OCUK RTX4060TI review thread

[random_matt]

NVIDIA AD107 Silicon Powering GeForce RTX 4060 Pictured

The "AD107" is expected to be the smallest client GPU based on the NVIDIA "Ada Lovelace" graphics architecture. The upcoming GeForce RTX 4060 (non-Ti) is rumored to be maxing this silicon out (enabling all available shaders). MEGAsizeGPU scored one of the first pictures of the "AD107" in the...

www.techpowerup.com

Nvidia has a 65% profit margin apparently.

 

[Brackish]

Minus 64% after you factor in the leather jackets.

 

[CAT-THE-FIFTH]

NVIDIA AD107 Silicon Powering GeForce RTX 4060 Pictured

The "AD107" is expected to be the smallest client GPU based on the NVIDIA "Ada Lovelace" graphics architecture. The upcoming GeForce RTX 4060 (non-Ti) is rumored to be maxing this silicon out (enabling all available shaders). MEGAsizeGPU scored one of the first pictures of the "AD107" in the...

www.techpowerup.com

Nvidia has a 65% profit margin apparently.

Next quarter 70% and Apple is in the low 40% range.

 

[Minusorange]

Yikes

 

[CAT-THE-FIFTH]

Yikes

What's even worse is that the RTX3060TI can undervolt quite easily too,so the power difference is even less. You can get an RTX3060TI for around £300 now.

 

[eddyr]

As they say ship in ship out misquoted as before the watershed So far I've seen mentioned training on Reddit and Twitter data, I'm sure we'll get all sorts of pearls of wisdom from that

That reminds me of something I watched about how/why certain words can cause chatGPT to get confused and spew gobbledygook. I'll see if I can find it.

Here we go: Glitch tokens

Glitch Tokens - Computerphile

Language Models' Achilles heel: Rob Miles talks about "glitch" tokens, those mysterious words which, which result in gibberish when entered into some large l...

www.youtube.com

Whole thing is worth a watch, but from about 13 mins in they explain that it was eventually discovered that in the dataset used to generate word tokens there was some junk data from reddit and elsewhere, like debug logs etc, that made it in.

 

[FoxEye]

I own a 3060ti, largely because I was badly in need of an upgrade and managed to get one for a "good price" in the market at the time of purchase.

Which is to say I still felt a little ripped off, the 4060ti is a flat out insult.

Hard pass, I'm happy to wait a generation or two until the GPU world gains some sanity.

If I was buying now it'd be a 6700XT, after cashback they can be had for under £300.

The last couple years don't exactly fill me with confidence wrt that...

 

[Alt0153]

The last couple years don't exactly fill me with confidence wrt that...

nvidia created covid to push profits! they spend all their r+d budget in something called the hive

 

[Phixsator]

The 6700XT and 6950XT are a bargains compared to what the RDNA 3 and RTX 3000 and 4000 sells for. I'd never thought I would end up buying a new last gen product and even less that it would offer better price to performance value.. The market is "!#!" at the moment and the lower down the stack you go the worse it feels. The 4060ti is absolutely a garbage GPU and I really hope nvidia is punished for it. There is nothing preventing nvidia earning the big bucks on 4070ti/4080/4090 while still being the midrange champion offering sensible performance for sensible money in the 4060 catagory. Now you know why I absolutely despise Jensen as CEO and the rest of upper management responsible for this.

nvidia created covid to push profits!

they spend all their r+d budget in something called the hive

Haha.. then again.. if Jensen could ... he would

 

[Murphy]

The 4060ti is absolutely a garbage GPU and I really hope nvidia is punished for it.

It's not, it's just price wrong, very wrong.

 

[RavenXXX2]

 

[Phixsator]

It's not, it's just price wrong, very wrong.

I have to disagree with you on that. The memory configuration is not worthy of a x60 badge and I'm not even refering to the 8gb but the bandwidth issues due to a limited bus width. Then there is a lackluster PCIe config of only 8 lanes meaning any potentiel buyer better be rocking a PCIe 4 ready system so AM4 gen 2 chipsets and are is out of luck and the same for comparable intel platforms. It's the same move AMD made with the 6600 except its even worse now because the GPU core is clearly more powerful and therefor easier to bandwidth starve.

 

[RavenXXX2]

Nvidia packaged a mobile chip as the 4060 ti desktop class, the disdain is palpable.

 

[Murphy]

I have to disagree with you on that. The memory configuration is not worthy of a x60 badge and I'm not even refering to the 8gb but the bandwidth issues due to a limited bus width. Then there is a lackluster PCIe config of only 8 lanes meaning any potentiel buyer better be rocking a PCIe 4 ready system so AM4 gen 2 chipsets and are is out of luck and the same for comparable intel platforms. It's the same move AMD made with the 6600 except its even worse now because the GPU core is clearly more powerful and therefor easier to bandwidth starve.

Having those doesn't make a card garbage, it displays an image on your monitor and allows you to play 3D games, it doesn't constantly blue screen Windows, it doesn't catch fire or anything else like that. Having 8GB, being bandwidth limited, using 8 PCIe lanes doesn't make it garbage. It makes it a 'low end' card that should at least be $100-150 cheaper.

If all those things made a card garbage then you wouldn't have similar configuration cards and it wouldn't even be worth buying if it cost $50, that's obviously not the case.

 

[Brackish]

How much difference could in increase in cache size have in future games and would rebar benefit from it too?

 

[Aegis]

How much difference could in increase in cache size have in future games and would rebar benefit from it too?

The increased cache allows the 40-series to read/write to the VRAM less often than the 30-series and scales across the line-up (the 4090 has 72mb vs. the 4060 Ti's 32mb) - its inclusion is one factor that allows the 4060 Ti to be competitive vs. the 3070 even though the architecture is massively cut-down in comparison (another factor being the 4060 Ti's 1GHz higher core clocks). It's not something that will be utilized any more in the future than it is now - it allows the 4060 Ti to reach the performance it's currently reaching - nothing more.

Unfortunately Rebar seems to be an afterthought for Nvidia - they added support when AMD announced SAM support on their cards/motherboards but on Nvidia it's based on a whitelist in the drivers that Nvidia rarely seems to bother updating these days (and any gains have generally been an absolutely maximum of around +5% performance but typically it's less).

 

[CuriousTomCat]

NVIDIA AD107 Silicon Powering GeForce RTX 4060 Pictured

The "AD107" is expected to be the smallest client GPU based on the NVIDIA "Ada Lovelace" graphics architecture. The upcoming GeForce RTX 4060 (non-Ti) is rumored to be maxing this silicon out (enabling all available shaders). MEGAsizeGPU scored one of the first pictures of the "AD107" in the...

www.techpowerup.com

Nvidia has a 65% profit margin apparently.

That's gross margin and it was 56% when I last checked. However, net margin is only about 16% which is the same as what they were making 12 years ago. If 16% is STILL all they make then the current prices are because it genuinely costs much more money to make a GPU.

 

[CAT-THE-FIFTH]

That's gross margin and it was 56% when I last checked. However, net margin is only about 16% which is the same as what they were making 12 years ago. If 16% is STILL all they make then the current prices are because it genuinely costs much more money to make a GPU.

Their current Gross margins are close to 70% and will be 70% according to Nvidia next quarter.

If you are talking about the yearly net figures,it usually is between 25% to 35% net margins:

NVDA (NVIDIA) Net Margin %

NVDA (NVIDIA) Net Margin % as of today (November 15, 2024) is 55.26%. Net Margin % explanation, calculation, historical data and more

www.gurufocus.com

That is higher or on par with Apple:

Apple (AAPL) Operating Margin %

Apple (AAPL) Operating Margin % as of today (November 15, 2024) is 31.17%. Operating Margin % explanation, calculation, historical data and more

www.gurufocus.com

Last reported quarter was almost 30% net margins which is around Apple level.

That 16% figure is for last year and Nvidia overproduced too much stock for crypto(probably):

Nvidia Stock: Inventory Glut And Revenue Decline (NASDAQ:NVDA)

Nvidia's Q4 revenue decreased by 21%, significantly impacted by its gaming market platform which saw a YoY decrease in revenues. Read my thoughts on NVDA stock here.

seekingalpha.com

Consumer dGPU sales are down and there was over $5 billion of built-up inventories. That most likely was factored into last years margins as a GAAP cost,which is hinted at late last year.

This has to do with a softening consumer market and too much inventory. It also affected BOTH AMD and Intel too because of unsold inventory. AMD did something similar last quarter too AFAIK.

PCMR has very short memories. In 2018,Nvidia overproduced too much Pascal dGPUs for the mining boom of 2017,so had similar inventory build-ups. So Turing was also released at a relatively high price.

Everytime dGPU sales have dipped a large amount and there is a large backlog of inventory,Nvidia has gone and released a sidegrade generation.

The question is,if the consumer dGPU sales slump continues will they refresh the range and drop margins a bit,or try and produce less dGPUs at a higher margin? We will find out in the next 6~9 months.

 

[Nexus18]

What's amd profit margin now? Given how they and their fan base keep banging on about chiplet design allowing them to make the gpus for dirt cheap, they must be making a killing?

Unfortunately Rebar seems to be an afterthought for Nvidia - they added support when AMD announced SAM support on their cards/motherboards but on Nvidia it's based on a whitelist in the drivers that Nvidia rarely seems to bother updating these days (and any gains have generally been an absolutely maximum of around +5% performance but typically it's less).

That's not quite as true now ever since nvidia released their fine wine driver for dx 12/rebar, which closed the gap significantly in titles like assassins creed valhalla (which many put down to being amds sam/rebar implementation....). Some games there is a huge boost e.g. dead space remake but yeah you have to manually go and enable it.

Dead Space Remake gets a 25% free performance boost with Resizable BAR

We can now confirm that Resizable BAR can increase performance by up to 25% in EA Motive's latest horror game, Dead Space Remake.

www.dsogaming.com

In my testing, the boost was huge:

Dead Space (Remake)

Gone back 4 pages and can't find the answer - how long does it take to complete the game if you take your time over it? I’m 4 hours in and only level 2 still. I’m tippy toeing around, getting lost and finding all types of goodies.

forums.overclockers.co.uk

Do agree it is an afterthought in terms of the whitelisting etc. however, it's probably for the best as it can cause some issues in games, hub even showed some issues on all amd hardware with rebar too.

Essentially rebar just remove the 256mb temp swap buffer to allow direct access for the cpu to vram.

 

[Murphy]

If 16% is STILL all they make then the current prices are because it genuinely costs much more money to make a GPU.

The rainforest's net margin are something like 1%, does that mean it genuinely costs much more money to run a multinational technology company?

Or is it that they're using the profits from selling books to branch out into selling more than just books?