 1 though, need to see if the log jam starts to clear up and ancedotal evidence of how many people actually claim to be getting orders in before I claim the sky has fallen. It's clearly a paper launch for 3080/3090... but hasn't been proven quite yet for the 3070. This will be a good indication of how much of the supply problem is Samsung and how much is Micron.
Will be interesting to see how AMD can do on Nov 18 - there will be a lot of demand for the 6800 and 6800XT as well. I did find it curious they left nVidia room to breathe with the 3070 and pricing on the 6800, but were very aggressive with competing with the 3080/3090.
Internet rumors say that TSMC was the problem. One of their production lines had something go wrong that destroyed all of the chips that were in production at the time. Coinidentally, that production line was producing Cypress chips (5800 series), while a different production line producing Juniper chips (5700 series) was unaffected.
Proshop has received average of 288 new RTX 3070 GPUs a week. At this speed their current RTX 3070 order will be fulfilled on March 2021.
The shortage isn't as bad as RTX 3080 and 3090, but it's still bad.
Imagine that you're AMD. You have an agreement to buy some particular number of 7 nm wafers from TSMC this month. TSMC doesn't care what the wafers are for, so you could allocate them as any mix of CPU and GPU wafers (or APUs or consoles or whatever) that you like.
Suppose that you can sell an average Zen 3 chiplet for $200. You can also sell an average Navi 2X GPU die for $200. An average Zen 3 wafer gets you 800 chiplets. Because the dies are much larger, an average Navi 2X wafer gets you 120 chips. You can easily sell all the chips you get as soon as you get them. How do you want to divide your wafer allocation between CPUs and GPUs?
This isn't a trick question. Once you parse through the situation, it quickly becomes clear that it's a huge waste of money to produce any GPU wafers at all until you've got all of the Zen 3 CPUs that you can sell. You could double the stated MSRPs of the new GPUs and it still wouldn't make any sense to build any until you've got all of the Zen 3 chiplets you want.
And AMD is going to need really a lot of Zen 3 chiplets. When AMD made their deal with TSMC of how many wafers they'd buy today, they weren't expecting that Intel would still be selling slightly modified versions of their 2015 product line as their state of the art. EPYC Milan is coming soon, and AMD surely doesn't want to starve a market that pays several thousand dollars per CPU.
I don't think that Navi 2X will be purely a paper launch, but I do expect a soft launch. A strictly paper launch would be too embarrassing for marketing reasons. And beyond being contractually obligated to provide particular numbers of chips to various partners, there are also longer term considerations that you don't want to make business partners hate you because you preferred to use your wafer allocation for a different product line. So AMD can't just stop making laptop chips entirely now that they're finally getting some real traction in the laptop market. But the consumer retail market for standalone GPUs? That's a much easier market to starve if you have to cut somewhere.
Let's put it this way: if AMD is able to keep Navi 2X CPUs in stock at MSRP before they do the same with Zen 3 CPUs, then that will mean that they goofed badly on their wafer usage.
I was skeptical of AMD's chiplet approach for consumer use, as you'll get a better product from a single, monolithic die. But the chiplet approach allowing the I/O die to be made on an older node that doesn't have the capacity constraints might mean that AMD can produce twice as many completed Zen 3 CPUs per TSMC 7 nm wafer than if they had a monolithic die. Quite apart from yields and cost of production, that's a huge win.
People mostly don't realize or appreciate just how tremendously profitable Zen 3 is. It's not just that it's a market-leading product. Zen 3 is:
1) a tiny die, at about 80 mm^2,
2) that readily sells for hundreds of dollars per die, and
3) in a huge market.
I can't think of a time that that combination has ever happened before. I don't just mean for AMD. I mean for any computer chip ever, whether CPU, GPU, FPGA, DRAM, NAND, ASIC, or anything else. There are plenty of times that two out of the three have happened, but two out of three isn't necessarily hugely profitable, or even profitable at all.
AMD is soon going to be pulling down multiple billions of dollars per quarter in net profit. And that's even though the console chips are only modestly profitable and their GPU division might not even be profitable at all on net.
Last Thursday they had received 100 units of RX 6800 and 25 units of RX 6800 XT.
Proshop has total of 22 RX 6800 or 6800 XT GPUs that are incoming (incoming means the manufacturer has confirmed to Proshop that delivery is happening). For comparison they have 753 RTX 3070/3080/3090 GPUs incoming. So it looks like AMD's availability is much worse than NVidia's and will remain much worse for at least some time.
There always seems to be a bottleneck,spend some $$$ here and then this game is struggling in a different area so spend more money.
The bottom line is that we seem to never get far enough ahead of the game designs to make our purchase worth it and to last long enough before devs are making new crappy games that require even more unnecessary horsepower to run.
I am not seeing games that are much better than what i saw years ago on PS2/3 platforms,so why are we now spending tons of money to play basically the same games?Oh we have higher resolution,we have some ray tracing,neither of which is a determining factor on a game being good or not.
Linus recently pulled out a 10 year old PC and was running top end games rather well so WHY?Well as stated you might have higher res some new lighting features,so after all these years basically the same games or worse but more things to spend money on lol.
Never forget 3 mile Island and never trust a government official or company spokesman.
Hmm, I agree with you, up to a point.
Part of that equation is what the user/player/consumer is willing to put up with. I built my current computer in 2014. I've got the itch to upgrade, but mostly just because I like to build systems, not because my computer really struggles with anything that I want to play. That said, I'm willing to turn down graphics a bit, I don't need 120+FPS minimum, and that has let my current computer stay relevant for a long time.
So, when you say we never seem to get far enough ahead to make it worth it, I would argue we have systems now that are worth it. The only thing the extra $$$ makes possible are the bits of eyecandy. For some people - the eyecandy is worth chasing the hardware. But I can't think of a single game released today that my 6 year old computer won't play - it's just a matter of what I'm willing to accept as minimally acceptable performance vs graphics.
For the most part, a high end GPU doesn't let you play games that a lower end one wouldn't. What it does do is to allow you to play those games at higher graphical settings than you would otherwise. It's fine if you don't care about that, but insufficiently powerful GPUs haven't meaningfully limited game mechanics in a long time.
The days of hardware leapfrogging itself and posting 100% gains every 18 months is long since past. Now we're on 10-15% gains, if we are lucky, and seeing anything significant come from increased parallelism.
Even in the GPU front, the latest cards are pretty much just throwing more cores at it. It's just that GPUs scale easier with more cores than CPU code does. Going from 128 cores on 90nm to 10,496 on 8nm in 14 years, while CPUs in the same time span went from 65nm 2-core Core 2 Duo to... 6 different Skylake refreshes
CPUs have pretty well stalled out hardware-wise, and GPUs are constrained by process density and power - that's the only thing keeping us from having an eleventy-bajillion core GPU right now.
So back to Crysis - sure, you could release a game that "pushes hardware" forward like Crysis of old. The difference between then and now is that then, you could expect hardware not only to catch up, but to greatly exceed the needed performance in a short period of time. Today... hardware is just inching along and for a game to push current hardware, it could be a decade before the hardware is able to get fast enough to catch up.
More info here:
https://hexus.net/tech/news/graphics/147074-nvidia-rtx-30-supply-issues-just-due-gpu-shortages/
"Be water my friend" - Bruce Lee
RX 6900 had pretty much only paper launch: Proshop received only 2 units.
During last 2 weeks Proshop has received 13 more RX 6800 XT cards and 5 more RX 6800 cards. At this speed it it will take manufacturers until 2026 to deliver Proshop's current order for RX 6800 XT and until 2036 to deliver their order for RX 6800.
And I did expect better from AMD. I had hoped that most of AMDs bluster and overhype that had got them into so much trouble with Vega had moved over to Intel, and with the recent success with Zen, I had thought AMD was starting to shoot straight with the consumer. When Frank Azor said they would be shipping RX 6000 product, I took that at face value, mostly because his name wasn't Raja and they did ship what they said they were going to with Zen each generation.
Now... I have been silent on AMD mostly. Not because I condone their actions - Hell no. I think they should fire Frank Azor, or at the very least muzzle him from anything public ever again. Especially after his "$10" and "I can hit F5" posts.
But I've been silent because I'm mostly numb to the situation now. You can't buy anything. Not GPUs. Not CPUs. Power Supplies are scarce. Consoles are non-existent. There are days my wife comes back from the grocery store and there's been another run on bleach and toilet paper.
But yeah... between the two GPU manufacturers, nVidia has been dribbling out product. Very slowly, but you can see numbers creep up - things like the Proshop release, people mentioning getting them in various forums, Steam hardware survey results, various AIB SKUs popping up at retailers (even if out of stock), etc. AMD is ... next to nothing.
Makes me think Quiz is probably right again - AMD only has so much 7nm wafer capacity on contract, and they are using that on CPUs, not GPUs. WHich wouldn't be so bad, except AMD (well, Azor) said they would be available - and that's what I am disappointed in most... not that the cards aren't there, but that AMD yapped off and it proved to be just another hyped up ball of nothing. I had thought AMD had learned that lesson, but apparently not.
Maybe AMD doesn't need GPUs as much as they once did to survive - they have a thriving CPU business right now... although I wouldn't hang my hat on that, that is very much just a flavor of the moment thing and isn't any guarantee of long term anything with Intel in the hunt, Apple throwing out the ARM bomb, and nVidia set to swoop up that ARM bomb.
But we need AMD to be in the GPU business. And to be competitive. If you want no better evidence... mainstream GPU prices have shifted up by hundreds of dollars over the course of just two generations. And there's the long rumored 3080Ti/3050 -- nVidia has no pressure to release either of those products now, and every incentive to hold them back until they can get their own logistics squared away.
But paper launches of runs of 2 cards doesn't give you competitive environment... Shame on you AMD, for letting Frank Azor run his mouth and toss out another hype bomb that you've let blow up in your face, after just getting out from under Raja and others trashing your reputation for years.
RTX 3070 availability seems to be improving: At the time of the update, they had total of 733 customer orders for RTX 3070 on queue and 543 incoming cards. ASUS seems to have trouble delivering RTX 3070 card for them, with a lot of launch day orders still unfulfilled, but for example for MSI the oldest RTX 3070 order still unfulfilled was only 12 days old.
It looks like if you're looking to buy RTX 3070 and are not particular about which model to get, it's now possible to get one with reasonable wait time.
For RTX 3080 the stock situation is still awful with a lot of undelivered orders since launch. For 3090 and 3060 Ti the stock situation is not as bad as 3080, but worse than 3070.
AMD cards are still super rare and will be for the time being: At the moment Proshop has 60 AMD's new generation cards incoming from manufacturers, compared to 1 192 incoming NVidia's new generation cards.
NVidia's CFO (chief financial officer) commented:
"We expect the overall channel inventories, meaning the inventories that are with our AIC partners [graphics card makers] as well as in our e-tail and retail channels will likely remain lean throughout Q1.”
For AMD their CEO commented:
“We are shipping lots of parts, and volumes in all segments are increasing, and that will happen through 2021. There will be tightness in the first half of the year..."
NVidia's financial Q1 is February - April. So at least 3.5 months more shortages for NVidia and 5.5 more months shortages for AMD expected, possibly even more. It's not looking good.
Source:
https://www.techradar.com/news/nvidia-rtx-3000-stock-may-not-improve-until-may-and-amds-gpu-outlook-isnt-great-either
All time classic MY NEW FAVORITE POST! (Keep laying those bricks)
"I should point out that no other company has shipped out a beta on a disc before this." - Official Mortal Online Lead Community Moderator
Proudly wearing the Harbinger badge since Dec 23, 2017.
Coined the phrase "Role-Playing a Development Team" January 2018
"Oddly Slap is the main reason I stay in these forums." - Mystichaze April 9th 2018