Jump to content

Talk:GeForce 40 series

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

RTX 4080 12GB SKU controversy

[edit]

There is no controversy, stop trying to manufacture fake one. 42.190.191.119 (talk) 19:45, 22 September 2022 (UTC)[reply]

please do not abuse your power to edit and remove content. if you can provide sources that rebuttal the claims of a controversy we will gladly listen but until then please do not attack the people trying to make sure this site is as truthful and neutral as possible. Tdfem 19:51, 22 September 2022 (UTC)[reply]

https://www.techpowerup.com/review/nvidia-geforce-gtx-1080/images/front.jpg https://www.techpowerup.com/review/nvidia-geforce-gtx-1080/images/gpu.jpg

GP104 was sold as 1080 class product, not the first time and won't be the last time 104 class GPU will be sold as 80 class card, again stop trying to manufacture fake controversy, AMD shill trolls that are extremely jealous of Nvidia's success. — Preceding unsigned comment added by 42.190.172.10 (talk) 20:03, 22 September 2022 (UTC) [reply]

So I've seen plenty of tech sites reporting with opinion pieces and user comments on those upset with the pricing. I think there's a legitimate-enough claim there to be worth covering, assuming the sources cited are improved (plenty of better options from legitimate tech publications rather than youtube/reddit). That is also assuming others feel it is useful for the article to have a section on discourse surrounding the product in the first place.

My problem though is then going on to mention the reduction in CUDA core count, which ignores the massive clock speed increase over the 30 series (~900 MHz, while clocks were mostly stagnant from 10 -> 20 -> 30 series), and the reduced bus width, which ignores the massive L2 memory cache increase of literally 16x. My point being that these products are not yet released and it's pedantic and extremely speculative to point out spec changes that may not really affect the end product in any meaningful way. At the very least provide a more neutral "some sources believe that the reduced memory bus width is not in line with the card's branding." --Bobrocks95 (talk) 22:30, 22 September 2022 (UTC)[reply]

I agree with that it's still speculation at the moment whether all of this will matter once benchmarks get released, if it gets shown that it really doesn't matter that these certain specs matter in framerate then it should definitely be either reduced in how bad it was actually or just put as a footnote that this controversy at least happened TurboSonic (talk) 02:09, 23 September 2022 (UTC)[reply]
We don't interpret or weigh the value of coverage in reliable secondary sources, if it exists it exists. If there are contrary opinions they can be given commensurate coverage in due proportion. If in the long run it is just a footnote that is fine, but it's not our place to decide it is merely a footnote because we feel in retrospect that it was misguided. —DIYeditor (talk) 02:15, 23 September 2022 (UTC)[reply]
If valid sources are discussing the bus width, core count, etc. as an argument for the series being improperly branded (and thus overpriced) in their opinion, sure. It makes sense that the coverage would be even for Wikipedia- this is something being talked about pretty widely after all, even though it is speculative. The section definitely needs cleanup and proper citation though. Bobrocks95 (talk) 03:47, 23 September 2022 (UTC)[reply]
I'm not sure why you're comparing the 30 series with the 40 series. The entire issue is within the series itself. And a massive reduction in the CUDA core count cannot be offset with higher frequencies of the 12GB SKU. We are just a few weeks away from reviews, so if the 12GB and 16GB SKUs have performance within 3% of each other, then this whole paragraph could be removed. As of now, it has caused a massive amount of negative press and comments. The title of this discussion is extremely derogatory BTW. I added the paragraph and I've been using NVIDIA GPUs since Riva TNT2. I wanted to be as unbiased as possible and added enough information to confirm the issue. If you hate the wording, you're free to change it but I see no issues with it whatsoever. It's just facts, nothing else. Artem S. Tashkinov (talk) 09:13, 23 September 2022 (UTC)[reply]
I also agree that Bobrocks95's eliciting 30-series specs was almost completely irrelevant. Core clocks between 3070 Ti and 3080 were quite close together, and so are the 4070 12GB and 4080 16GB. The fact that numbers have increased proportionately over last gen is only more ammunition in favor of the controversy here. Just really bizarre logic. A metal shard (talk) 10:34, 26 September 2022 (UTC)[reply]
The section has been largely rewritten (not by me) and it now seemingly addresses all the concerns voiced in this discussion. Anyways it's WP, so you're welcome to contribute. Artem S. Tashkinov (talk) 18:22, 26 September 2022 (UTC)[reply]
Due to the OP (IP editor)'s continued persistent behaviour of disruptive editing and attacking/harassing other editors, I have reported them at AN/I a month ago (link to archived thread), the result being a 2 year block for the main IP range they edit from (42.190.128.0/18).
Turns out this has been happening for several years (see the "list of incidents" table in the archived thread) on various articles, GeForce 40 series isn't the only article they have disrupted here. The behaviour even goes as far back as 2015 from what I've seen, see GeForce 900 series history here. Not sure exactly how long this has been going for, but that's the earliest incident I've come across so far.
Let me know if they come back as a different IP, evade the block and start disruptively editing / edit warring and attacking other editors on any articles again (not just GeForce 40 series), and I'll file another report at AN/I about it. From what I've seen they've also used the 219.92.192.0/19 range to briefly evade a block and disrupt once before.
AP 499D25 (talk) 00:29, 28 March 2023 (UTC)[reply]

Isn't "Canceled" the wrong term here for the "4080 12GB" as was effectively rebranded as "4070 TI" [1] and also apparent by the specs? I believe the table should reflect that, e.g. "released as 4070 TI". 81.217.246.229 (talk) 14:02, 24 May 2023 (UTC)[reply]

References

Products table: Launch next to Launch MSRP (rework proposition)

[edit]

Maybe it's just me, but launch next to launch MSRP looks strange to read (first world problems, right?).

Launch Launch
MSRP
(USD)

How about:

A) We merge launch parts? B) Remove launch from MSRP and leave launch date (or (C) change it to release date)? D) Merge both into a single column? E) Change MSRP to price or move MSRP column behind TDP?

Examples:

A: Launch B: Launch
date
MSRP
(USD)
C: Release
date
MSRP
(USD)
D: Launch date
and MSRP
E: Launch Price
(USD)
Date MSRP
(USD)

Rando717 (talk) 19:35, 9 October 2022 (UTC)[reply]

1. IMO this should be two columns anyways, not just one because otherwise you won't be able to sort by price :-)
2. The launch price being at the end makes a lot more sense, why? E.g. the GeForce 20 series cards were not available at MSRP for almost two years due to the crypto craze. No one cared about the launch prices.

To be honest, I'd move both the launch date and price to the very end/right. Why? Because mobile users have very narrow screens and these two pieces of info are hardly important. Wikipedia normally lists products which are/were available at retail/etail, which means the launch date is not too important and launch prices are only relevant for US consumers. The rest of the world lives by different rules. Artem S. Tashkinov (talk) 18:36, 12 October 2022 (UTC)[reply]

1a. My initial so called issue was with repeating Launch word and merge suggestion came from Amd tables.On Amd's side modelname and code name columns are merged, also date and price, so 4 columns cut down to 2 (in width).
1b. As for sorting by price, mostly (but not always) products are sorted from entry-level to high-end (or vice versa) and price usually follows it.Only dates are not sorted, but you can still sort by dates or go back to by model/price. :-)
2. I agree with you, I am from Europe and for me MSRP doesn't have much of a meaning (but it's not the reason why i would move it), IMO something like TDP or memory size are more important than code name, date and price (or even transistors/die size).
At the same time those five are the first columns that show.
Also when looking at list of gpus, prices are on the right.Even looking back at the 20 series (since you mentioned it), price was on the right side, but somebody decided to move it left and create this 3rows/3colums "mess" (IMO).
3. As for moving date to the end(right), I disagree and only because of one reason.When we look at gpu, cpu, chipset tables...release (launch) date is always first (actually second) column behind modelname.It feels like unwritten rule or something.
Based on that moving the date to the right would look off (IMO).I am aware that you shouldn't copy style from other articles or force same style across articles.But it's second column always...so I don't know. :-) Rando717 (talk) 12:35, 13 October 2022 (UTC)[reply]

Products

[edit]

I've got an idea. We have so many data points and so few products it makes sense to reorient the table and swap the products with their characteristics, e.g. transpose columns and rows. It will make the table 10 times more narrow but a lot longer which will aid reading and understanding. Also all the product names can be shown vertically. Artem S. Tashkinov (talk) 03:13, 25 October 2022 (UTC)[reply]

I'm talking e.g. something akin to the this GeForce 40 Ru Wikipedia. Casting @AzureNeptune:, @Purplneon486:, @TurboSonic:, @Rando717:, @Alexysun:. Artem S. Tashkinov (talk) 07:47, 29 October 2022 (UTC)[reply]

I stopped editing this article, ever since my last edit (when I moved price col and notes out of the way) was reverted with 0 explanations (so I can avoid making same mistake twice).
But I do have one concern about vertical sorting.What happens when rest of the series is released eventually?
I assume it's gonna look something like this GeForce 30 Ru Wikipedia? If so are we saving any space at all? Rando717 (talk) 08:16, 29 October 2022 (UTC)[reply]
This looks fine actually. Besides, like I've already said, product names can be listed vertically. This table looks a lot better than what the English wikipedia has. Artem S. Tashkinov (talk) 01:46, 31 October 2022 (UTC)[reply]
Casting @UKER: as well. Artem S. Tashkinov (talk) 01:48, 31 October 2022 (UTC)[reply]
It looks better to me, and solves having to have vertical/sideways text to avoid having wide columns for small data points. I suspect it looks better on mobile devices too. —Locke Coletc 04:12, 31 October 2022 (UTC)[reply]
Yeah, I didn't think it would look good, but it's not bad at all. --uKER (talk) 15:48, 31 October 2022 (UTC)[reply]

Nvidia 12VHPWR adapter

[edit]

Just in case anyone wants to write up about the OEM 12VHPWR adapters that came with the FE variants and third-party AIB variants of the RTX 4090, I've taken a few pictures of mine.

Feel free to go ham with these. --benlisquareTCE 08:32, 10 November 2022 (UTC)[reply]

Resurrection of RTX 4080 12GB (as RTX 4070 Ti)

[edit]

An AIB vendor - Colorful had updated the page that RTX 4080 12GB will be resurrected as RTX 4070 Ti. There's a rumor that Nvidia will relauch the RTX 4070 Ti on January 5th.

Link can be found here: [1] Nobi-Wiki (talk) 23:55, 7 December 2022 (UTC)[reply]

Inclusion of 'GeForce' in branding & model column

[edit]

Not having 'GeForce' in the branding and model column is like referring to Ryzen 5 as "5", or calling the Core i5 series a "i5", which can be a bit confusing and bamboozling to some new readers learning about the subject. It also makes it harder to distinguish from the Quadro line of professional GPUs.

Take a look at the box of any GeForce 40 series graphics card. Clearly you have on one line, "GeForce RTX", and then on the next line, the model number like "4080" or "4090", without the "RTX" brand name. For proof, go to google images, and search for "rtx 4080 box", "rtx 4070 ti box".

Lastly, here are three sources that show that the RTX brand name is trademarked as "GeForce RTX", and NOT simply "RTX": [2] [3] [4]

Extra one: [5] "GeForce RTX is a trademark and/or a registered trademark of NVIDIA Corporation in the U.S. and/or other countries."

If it was simply trademarked as RTX then it would make sense to omit the GeForce from the model column.

Hope that makes sense.

AP 499D25 (talk) 03:06, 7 February 2023 (UTC)[reply]

The 4080 card has TDP of 340W - or 320W?

[edit]

according to the German Wikipedia it is 340W and not 320W. Also Palit 4080 gamerock Omniblack has specification listed as 340W: https://www.palit.com/palit/vgapro.php?id=4658&lang=en&pn=NED4080019T2-1030Q&tab=sp

But looking at NVIDIA shop I see a Zotac 4080 to buy and the link from NVidia leads to me: https://www.x-kom.pl/p/1085927-karta-graficzna-nvidia-zotac-geforce-rtx-4080-gaming-trinity-16gb-gddr6x.html#Specyfikacja

where 320W is specified.

So... what is it now? 340W or 320W? Or both not? See:

"The GPU's power efficiency was positively received with Digital Trends finding that the GPU had an average power draw of 271W despite its rated 320W TDP" 81.161.204.104 (talk) 10:53, 13 February 2023 (UTC)[reply]

RTX4080 positioning

[edit]

RTX4080 with its current pricing should be considered as enthusiast segment card. Also it's mentioned that 7900XTX is the competitor and it's labeled enthusiast (see Radeon RX7xxx series Wiki). 87.207.179.21 (talk) 19:06, 25 March 2023 (UTC)[reply]

PCIe connection to PC not mentioned

[edit]

The article doesn't mention anywhere what kind of PCIe connection to the PC the card uses. As far as I know its 16 lanes of PCIe 4.0. Christoph (talk) 10:27, 6 July 2023 (UTC)[reply]

4090 ti/ TITAN ADA leaks?

[edit]

There have been several leaks about the 4090 ti/TITAN ADA for a couple of months now, should there be a line discussing it?

(https://videocardz.com/newz/never-released-nvidia-rtx-4090ti-aka-titan-ada-has-been-spotted-again-but-this-time-with-pcie-interface, https://wccftech.com/nvidia-geforce-rtx-4090-ti-titan-ada-graphics-card-pictured-once-again-massive-quad-slot-cooler/, https://www.notebookcheck.net/Nvidia-Titan-Ada-images-resurface-with-gargantuan-cooler-and-unique-PCB-design.728318.0.html) Tukok (talk) 22:19, 16 November 2023 (UTC)[reply]

Incorrect Super launch dates

[edit]

As stated in the cited source (https://nvidianews.nvidia.com/news/geforce-rtx-40-super-series), the 4070 Super launches on Jan. 17, the 4070 Ti Super launches Jan. 24, and the 4080 Super launches Jan. 31. 216.243.58.249 (talk) 22:21, 8 January 2024 (UTC)[reply]

Incorrect RTX 4080 Super Clock Speeds

[edit]

As per NVIDIA, the boost clock speed should be 2550 MHz, not 2505 MHz. See here: https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4080-family/

Click the Specs tab, then View Full Specs, and you'll see Boost Clock (GHz) is listed as 2.55, not 2.50/2.51. Likely just a typo by whoever entered it. FlyingSuicide (talk) 08:48, 17 January 2024 (UTC)[reply]

To add to this, if you look at any Gigabyte RTX 4080 SUPER product page you'll see they list their SKU's core clock with reference card in braces next to it. All are listed as " (Reference card: 2550 MHz) "
Here is an example: https://www.gigabyte.com/Graphics-Card/GV-N408SAORUS-M-16GD/sp#sp
In this example, it says:
Core Clock 2525 MHz (Reference card: 2550 MHz) FlyingSuicide (talk) 08:54, 17 January 2024 (UTC)[reply]

Addition of 40X0 Super class gpus

[edit]

add 4070 Super, 4070 Ti Super and 4080 Super to the gpus. 2A02:587:1A15:5900:C843:687D:502B:1039 (talk) 19:44, 22 January 2024 (UTC)[reply]

 Not done: the GPUs you talk about aren't notable for inclusion on the article. The 🏎 Corvette 🏍 ZR1(The Garage) 17:42, 23 January 2024 (UTC)[reply]
@The Corvette ZR1 ...what? —Locke Coletc 19:47, 23 January 2024 (UTC)[reply]
 Done by Rando717 with this edit. —Locke Coletc 22:06, 24 January 2024 (UTC)[reply]

AI modified images against MOS

[edit]

Several images on this article come from the user "marcusburns1977" on Deviantart, While they were originally uploaded as creative commons, they appear to have been modified with AI upscaling and monochrome effects, which are against the Manual of Style. RealNZZN (talk) 14:36, 2 February 2024 (UTC)[reply]

Layout Bug

[edit]

The most important table on this page, “Products / Desktop”, is unreadable because the tall header at the top right corner of the page makes the table very small in width. It only has 7 columns visible out of 21. The table has a horizontal scroll bar, however horizontal scrolling is borderline useless because it immediately hides the most important first column of the table, with the model name.

Tested in Edge and Firefox browsers, same issue. 79.140.150.233 (talk) 22:34, 14 August 2024 (UTC)[reply]

Introduced in revision 1237380748 (2024-07-29), with <div class="noresize">. ToraNeko (talk) 12:59, 21 August 2024 (UTC)[reply]

Mating cycles unchanged

[edit]

"It was also reported that the new connectors have a limited lifespan of around 30–40 mating cycles before contact potentially becoming unreliable.[68]"

To my understanding (well, I had a look at the specs) this is unchanged from 6pin and 8pin connectors. The required nominal lifespan was always just 40 cycles. Nonetheless this limitation becomes much more relevant with the too high ampere pushed through them now. The article should probably go into more detail here. --SchallundRauch (talk) 03:10, 4 November 2024 (UTC)[reply]