Nvidia - A New High in Low Morality
Vložit
- čas přidán 25. 09. 2022
- ♥ Check out adoredtv.com for more tech!
♥ Subscribe To AdoredTV - bit.ly/1J7020P
► Support AdoredTV through Patreon / adoredtv ◄
Buy Games on the Humble Store! -
►www.humblebundle.com/store?pa... ◄
Bitcoin Address - 1HuL9vN6Sgk4LqAS1AS6GexJoKNgoXFLEX
Ethereum Address - 0xB3535135b69EeE166fEc5021De725502911D9fd2
♥ Buy PC Parts from Amazon below.
♥ NEW USA Store! - www.amazon.com/shop/adoredtv
♥ Canada - amzn.to/2ppgYsX
♥ UK - amzn.to/2fUdvU7
♥ Germany - amzn.to/2p1lX6r
♥ France - amzn.to/2oUAK2Z
♥ Italy - amzn.to/2p37Uui
♥ Spain - amzn.to/2p3oIBm
♥ Australia - amzn.to/2uRTYb7
♥ India - amzn.to/2RgoWmj
♥ Want to help with Video Titles and Subtitles?
czcams.com/users/timedtext_cs_p... - Věda a technologie
He’s back, not the hero we deserved but the hero we needed.
Nothing less than a knight, Scottish.
Boy you're in for a show tonight
Just when he thought he was out THEY PULL HIM BACK IN!
Amén 🙏 ❤
Indeed
I find it funny Jensen is acting like the new 4080s cost him loads when the dies are only 295mm and 379mm down from 628mm even if wafer prices are doubled it doesn't cut into his leather jacket margins too much.
MLiD thinks TSMC gave Nvidia a special offer Jensen could not refuse, after all they offended Samsung and new AMD would go brrrrr on N5
All because Nvidia refuses to cut margins so the execs will just have to buy one less yacht this year lol
The wafers are $16000 a piece while the previous generation were in the $5k range.
@@VoldoronGaming they are 9k each
@@VoldoronGaming from what ive seen, the yields on the new process is higher and its yields is expected to increase.
Plus with the smaller die sizes its possible the price increase is much less than it is on paper.
Heck, if you compare the 3080 and the 4080 12gb it's 46% of the die size and the yield rates are better. And the actual gpu die is only a small part of the cost of the gpu.
If anything the prices on some tiers are not going to be costing nvidia anything more than the last generation.
Especially considering the 3000 series gpus got released in a time of major memory shortages and price increases in that area.
If it cost nvidia $100 to produce a 3080 gpu die, I wouldn't be surprised if the 4080 12gb die is $120... and you're paying $200 more for that luxury.
Of all the tech channels, Adored seems to be the only one to acknowledge NVIDIA's rising profit margins in all of this. Other channels seem to justify these price hikes stating rising bill of materials costs or acknowledging that board partners are barely making a profit; they fail to recognize NVIDIA taking an ever increasing cut of each sale generation over generation. They pass the extra costs on to board partners while raising prices.
We can only hope that consumers will slow down their purchasing of these cards and force NVIDIA to bring their margins back to a sustainable level.
Another argument that's gotten very popular over the past year or two is, "Of course it's more expensive! It's faster! It has more transistors!" If you apply this logic back to the beginning, computers would cost billions of dollars. A related argument is rising wafer costs with new nodes, but, trust me, hardly any business is going to sacrifice margins to ideologically chase a node shrink, especially a fabless company like NVIDIA. There is a reason to upgrade or they wouldn't be doing it.
Hmm, that doesn't jibe with what I've been seeing. Hell, the whole EVGA debacle pointed out how Nvidia is slowly but surely moving to become the sole supplier of Nvidia cards (if Jensen gets his way.) The only thing (for now) that is going to reign in Nvidia is if AMD has a winner in thier RDNA3 GPUs and don't decide to match Nvidia's prices. Otherwise we have to hope that Intel stays in the marketplace and over the next couple of years they can have a competitive product with high quality drivers. A third player in the marketplace would do wonders to keep both Nvidia AND AMD in check.
Market sets the price. As we saw during the plandemic, system builders will buy en-masse to reduce the per unit cost and people will be looking there to get a new card. Lots of people on multiple generation old GPU's still, due to these historic high prices. Lots of long time working class gamers have simply been priced out, continue to be priced out, in the West. But there are the Indian and Chinese middle class there to splurge silly money on graphics cards. Nvidia will simply divert their cards East. Money is money, and Nvidia love money. Not to mention Nvidia got a huge chunk of that $50bn handout from Biden for the microprocessor industry Stateside, they will be in no rush to sell, being 1/2 of a duopoly after all.
In a World with +15% inflation Nvidia wants to make +80% profits yoy.
"Of all the tech channels, Adored seems to be the only one"
That is not true. Others have pointed this out as well.
For example Jaystwocents talked about his in one of his EVGA videos. I say that, although he is not my favorite CZcamsr :)
Nvidia slowly builds up their own sales distribution and eventually just will kill off its "partners".
When I saw what happened to EVGA, all I could think of was your videos from over the years. Can't wait to see the chiplet stuff on GPU finally come to fruition.
RDNA 3 is introducing it.
If it works, we'll see it even more
@@joaogabriels.f.5143 If it didn't work,it wouldn't have been coming out in a month. Rest assured,it just works.
Chiplettes are inevitable because what constitutes a wafer defect is ever expanding as feature size shrinks. The real question is will GPU's get to peek chiplette before CPU's get to the same levels of parallelism. Both core counts and the width of SIMD seem to keep increasing, after all. Software rendering is already very much back if you want to do it.
@@adnan4688 Which calls into question the entire evga narrative, they had an option to keep going, instead they went out screaming like a baby. Inflation renders the rest moot.
@@churblefurbles I am referring to chiplet part of OP's comment. What does EVGA have to do with it? Do you think that EVGA is changing colors or something?
I just rewatched your hour long history of Nvidia's envy from 3 years ago, what an absolute treat to see this. Thank you so much Jim.
I've watched that one a handful of times now because I needed some specific piece of information from it... I should just take notes, but... I can also just watch it again :P
About bloody time.
Good to see you back, Jim.
At 192bits, the 12GB 4080 is more in line with the x060 lineup. A complete insult to gamers and professionals... but then again, Nvidia's stinginess with memory continues to mystify me. I've been a customer for 6 generations, but AMD has won me over with RDNA2, and with their RDNA3 lineup, they have a lot of flexibility with fabrication allocations, because they can switch between CPU and GPU production, and also, they aren't dealing with a ticked off TSMC. It also looks like RDNA3 cards will be cheaper to make. NVidia over-produced, because their greedy execs apparently thought the crypto-mining boom (and those juicy whale customers) was a sustainable growth model - now they expect the gamers to make up for their gambling losses. Nvidia's C-Suite is apparently so stupid, they don't see the coming wave of used GPUs flooding the market, competing with their overpriced and obsolete 3000 series cards (which is a problem THEY created), and arrogant enough to think AMD won't be able to capitalize on Jensen's idiotic pricing.
Make no mistake: There was no SHORTAGE of GPUs during the pandemic - only a lack of availability for gamers. Miners got theirs by the pallet - and Nvidia was more than happy to sell them. Nvidia sold more cards per month during the pandemic than ever in their history - I'd even be willing to bet that rate increased right up until December of 2021. All those used cards are about to swamp the market and directly compete with Nvidia's "new GPU stack" that attempts to price the 4000 cards as high end options on that stack.
so true. Jensen sold directly to miners at a highly inflated price.
Pepperidge Farm remembers the bulk listings on eBay where the sellers would include photos of skids, sometimes an entire warehouse, as proof of the cards' existence. Around this time major retailers were reporting getting mere dozens of cards. I think some European vendors reported something like 8, lol. I wonder if any of them are as pissed off as eVGA is. You have to consider all the shelf space, advertising, customer service, etc. dedicated to selling product that doesn't exist, and you know at least some executive somewhere witnessed the pallets on eBay or Alibaba.
it is in fact a 4060 in disguise. you can see from the naming scheme hikes starting at the 700 series with the introduction of the 80ti models, and then the 3000 series with the 90 and 90ti models, which are all really renamed 80 models. Of course, the specs only go to prove that. paying close to a thousand bucks for a 60 series card is just completely unreasonable.
@@dycedargselderbrother5353: thanks for the context! Makes a lot of sense.
All of this
Damnit Jim. I didn’t realize how much I missed hearing your analysis of technology. Thanks for making a great video.
I feel same. I love his videos
Same here, His perspective is very unique and I definitely missed his videos
I just want to second this comment! Like a breath of fresh air. 😮
God man... I remember why I started watching you. Love it. Never stop doing it. Just like Tom from MLID, you've been more right than wrong by a huge margin on your predictions and, like Tom, you've got another follower for life. :D
Jim has another channel called Freethinker, and I wish he would do a deep dive on...
plasmasphere (aka Van Allen Belts) which enters into the magnetosheath the region of space between the magneto- pause and the bow shock (roughly between 15,000 to 25,000 miles where you start to reach the bow shock) of a planet's magnetosphere.
What I'm getting at is a conversation about electromagnetic shielding and why humans can't travel beyond that point.
Nvidia put all their bets on crypto still being a thing at launch of the RTX 40 series. No doubt the fanboys and whales will still buy them and scalpers are salivating at the thought of hoarding them. I would dearly love to see Nvidia get a black eye for their anti-consumer practices but we all know how their customer base are.
Lmao I'm going to wait 4 to 6 months before buying and see what's going on if the new nvidia gpu deserved to be bought
I've been with nVidia since the 6800, only once going Red with a 4850 x2, and after this launch, I'm done with nVidia. All PC games are anymore are console ports with a few barely-noticeable upgrades, so I see no point in spending so much on a gaming PC anymore. At least AMD isn't ripping off its customers.
crypto mining has ended bro, the only mining done at the moment is bitcoin, and bitcoin is using ASIC miners, not graphic cards.
Dlss 3 and improved ray tracing. They are innovating ahead of competitors. That’s going to get a premium price. And has no answer for dlss 3 frame Generation.
@@shashank1630 dlss is just cheap smartphone tricks, no one who buys a 4k monitor would ever embarrass themselves by using dlss
Great video, Jim. The 4080 12 gb is going to get savaged in reviews and I doubt they will sample it.
Like the review cards won't ship with a sweaty wad of money and a wink...
@@TheVanillatech ugh true
@@christopheroconnor4665 They all take the money, except a handful of CZcamsrs who don't need it. Even worse now Biden just handed out 50bn to Jensen and his cronies.
@@TheVanillatech I mean they can but the charts and the price will speak for themselves. Both 4080s are a scam and dogshit value, the 4090 is trash value as well but at least it provides a generational leap unlike the disguised 4060/4070
Morality doesn't get you leather jackets now, does it?
Jim I hope you are doing well. You're a good man
I am also a good man.
Figured an AdoredTV video was coming soon after GTC. Thanks AdoredTV, have been revisiting the history lately...
Given the 4080 shenanigans this video was an inevitably 😉
You mean 4070😜
"4080 12GB" has pretty much the expected characteristics of a x060Ti card regarding memory bandwidth and percentage of cuda cores compared to the flagship. That thing is a disgrace, and an insult to everyone.
@@odaharry You mean 4060! 🤣
Great to hear from you again, Jim, best tech-content provider out there.
is it 2018 again?
this is the type of video that the tech world has been missing. the question is, will they listen this time?
Linus poo-pooed the pricing on the WAN Show but good money is on him jumping around and shouting "Insane performance! Buy now! It's a STEAL even at $3000! WHOA!!!!" in the actual review. JayzTwoCents, too, but that's a given.
@@dycedargselderbrother5353 I do trust the Gamers Nexus and Hardware Unboxed reviews though.
@@eudyptes5046 Agreed, those two are two of the best.
If the past is anything to go by, absolutely not.
If this video is not a good reason enough for people to wait and see what AMD has to offer I literally have no hope for the PC Master race
Best part of my day, a new video from AdoredTV
this one's 25 minutes, this seems promising (really hope this isn't a one off even if the frequency isn't very high)
Jim has another channel called Freethinker, and I wish he would do a deep dive on...
plasmasphere (aka Van Allen Belts) which enters into the magnetosheath the region of space between the magneto- pause and the bow shock (roughly between 15,000 to 25,000 miles where you start to reach the bow shock) of a planet's magnetosphere.
What I'm getting at is a conversation about electromagnetic shielding and why humans can't travel beyond that point.
@@AwesomeBlackDude he'd probably get cancelled
@@AwesomeBlackDude you think we havent been to the moon and mars?
@@AwesomeBlackDude I believe the Apollo missions to the moon relied on minimizing exposure to the Van Allen Belts by tracking it and traveling through the thinnest regions of it for just a few minutes and developing the best electromagnetic shielding equipment available at the time in their spacecraft.
It would be easier to do it once again. I'll remind you the entire international scientific community + competing nation states including the Soviets were tracking the spaceships carefully with telescopes the whole way through (especially Apollo 11) and would've called out any foul play by the US so they wouldn't get to claim victory over the space race.
I'm happy to see you back for another video. Hope you're doing good and we can get some more videos from you when you're feeling well enough to make some.
I was hoping you'd return for this season of tech. I really needed that, "Alright guys, how's it going?" today.
I missed this sooo much. Don’t ever leave again Jim!!!
Jensen like: "what if Moore's law, but price instead of performance 🤔"
This is the best video I’ve seen covering the shortfalls of Lovelace! Those two graphs at the end do such an amazing job at providing context for the history of the x80 class gpus and how the 4080 12gb stands out so clearly. This video is great for people like me who haven't been following the market for too long.
Nvidia learned for their "mistake" with the 30 series. The 3060 Ti, 3070 and 3080 blew away the Turing cards because they were worried about RDNA 2. Even though RDNA 2 offered better preformance/value for rasterization - it had almost no impact on their ability to sell the expensive 3090s. The decreased value of the 3060, 3070 Ti, 3080 12GB and 3080 Ti reflected Nvidia's lack of concern about competition from RDNA 2.
Even if AMD release a 7800XT for $799 that matches the 4080 16GB in rasterization, Nvidia won't give a jot. If the market does start to lean torwards AMD they can more easily trim down the prices, like they have with the 30 cards in the past month. I have no idea how badly Nvidia would have to lose in rasterization value for their cards to not sell out.
I expect the 102 die will soon become exclusive to the 90 cards, so they can charge >$2k for a 102 die. 90 is the new 80.
Sad but true
It'll take beating them with a new version of FSR for Nvidia to start losing sales to the competition, but not only that since people are just complaining that AMD doesn't have CUDA for work. They're also consistently complaining about AMD's encoders. Wanting something on par with Nvenc.
The problem is even if AMD did significantly beat them in value, it probably still wouldn’t hurt Nvidia’s sales much. I’m sure you’ve seen “the GPU war is over” too. Nvidia, their marketing and their loyal fans have destroyed the GPU space. There will be no good value card from AMD because they’ve given up and just follow Nvidia with their pricing now.
Great to hear from you again, i had been regularly checking your channel to see if i missed a notification but I know you wanted to take a bit of a break, luckily I checked the day before you uploaded and decided to check the next day, hope to hear more from you soon. Keep up the good work.
the man the myth the legend.
I honestly blame the consumer for actually buying this crap.
PC gaming has a ton of new blood after the pandemic, unfortunately of the dumb kind that will buy new iPhones every year even when the processor is the same.
This time they will pay for their low morality. Raising prices during an economic downturn is business suicide. EVGA's departure is timed perfectly.
EVGA turned out to do the right move since at this rate we're all going to need new power supplies lol
idk people were buying 2070 and 2070 supers like it was good value. And everyone always has an excuse for why they wont buy amd. I only buy nvidia and apple but waaah waaah monopoly.
@@Heatranoveryou I bought a 5700 XT just because I was priced out of the 2070, and now I currently use a 6800 XT as my GPU
@@Carnyzzle 4090ti and 7950X combo 💀
Those GPUs are not for people like you or me. NVIDIA sells those GPUs also to people like for example video designers, streamers, architects, universities, government agencies, weather stations and so on. Those people have the money for the newest and even if its only slightly better its still better. That's enough for those people. 2k$ isn't much for a big streamer or architect or Hollywood video designers. For example the Oscars this year. All 5 nominations for the Oscars in best visuals this year used NVIDIA GPU and Software for their movies. All 5 nominations! A lot of big architecture companies use high end NVIDIA GPUs for CAD programs and more. NVIDIA knows that those companies have the money for high end GPUs. So of course they use it. Another aspect is that NVIDIA was affected by the trade war between USA and China this quarter. -400Mio$ for NVIDIA because of this trade war.
BUT the more important part is that NVIDIA uses other companies for manufacturing those chips. The 4000 was made in Taiwan from TSMC: TSMC ist hard affected by the energy crisis (high energy prices) and climate change (less water). Do you think that TSMC doesn't want more money?
"Taiwan will raise electricity prices by an average of 8.4% starting from July 1 in response to rising global energy prices including a 15% increase for large industrial users, the Economy Ministry said on Monday." (Source: TAIPEI, June 27, Reuters)
ALRIGHT GUUISEEEE
Nice to see your back. I have long appreciated your insight into the PC Gaming industry. Hope you stick around and keep making videos.
I can absolutely see nvidia slashing their prices in one year and acting like they are a savior of the poor and we should thank them.
Thank you. Especially the last two charts were real eye openers.
Jim, it always brings a smile to my face to hear your voice even after the six years that I've known you. You're still the best and always will be! Cheers!
I literally just watched some of your old videos yesterday for nostalgia and you upload the next day 😂
So glad to hear that voice and that accent again in a very well made and informed tech video!
I can only hope we get more!
Love your analysis videos. Always welcome, always enjoyed. Please keep doing what you love.
Nice one Jim, really enjoyed this, looking forward to your take on AMD and Intel's positions.
I currently have a 1080 that I bought almost 6 years ago. I bypassed the 2000 series as a point of principal when Nvidia raised that pricebar (RTX 'tax'), and bypassed RDNA2 and the 3000 series because...well, need I say? With the insane 4000 pricing just revealed, it's stupid power requirements, THAT 4070-->4080 insane switch, I ain't one bit interested in Nvidia.
Add to that Nvidia's utter determination to totally destroy that 'bastion of gaming performance', FPS, with that DLSS 3 fake frame injection bullsh#t and we all have some serious issues ahead in GPU land.
It's damning on Nvidia and AMD that they have not persuaded me (and many others) to part with available cash to upgrade for two and now, possibly three generations.
Ok then, it's over to you AMD. You have two choices...
1) Aggressively undercut Nvidia's 4000 pricing (as you did with Zen2), recognising today's insane cost of living, and win significant market share and hearts and minds.
2) Price 'effectively' the same as Nvidia, displaying to the world that AMD does not recognise Nvidia's shocking pricing or consumers cost of living crisis, and thus prove that AMD and Nvidia are clearly working together to price-fix, and to insane levels.
AMD, do you, as you keep saying, "have gamers at heart"?
[Interestingly, my 1080 still performs very well at 1440p with the latest games! - that 1080 was indeed a good purchase, thanks Nvidia ;) ]
Given how AMD announced price cuts for current Radeon GPUS I think they might be going for the former rather than the latter. AMD did something similar with the launch of alder lake. Even if they keep the prices as they did before I think AMD will take a chunk of marketshare
Everyone complains about nvidia but that never translates into sales for amd so I can't blame them if they stick with nv pricing...
@@Ammothief41 This time could be a real difference though. I think I can say most PC enthusiasts don't want a space heater in their room just so they can have extra performance their monitors can't display. Current gen GPU's from NVIDA and AMD already offer amazing value, hence why NVIDIA still advertises 30 series alongside 40 as "cheaper" options. If AMD's newest chiplet GPU's don't try to put the pedal to the metal and eat watts and shit temperatures like Ryzen 7000, Intel 12th gen, and NVIDIA 40 series all while providing easier to swallow prices. Then it won't matter if the performance gains over the previous gen are big or small.
With crazy clock speeds and temperatures being pushed through shrinking nodes this seems to be a heavily experimental new tech generation being desperately marketed as "better" when in reality the new Performance gains aren't that noticeable compared the hole in your wallet from buying the thing and paying your power bill.
I hope Intel 13th gen and RDNA 3 offers a better value proposition, but given the trend for this generation so far that may not be the case. With chiplets though for AMD it does give me hope. Time will tell.
@@Koubles Great analysis but I believe you're overthinking it.
Nvidia is about like Apple with their branding and features and if you want the real thing then that's what you have to get.
All the 4000 series means is that there's going to be lower cost 2000 and 3000 cards hitting the market.
They won't even consider that amd exists regardless of the price.
@@Ammothief41 Intel branding didn't stop AMD from taking a large cut of their market base with Ryzen. I don't think it will be a landslide because as you said, NVIDIA is like Apple in that they are the first brand people think of for graphics cards. However I think we'll see a rather significant shift in market share for AMD if RDNA3 doesn't crap the bed.
I was hoping to see a new video on the industry soon, so good to have you back Jim!
Thank you Jim! Great to hear your voice again!
Hey Jim it’s good to see you back I’ve always loved your videos and hope you make more. Thank you!
*sees adored notification*
*comes to video*
me : say the line 🥺
Jim : *says the line*
👼👼👼
Your videos are like an oasis in the CZcams desert.
Welcome back Jim and thank you for an other great video …we ll be waiting for the next one
I keep thinking back to Jim's old analysis on Nvidia, and pretty much all those points have been echoed by others recently...
For Nvidia and AMD indeed more cash is the way forward lol.
Jim is back, and I still revere him as the best Tech Reporter in the Industry. All the love in the world goes to you Jim. You are an amazing gentleman.
Yo Jim, it's good to have you back. Love your work mate.
is it just me, or is it that every time a gpu launch follows a mining wave, we get absolute crap? nvidia gets a year or two of great profits there, and then they try to carry those margins over to the next generation, just expecting us to swallow it. this time, they even flat out refused to make lower-tier products and are just selling the overproduced 30-series inventory for a while, for the same msrp those were released at two years ago
i hope their own cards on the used market kill this whole thing off.
Good to see Jim back ,the ball is now in Radeon's court on how they will win this gen and still make profit probably we will see newer segment cards such RX 7970 XT and RX 7990XT being the ultimate card to take on and beat the Nvidia's ultimate gaming card , interesting generation this time
I honestly still find this a stretch. Many newer people building their first PC are going to still buy an Nvidia GPU as Nvidia has been known as better for year and any guides they find will suggest one. We also know similar happened with the 20 series and while sales were lower, they still made a decent profit.
Lastly, I highly doubt AMD will be the good guys. AMD has continually claimed to be "leading the industry" and as a result raising prices across all their markets for higher profits. AMD has continued to either slowly lower the price vs Nvidia if it is slower and slightly raise it if faster. I highly doubt the rage from fans will make them change their minds about it. While it may not get them instant satisfaction, the idea people are stuck with these prices will make people suck it up and give in to these obnoxious prices, even if they really shouldn't.
No the ball is on US not to buy any Nvidia GPU no matter what. Death to the fanboys.
The new Nvidia cards don't fit my sff PSU and case. No thanks!
AMD will be getting its dick wet from Nvidia, while Intel Arc is raughdogging them from the back this generation.
Maaaan it's si good to hear from you again. Hope you're well Jim 👍
Awesome video. Glad you're back with these videos.
Aaaaaand....
"The RTX 4080 12GB is a fantastic graphics card, but it’s not named right. Having two GPUs with the 4080 designation is confusing.
So, we’re pressing the “unlaunch” button on the 4080 12GB. The RTX 4080 16GB is amazing and on track to delight gamers everywhere on November 16th.
If the lines around the block and enthusiasm for the 4090 is any indication, the reception for the 4080 will be awesome."
😂🤣😂😂🤣😂
for dam near 3090 prices lol no way in hell
I think this 4080 12GB will be tarnished forever among PC enthusiasts, so it would make sense for NVIDIA/AIBs to push this to OEMs.
wassup Jim!! how lovely is ho hear from you mate!!! we missed you!!
Awesome to hear from you again Jim!
Ah the 4080 12gb the new consumer champion, I have no doubt it will be tearing up the steam ownership survey soon enough. Everyone knows Lovelace is a turkey but nvidia are going to market the hell out of it, the tech reviewers will be pushing it very hard.
Really looking forward to whatever info you have on RDNA 3 and your take on it, because from my armchair, it's looking like AMD is going to absolutely destroy nVidia this time around.
Just look at Navi 32, which the leaked specs show as having 7680 FP32 units - dedicated ones, not the half-and-half of AD104.
Ampere performance shows that the dual-function ALU's contribute about 25% on average to gaming performance at 4K (highest I saw reported in any game was 40%), which is easy to compute just by comparing the performance of the 2080 Ti to the 3080. Both cards have the same CUDA core count and run at the same clock speeds, but the 3080 has the extra FP32 capacity. So as a general rule, to get the effective shader count of an Ampere card, divide the marketed "CUDA core" count by 1.6. The performance shown by nVidia, in the only three games not using trickery, suggests this hasn't changed at all. But let's pad their numbers. Rather than 25% of that extra FP32 capacity being useful in games, let's say it's 50%. So divide the "CUDA core" count by 1.33 instead.
That means the 4080 12GB has an effective shader count of 5760. It has a rated boost speed of 2.6GHz, but ever since they added their auto-overclocking feature, nVidia have never given accurate boost figures. Let's assume the card actually runs at 2.8GHz. That give it an effective compute capacity in games of about 32.25 TFLOP's (more than the 3090 Ti, but we're pretending the 4080 12GB is better than it actually is here).
Now let's go back to Navi 32, with it's 7680 FP32 units. What clock speeds will it have? Given how high RDNA 2 clocks at 7nm, and how high Zen 4 clocks on 5nm, I think it's beyond reason to believe that AMD won't clock higher. But let's pretend it'll only run at 2.8GHz as well. That gives it a compute capacity in games of 43 TFLOP's.
So using those guesstimates, which are more favorable to nVidia than I think is remotely realistic, we get a full Navi 32 card being 30%+ faster than the 4080 12GB in games.
And how much does AMD need to charge for that Navi 32 card? The total cost of the GCD and MCD dies is around $25 higher than the cost of the GPU die of the 6700 XT, which was launched with mid-pandemic with a FUBAR supply chain at $479. Today, I think AMD could make a sizable profit pricing that Navi 32 card at $500.
So it looks to me like AMD can put out a $500 card that will be 30%+ faster than nVidia's $900 card.
It also looks like a full Navi 31 will be more than 20% faster than the 4090 with a total die cost $100 lower, but I'm less confident in performance scaling as the shader count goes so much higher. The new Infinity Cache is supposed to be much improved, but that's still a lot of shaders to keep fed with data.
It just looks like AMD could crush nVidia on performance while undercutting them by a massive amount, but I can't help but wonder if there's a catch, or if the suits at AMD will utterly fail to recognize the potential in going for volume over individual unit profit.
they will price match nvidia and take in the cash.
always amazing to see you posting Jim, hope you are doing well!
The post title cracked me up. You really have a way with words and I like it!!
As always, Jim nails the analysis and sadly the reality of nVidea marketing being parroted by mindless tech reviewers. Keep shining the light Jim. We need to see clearly.
I'm soo onboard of the AMD train right now. Chiplets, efficiency and trying to beat 2 giants in the form of Intel and Nvidia that in the past (and present) have tried to destroy their competition with the worst practices ever, how some of the stuff they did was even legal and how AMD is still able to do what it does I can't really understand. I just hope AMD can take over more mobile and server marketshare (zen4(c) seems to be able to do that) and really hope they have an answer for CUDA/OptiX because Nvidia has an almost unbreakable armor in terms of that and the deals they can make with it (Vulkan/OpenCL/SYCL ?). I will still remain (perf/watt)/price but yeah... Lets hope Intel can make their GPU's competitive as well, I guess...
Zen 5 just came out, and I bet it'll shake things up going forward.
There is also ROCm for gpus.
great to hear from you again Jim.
Glad to have you back Jim, great content, as normal - No one does it quite like you do! Excellent.
From memory FP32 was locked on Ampere to 1/2 of full potential and only available on professional cards. Maaaaybe if they use full AD102 to make Titan this time, instead of 4090 Ti, maaaaybe we will see it unlocked with drivers. If it doesn't require different die altogether.
Also - congrats NVIDIA, you managed to drive out your closes partner EVGA. With friends like this who needs enemies?
It depends on AMD. Tom from MLID has an nVidia source who says they won't release a titan unless they are absolutely sure they will stomp whatever AMD can throw down. They do not want headlines about the Titan losing.
@@Hotrob_J Oh, I know. I watch MLID religiously.
@@Hotrob_J well there wont be a tital then.
@@Hotrob_J how sad, they won't produce it if it actually loses. They just want to maintain the illusion that there can never be a titan killer.
Wait wot, I only remember around 10 years ago when they locked down or got rid of FP64 in customer cards ad now the customer cards are even too fast, so they lock down the FP32 performance? o.O
Been expecting this ... 😁😁
The 3090 Ti set an abysmal power & price : performance ratio, now they pick that ridiculous SKU to justify RTX 40. RTX 4078 192bit bus is -10% on a '90 in previous generation, around a 3080 so definitely only an '80 for $899 reasons.
I am wondering how GPU chip scarcity turned into massive stocks of cards at AIB partners and apparently warehouses with stored GPUs chips within a few months Q2.
MLiD talked often of "supply chain management" so possibly this price gouging tactic has backfired now.
They were selling by fucktons to miners and artificially limiting supply at retailers to drive price up, choosing to stockpile in their warehouses instead. Today we see the results of their 200IQ play.
I always miss you man. And when ever you upload its like Christmas for me. I love you commantary.
Yesterday I checked the channel cause I thought I hadn’t seen a vid from Adored in a bit. Great to see he’s back!
Founder's Edition cards are still available at Best Buy. It's comical that these used to be "Limited". Their AIB partners can't be happy that even at the end of a series run Nvidia are still in direct competition with them. Crypto sent them into a feeding frenzy and they can't even distinguish between friend or foe.
The real reason why EVGA said bye bye.NVIDIA is taking manufacturing to monopoly level.
There's a scary thought, triple slot gpu size for the founder edition of the 4000 series and you really have to wonder how big the custom cards are going to be like, the card is too big and probably going to be bigger once the partner cards are out, they also could be more expensive to boot lol.
Everything I'm seeing about the new gen from Nvidia is looking like once massive blunder and It's going to be interesting to see if the fanboys fall for it again, especially with them naming the 4080 12 and 16GB models the same when looking at the specs, the 12GB model is more like the 4060 or 4070 but with a much higher price tag, this is all designed to push up the lower tier cards to the price point of the tier above and this is Nvidia treating its loyal fans like idiots.
This is a carbon copy of what Apple did to smartphones, keep pushing the price up, create a cult following and keep squeezing them hard by bumping the price up gen on gen, loyalty does pay and that blind loyalty that many gamers had to Nvidia is paying hard against them and if they are not careful, this could really hurt the PC gaming industry as it's pricing it out for most gamers.
It's going to be interesting to see what AMD does, because if they deliver a doubling in performance, they have a massive opening here to score a PR win against Nvidia and it could move the needle in AMD's favour to gain market and mind share which is desperately what they need and all AMD has to do here is deliver, offer an attractive price point and some good marketing as Nvidia are messing up so badly here that it's more or less giving the game to AMD, let's see if AMD are smart enough to see it and take advantage of it or will they do what they've been showing the last two years and get greedy.
I do know one thing, if AMD charges a little less than what Nvidia is charging for the next gen of cards, Nvidia will be laughing all the way to the bank because they know it's not going to change anything on the market share, if AMD really wants to change the game and gain market and mind share, they either have to blow Nvidia away with much better performance or they have to offer a much better price point, they've not done either the last few gens and the market share has not moved.
So this is your chance AMD, don't mess it up, you've got a chiplet design, likely a smaller die, AMD likely can afford to sell them much cheaper than Nvidia can, this is the right time to think big picture and not short term profits, gain as much market and mind share as they can because that will deliver a lot more profit long term.
As for those partner cards, they've only got themselves to blame for getting into bed so much with Nvidia when really the smart thing for them to do is to push rival cards hard in the market with the aim of balancing the market so they can play them off each other, all they did is become the lapdog of Nvidia and Adored did warn them about this years ago, maybe now they might start to learn but I doubt it lol.
The custom cards ("AIB") have been announced, and they are indeed huge, 4-slot monsters. (GamersNexus did a video detailing some of them). Although there will apparently be some AIB cards at MSRP, the "desirable" cards will be 10-20% above MSRP
To put it simply, NVIDIA went big, expensive, and power-hungry. Even if their efficiency is theoretically pretty good RDNA3 even if it doesn't offer a 2x performance bump(AMD is known to hit their targets though...) is expected to offer basically disgusting levels of power efficiency.
AMD doesn't necessarily have to take the crown to take a W even if they can win. They just need to offer competitive cards which don't break your power supplies, your breakers, and any need anyone has ever had for a space heater. People are aware now more than ever of how much power really costs.
AMD aren't exactly the good guys either, but NVIDIA is pretty decidedly still behaving worse, and hopefully NVIDIA gets a nice black eye from this one.
4 slot AIB cards, have yet to see a 5 slot produced. Yeah the cards are getting so big it'll be better to run them externally, especially if you want to mount any other PCI-e device.
When I saw this video pop up on my feed, I initially thought it was just CZcams recommending an old video. But what's this? It was posted today!?. I click, and my ears are blessed with the dulcet tones of "all right guys, how's it goin?" And suddenly, it was a good day.
Glad to hear back from you good sir.
For me as consumer bigger problem is that Nvidia is selling 2 cards with same name (only different memory size) which are not same in performance. For me is misleading, and this should end in court IMO. What chip name is, bus size etc is not important from this perspective, only that they misleading consumers.
Macbook Pro 13 inch is vastly different than the 14 & 16 inch. Yet they're both Macbook Pro. I think Nvidia will get away with it tbh.
@@kaleofthegods different size in laptops matter, because you have different physical space to works with. In this case they would not have this excuse. But it all will depend if someone go with it to court and how good layers they will have. I think in EU it would be easier. Time only shows.
1100 euros for worse than 3090ti performance... yikes. I bought a 6750xt for 450eur last week, since there's nothing priced sensibly where I live second hand (or new) and mid range rdna3 shouldn't come out until next year plus we'll get screwed by currency conversion rates as well.
Btw really nice surprise to see another upload from you Jim, I was literally rewatching a video of yours about undervolting the rx 480 an hour ago, that I had originally watched 6 years ago lol, since I'm using an rx 570 atm. Really looking forwards to that GPU chiplet speculation video ;).
Great to hear you again!
Regards Jim, good to hear your voice.
Can you make an video on AMD - how they promised more cores for less, but here we are - zen4, no increase in cores for desktop/mobile user while paying more.
What are you talking about?
Zen 4 significantly batter than Adler Lake and Zen 3 , aside the group of games which greatly benefit from 3D VCache on 5800X3d.
7950x down on 5950x
7700x down on 8c 7800x
Others same price despite inflation.
You're ignoring that CPUs generally fall in price after a year or so, the next generation recoups development costs by restoring the amounts.
Who would use a cheap box cooler with a thermally managed 105w high end chip?
@@tyaty I'm talking about core counts. The more you reduce the process, the smaller the cores get = the more cores you can fit on a chiplet = the more cores consumers get for the same money. Why hasn't there been an increase? It's because Intel doesn't compete and AMD won't say no to crazy profits.
He said at 24:34 that this video is too long already and he's working on a special video just for amd.
@@sulev111
You get diminishing return over 12-16 cores. Ask anyone who has a Ryzen 9 /TR system , how often they actually utilize all cores. As an early adopter of 3900x, i can confirm that not very often outside of workstation type loads.
For mainstream usage 8 core is the sweet spot.
Also while N5 is denser than N7, chips of of same size are significantly more expensive than on N7.
I would love an EVGA 7900xtx
Welcome back, after enjoying your content for sometime, I finally subbed, and liked ofc! , don't vanish for so long next time :)
OH GOD IM SO HAPPY TO HEAR YOUR VOICE AGAIN! Life is to short I miss your videos!
Don't forget that most of the gains that will come for the 4000 series, will be thank to crippling performance on previous generations due the upcoming drivers.
I half believe this too, but it's still so hard, if not impossible, to find any evidence from testing over the generations that actually stay 100% the same besides GPU's and their driver. Windows updates alone can drastically change things sometimes (and usually not for the better).
my 780 ti got gimped. I was gaming at 1440p all day. I done an update and I list about 30% in raw frames.
@@AceStrife as a 1080ti,2080ti, and 3090 user, I can confirm that it always happens. That's why I'm going to wait a see how good is RDNA3 and jump ships.
EVGA were the only reason I was a returning Nvidia customer. I'm gutted they won't be moving to AMD so sadly it's the end of an era... It puts AMD cards back on my radar but losing that good Evga warranty is a disaster for me.
Awesome to see a new vid from you, I'm looking forward to your take on the chiplet GPUs
OMG!! Jim, I'm so glad you made this video. While watching the RTX 4000 Livestream, I kept thinking about what your take on this would be.
It has been 5 years. 5 YEARS. WELCOME BACK!
Great video. Glad to see you back.
I re-watched your video about the 3090 a day or two ago, seeing as to how relevant it remained.
Glad to see a new video from you.
Waiting so long Jim, finally we got this type of video again :)
Glad to have you back with another banger. Great insight and information. Looking forward to your thoughts on RDNA 3. Angstronomics has a pretty good looking spec list for it that seems as close to factual as we've seen thus far this year. If Sky Juice is correct, the full Navi 31 die (7900XT) will be on a 384-bit memory bus and the cut down version (7800XT?) on 320-bit.
@AdoredTV thanks you for the effort put into this video as always.
I'm glad you have kept doing tech videos. You're the only youtube personality that makes anywhere near the research effort I do, thus you often reach the same conclusions. You doing these videos is the only thing that keeps me from doing them. And yes, that does make you Cassandra. But keep it up anyway. Some of us are paying attention.
gained a new subscriber with this one, cheers from USA!
Hiya Jim! Good to hear your voice...hope you are in good health and spirits these days :)
God, it's good to hear your analysis and voice once again, Jim!
I can't begin to explain how happy I am to hear his voice again. Made my eyes water
Glad you're here Jim
My hat is off to you and your whole team for this video. You have really gotten your message across with this one.
Thanks for using my picture from twitter at 7:49 :)
Long time no see, thanks for the video!
Great video as always. Thanks Jim 👍 😊