[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology

Name
Options
Comment
Verification
4chan Pass users can bypass this verification. [Learn More] [Login]
File
  • Please read the Rules and FAQ before posting.
  • You may highlight syntax and preserve whitespace by using [code] tags.

08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
05/04/17New trial board added: /bant/ - International/Random
10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
[Hide] [Show All]


Janitor acceptance emails will be sent out over the coming weeks. Make sure to check your spam folder!


[Advertise on 4chan]


File: 1763985773471974.jpg (41 KB, 738x384)
41 KB JPG
https://www.youtube.com/watch?v=-fZM9wOvbh0
>>
>run in the US
>ceo is indian
if intel cant break into the gpu market then rajesh wont either
>>
>>109161221
>uses RAM
>>
>>109161221
DoA
>>
>>109161221
lol no
>>
>>109161221
Jesus fucking Christ, after sucking valve's dick over their shitbox now this. GN just keeps getting worse.
>>
>>109161221
Why does it have a PCIe connector on both the top and bottom of the card? Is this an AI render, or is that an SLI-style interconnect?
>>
>>109161488
This is what turns you off from Shanghai Steve and not the incessant political spam in the past year? The valvebox content was a breath of fresh air more than anything.
>>
>>109161508
I didn't watch any of his politically charged videos, I only care about benchmarks. I have seen the thumbnails of his "investigations" and that was already enough for me.
>>
>>109161221
>6 years in the making
>Only have a render and some code
Vaporware. Very disappointing Steve is blowing smoke up his ass.
>>
lol gamer's nexus fully jumping the shark and giving a platform to a blatant scam
>>
>>109161726
nah, they covered nvidia products before
>>
File: 45654646.jpg (905 KB, 1920x1080)
905 KB JPG
>indian ceo
>zeus
>>
after watching a 10min of them talking in front of a rendered image I truly didnt understand why gn made the video, or why he lied in the video titles about it being for gamers

Literally the only audience is venture capitalists who might fund the company with half a billion it takes to rnd a chip and get it to tsmc and start marketing it.
Theres tenstorrent that made gpus for ai acceleration so its not impossible if you have jim keller money.
>>
>>109161499
>SLI-style interconnect
this
>>
>>109161975
Because Jensen is living in his head rent free and despite what he claims, he is a very dishonest and biased youtuber.
>>
>>109161260
ddr5 is too slow for both gaming and AI. Might as well use iGPU for games and cpu for AI then. It will be about as fast.
>>
>>109161221
Just use the huawei one if you’re okay with lpddr
>>
>>109161240
Most tech CEO's are Indian, they are just far better at running businesses than whites and thats just the truth
>>
File: 14075.jpg (21 KB, 204x246)
21 KB JPG
>>109162262
Only the shit start ups that fail in a year. The ones that stick around are white, which is what anon was noting.
>>
>>109161221
why the fuck does it have 2 pci express ports? unironically who the fuck asked for that? What kind of retard would use laptop ram on a GPU? And not even HBM? The fuck? the answer is no, this will be dead on arrival. Kill yourself street shitter.
>>
File: il_1080xN.994299053_29s7.jpg (175 KB, 1080x1105)
175 KB JPG
>>109162262
>and thats just the truth
Get better bait
>>
>>109162295
>Only the shit start ups that fail in a year
nah, those are usually helmed by white people and the occasional chink
>>
File: 1778496539657193.png (243 KB, 720x822)
243 KB PNG
>>109161221
>400GbE QSFP
>BMC
>LPDDR
>SODIMMs
>gaming
>>
>>109161221
Alibaba frankenstein hardware
>>
it's funny that they think they'll pitch this to offline 3d render people when:
a. most renderers don't even bother supporting amd gpus - cycles might the only exception. it's just not worth the development effort.
b. the industry is in the midst of a terminal decline as AI is beginning to eat its face and smaller shops are actively hiring 'AI native' artists
>>
File: 1779327600369084.jpg (134 KB, 1070x1070)
134 KB JPG
>>109161221
That fat fuck is truly desperate for clickbait. And here I thought the 3D printed steam machine front panels were already a new low.
>>
File: needmana.png (444 KB, 436x474)
444 KB PNG
What the fuck is this unholy combination of SBC, SOC and GPU that I am looking at.
>>
>>109162506
It's probably only going to get so much worse.
>>
>>109161221
The most obvious startup scam company.
>>
>>109162638
It's a render to scam uninformed investors out of their money.
>>
>>109161488
Ever since the RAM crisis started he became unbearable and I ended up unsubscribing after like 6-7 (haha) years. Now that he's platforming this scam that I remember hearing about a few years ago and clearly seeing how bullshit it is, I don't know how lower can he stoop from his once-"prestigious" title of a "trustworthy tech journalist".

Worst part is that he has a loyal army of ballwashers that'll call you a corpo shill every time you dare to speak against their "Tech Jesus". At this point I hope he smashes his dumb neanderthal face while MTX-ing, assuming he still does that given how fat he's gotten.
>>
>>109162638
its a gpu it just has networking like the nvidia ones except it uses a standard PCI-E ribbon instead of proprietery shit to connect to other cards

also 400 gigabit fiber optic is probably good if you wanna link up a bunch of them together in a data center or homelab
>>
File: 1778244225122.jpg (116 KB, 1024x683)
116 KB JPG
>>109161221
>the jeet says in the video that it's not for gaming
>put "gaming gpu" in the video title anyway to trick people into clicking on it
How are they not getting more shit for this? GN is trying to sell it as an alternative to nvidia gaming gpus even though the guy very clearly states that's not what the company is doing. They're just blatantly lying to people now for views why are they not getting shit for this?
>>
>>109165058
I'm just desperately trying to figure out what the fuck the point is.
>dedicated BMC on an accelerator card?
>400GbE single port nic has less than half the bandwidth that NVLink does one way
>SODIMM????
GN is pure fucking slop for nitwits and always has been but who the fuck exactly is that sca- product trying to target?
>>
what’s the ai performance?
>>
>>109161221
idk man.
>ddr5
is that good enough? genuine question. idk

is the goal to be a gaming card & not ai?
>>
>>109161221
With what software????????
>>
>>109165140
DDR5 Sodims are about 6~ Gbps. GDDR7 is 32Gbps.
So, no.
>>
It would need like 6 memory channels to get 300GB/s but if they did that or more for top end it could work
>>
>>109162044
>>109166097
Might be better than sysram for running moe models? Idk
>>
>>109165107
>>109165109
>>109165140
>>109165776
>>109166257
IT'S AN OBVIOUS SCAM WHY ARE YOU PEOPLE PRETENDING THAT IT COULD POSSIBLY BE LEGIT?
HOLY FUCK YOU GUYS ARE DUMB
>>
>>109161488
>tech channel reports on tech
... yes, it's the fall of civilization...
>>
>>109161221
Steve is no better than Linus if he's willing to make a video glazing an actual scam like this.
>>
>>109161240
This video made me realize I don't really hate Indians, I mainly hate their accent
>>
>>109161221
>saaar my 4090 burned down my house saar believe me sar
Fucking lying sack of shit jeet.
Now don't get me wrong, that connector is a piece of shit, but there is around a 0% chance that jeet isn't just lying.
>>
File: risitas.jpg (29 KB, 447x376)
29 KB JPG
>nvidiots quaking at the mere thought that their precious company could have competition
>>
>>109167467
People like (You) are exactly why Steve decided to platform these scammers. You won't think for a second or ask any questions, if it satisfies your urge to "show Nvidia that gamers matter" or whatever, you'll blindly believe anything. Cattle.

Now go ahead and call me a paid Nvidia shill because I dared to criticize your "Tech Jesus" e-daddy, I know you want to.
>>
>>109161221
>jeet gpu

Ofc it's a scam. He will take the investment money and disappear
>>
>>109161221
so, a xylinx fpga with ram?
>>
>>109167420
>I don't really hate Indians, I mainly hate their accent
You are either a pájeet yourself, or you are someone that never had the disgust of interacting directly with one. They are truly the goblins of our world.
>>
>>109168423
nta but I would absolutely take any chance I can to fuck over nvidia for going full jew
>>
>>109170295
>I would absolutely take any chance
Yeah that's exactly the problem cause scammers know that.
>>
>>109161221
I like the DIMM slots Nvidia should have started doing that years ago
>>
>>109162262
>they are just far better at running businesses than whites
That why they made literally zero contribution to the history of the evolution of digital computing, operating systems, propgramming langiuages, the silicon chip, telephony, digital networking amalogue computing, video games, the creation of the internet and its protocols or the establishment of ANY of the industry leaders today? But they made an enormous contribution to the rise of fraud, scamming and are notorious liars and conmen that strangely ';suceed' when they enter high trust societies??
>>
>>109168683
yeah we don't get many Indians arojdn where I live. But the guy form the vid seemed reasonable enough idk
>>
>>109170295
>any chance i can to fuck over nvidia going full jew
except that this one isn't
>>
The DDR5 SODIMM's are the litmus test.
If you think they're a cool idea because you hate that Nvidia locks away the high capacity RAM to the most expensive models and you get tunnel vision from it, you failed it.
If you think it's completely retarded as DIMM's cannot reach the same speed that you'd want for an AI GPU, nor is DDR5 even close to GDDR6 or GDDR7 in terms of performance which is what you actually want as otherwise you could just offload the models to system RAM, you've successfully sniffed out why Bolt Graphics has been a scam ever since they first popped up.

Of course, Steve Burke of Gamers Nexus is, in reality, a complete idiot that doesn't understand anything out of his narrow little world of game benchmarking, and he has an insane rage boner for Jensen Huang, so he instantly sucked off this scam as it aligns with his grudge against Nvidia.
>>
did anybody actually watch the video?

the prototype die (FPGA) they have is surrounded by soldered on LPDDR5, the SODIMMS are for extra memory, they're targeting professional market for video/cad workflows.
>>
>>109171123
When they popped up around 2020 they were aiming for the AI market, now the smarmy jeet is moving the goal post since AI is no longer cool to market towards.

But hey, anything to show Jensen that he's nothing! Even if it's making a fool out of yourself by propping up vaporware that wouldn't even compete with AMD if it were to be made!
>>
>>109171123
>video/cad workflows
not even a retarded company would buy from a scammy jeetslop company with questionable at best sales pitch that ticks all the red boxes
even chinks with real and proper engineering teams struggles to run a game at a moderate speed let alone professional creative workloads
>>
>>109161499
they want you to be able to plug it into two computers, effectively halving the price
>>
>>109166303
Same for the comments on the video. I got brain damage from reading them, but at least the more recent ones are more critical.
>>
>>109171086
16 channels give you 800GB/s and 128GB for the price of a 5090 which has twice the bandwidth but only 32GB. It's not obvious to me that the idea is retarded.
>>
File: 1602257306272.jpg (85 KB, 633x738)
85 KB JPG
>>109161749
>>
>>109172689
I mean for some workloads, especially AI nerds, its cool. Brings generative video to the masses albeit with slow progress. But doable vs not possible like on most current consumer desktops.

HOWEVER

For gaming, and current speed standards, Its not quite enough. Maybe if the GPU die itself has a mega cache like AMD RDNA does then it can be enough, but we're talking about a company with no experience here. In the GPU market its extremely unlikely for a nobody to come out and match the state of the art with their first commercial design, and there is no matching the polish of the drivers and tuning without sheer manpower (see: money)

ALSO

Ram cost still isn't free and for a product like this you still need everything populated. All that would actually happen is the markup from the GPU companies gets passed on to middlemen who will actually add more cost, not less. As we have already seen by the ram crisis of 2025/2026.

The true litmus test of hype is the entire system itself IMO. RJ45 and QSFP both on an accelerator board reeks of a solution looking for a problem.
>>
>>109172689
The problem is inter-RAM latency is going to be catastrophic compared to a real GPU so training is fully out of the question and inference is probably still going to suck dick with you likely achieving better results in terms of inference speed with an EPYC with half a TB of LPDDR4/5
>>
>>109175156
I meant LRDIMMs not LPDDR brainfart.
>>
>>109161221
This shit isn't out yet? They were already shilling it here before AMD 9000 series even came out. It's probably fake and will never even exist.
>>
>109161221
>LPDDR5
>dual channel SODIMM
>ethernet ports
this shit is rtx 2060 performance at best
>>
File: smug plebbitor faggot.png (284 KB, 450x450)
284 KB PNG
>>109176576
B-but it'll PWN Nvidia! Show Jensen who's boss! What are you, a paid corpo shill? You get paid for talking shit about potential Nvidia killers?
>>
>>109175156
But EPYC would have the same problems and it's limited to 12 memory channls as far as I can see. I mean at least EPYC actually exists so yeah you're right but in theory I don't think it's impossible, just difficult perhaps.
>>
Delet
>>
anyone starting a GPU project the last few years isn't after the gaymer market
and all those data center features like bmc and 400gbps networking should reveal what they are intending to target
>>
>>109176576
>this shit is rtx 2060 performance at best
10 rtx 5090 raytracing theoretical performance and probably under a B580 raster performance.
>>
>>109161221
I hate that he's flanderizing his entire channel by covering this out of spite for the industry, and I hate that he isn't being honest about it. Fucking hell, Steve.
>>
>>109179179
What I hate the most is that while to a normal person like (You) or me it's a really fucking bad look for a guy who used to be revered for being a trustworthy news source and an authority, the vast majority of his current viewers are yes-men that only fuel him further into delusion that he is not doing anything wrong, with any critique being shot down as "paid corpo shills".

He has become what he criticized Linus of and he is to full of himself to realize that.
>>
>>109161780
I hypocrite streams and shows his face. There is nothing wrong with any of those.
>>
>>109162297
>why the fuck does it have 2 pci express ports? unironically who the fuck asked for that?

I did. Problem?



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.