NVIDIA CEO Jensen Huang hints at ‘exciting’ next-generation GPU update on September 20th Tuesday

Flogger23m

[H]F Junkie
Joined
Jun 19, 2009
Messages
13,114
Hoping they use this as an opportunity to restructure both of the 4080's.
The 12 GB needs to be renamed to 4070 and both cards need a price cut of at least $200. Using the 4090 as a baseline.

The 16 GB model would probably do fine at $1200 but is there really a market for the 12 GB at $900? Nvidia must have seen the writing on the wall.

$700 for a **70 is still crazy. Needs to be around $500.

It is also 24-30% slower than the 16GB.

Typically the **70 is closer to around 20-24% slower than the **80 if I am not mistaken.

relative-performance_2560-1440.png


relative-performance_2560-1440.png


So if we're looking at 24-30% slower, this is a weak entry for a **70 as it is. Even if they rename this to a "4070", it will still be fairly lackluster, and it would absolutely need to be $500 or less. The 4080 12GB is more like a "4060ti" or "4060 Super".
 
Last edited:

Armenius

Extremely [H]
Joined
Jan 28, 2014
Messages
35,443

kirbyrj

Fully [H]
Joined
Feb 1, 2005
Messages
30,241
$700 for a **70 is still crazy. Needs to be around $500.

It is also 24-30% slower than the 16GB.

Typically the **70 is closer to around 20-24% slower than the **80 if I am not mistaken.

View attachment 518459

View attachment 518460

So if we're looking at 24-30% slower, this is a weak entry for a **70 as it is. Even if they rename this to a "4070", it will still be fairly lackluster, and it would absolutely need to be $500 or less. The 4080 12GB is more like a "4060ti" or "4060 Super".

And that bears itself out in the fact that the XX60 series is often 192 bit.

No one is going to buy a $900 4060.
 

TaintedSquirrel

[H]F Junkie
Joined
Aug 5, 2013
Messages
11,756
$700 for a **70 is still crazy. Needs to be around $500.

It is also 24-30% slower than the 16GB.

Typically the **70 is closer to around 20-24% slower than the **80 if I am not mistaken.
Nvidia has inadvertently given us a look into their margins with the 4090.
The 4090 is 64% faster than the 3090 while costing 6.6% more money. If we apply this to the rest of the 40 series we can see how much they should cost.

Assuming the 4080 12 GB is 3090 performance: It's 69.4% faster than the 3060 Ti. $400 plus 6.6% is $426. It's slightly faster so add a little bit, and the 4080 12 GB should be slightly ahead of the 3090, so we end up with ~$500.
For the 4080 16 GB, 20% over the 3090: It's 56% faster than the 3080 10 GB. $700 plus 6.6% is $746. It's slightly slower so subtract a little bit, ~$700.
(I used TPU's latest performance summary)

So... We end up with the exact same MSRPs as the 3070 and 3080.

You could probably calculate the exact perf-per-dollar of the 4090 and get more precise numbers.
 
Last edited:

Lakados

Supreme [H]ardness
Joined
Feb 3, 2014
Messages
6,958
It's not needed anymore and limiters of that nature will always have performance impacts elsewhere in the system and if they keep them there then they need to maintain them. GPU mining is essentially dead so why leave a potentially problematic timebomb in your drivers if you don't have to?
 

Lakados

Supreme [H]ardness
Joined
Feb 3, 2014
Messages
6,958
Nvidia has inadvertently given us a look into their margins with the 4090.
The 4090 is 64% faster than the 3090 while costing 6.6% more money. If we apply this to the rest of the 40 series we can see how much they should cost.

Assuming the 4080 12 GB is 3090 performance: It's 69.4% faster than the 3060 Ti. $400 plus 6.6% is $426. It's slightly faster so add a little bit, about $450.
For the 4080 16 GB, 20% over the 3090: It's 56% faster than the 3080 10 GB. $700 plus 6.6% is $746. It's slightly slower so subtract a little bit, $650-$700.
(I used TPU's latest performance summary)

You could probably calculate the exact perf-per-dollar of the 4090 and get more precise numbers.
Nvidia also needs to try and recoup that $10B they paid to get back on TSMC's production schedule and off Samsung. You better bet if they don't work those costs into the hardware while they can Investors will shit a brick.
But yes Nvidia is overcharging for the current lineup just as they said they would, once AMD has their parts available and the 3000 series overstock is gone they can reconsider their pricing. But until there is a reason too why would they, I mean people are lining up around the block at a chance at the existing 4000 series parts so why lower the price if they are flying off the shelves?
 
Last edited:

TaintedSquirrel

[H]F Junkie
Joined
Aug 5, 2013
Messages
11,756
CableMod has posted their 90-degree adapter. It looks like it may not be compatible with cards that use recessed power connectors, the PCB will block it. Also I'm pretty sure directly connecting a sleeved 12-pin to the card would still take up less space.

CableMod_90Degree.png
 

GoodBoy

2[H]4U
Joined
Nov 29, 2004
Messages
2,417
From CableMod regarding the 12vhpwr cable that has the straight connector:

"12VHPWR cable FAQ:

https://cablemod.com/support/#12vhpwr-pcie-gen-5-atx-3-0

One thing of note here is that due to the cable bend, you’re going to need roughly 4 to 4.5cm of space from your graphics card to your side panel in order to properly install this cable. Please check this measurement with your PC’s case."

So, that connector looks to decrease the required distance from the top of the card to the side of the case from 4.5Cm to 2.3Cm. I only have 3.2Cm clearance myself, so that appears to be a good solution.
 

RazorWind

Supreme [H]ardness
Joined
Feb 11, 2001
Messages
4,450
Card manufacturer will blame the PSU or cite board damage to deny RMA. Bet on it.
Is that your card in the photo?

I could probably repair the board, if the manufacturer won't, assuming we can get our hands on a replacement connector.
 

ChadD

Supreme [H]ardness
Joined
Feb 8, 2016
Messages
5,903
Sounds like the partner boards don't need just anti sag sticks... they need the Extreme Cable Mohawk Cap. A hunk of pink plastic to keep the cable coming out of the card standing up straight and only bending beyond 35mm from the connector.

Only half joking... throw some RGB on it. The POWER petal with RGB header. The Cable Mate cable erector.
How about one that was more a tube... the Tower of Power Cable Sleeve.
Or just put a battery in one for the RBG so you don't need even more power cables (sounds safe) call it the BOB. The battery operated (cable) BOOM.

In all seriousness though... Nvidia should probably recall these, and ship customers a free a new cable that has a RIGIDE covering extending to 35mm past the connecter so its impossible to bend or connect horizontally. Perhaps even print the basic... don't install in a horizontal configuration on the actual cable. It might be better PR then, I am going to bet good handful of such pictures getting post over the next few months. A lot of these installation issues will be fine for a little while... My bet is Third week of November. That is when the flood of WTF my card is on fire posts start becoming common.
 

Lakados

Supreme [H]ardness
Joined
Feb 3, 2014
Messages
6,958
Yep, don't follow directions for use and suffer the consequences.
That said if a wire fails that catastrophically by being bent in a confined space it is a poorly designed wire. I know 90 degree aftermarket connectors are a thing and I suspect they are the vastly better cable.
 

RazorWind

Supreme [H]ardness
Joined
Feb 11, 2001
Messages
4,450
That said if a wire fails that catastrophically by being bent in a confined space it is a poorly designed wire. I know 90 degree aftermarket connectors are a thing and I suspect they are the vastly better cable.
The problem is not the cable, but rather the pins in the connector. The smaller pins get loose more easily than the bigger ones in the old school 8 pin connectors, creating a loose connection that causes the pins to get hot when large current is passed through them, eventually melting the connector, or causing the plastic to catch fire.
 

Lakados

Supreme [H]ardness
Joined
Feb 3, 2014
Messages
6,958
The problem is not the cable, but rather the pins in the connector. The smaller pins get loose more easily than the bigger ones in the old school 8 pin connectors, creating a loose connection that causes the pins to get hot when large current is passed through them, eventually melting the connector, or causing the plastic to catch fire.
I was under the impression too much torsional force on them caused them to dislodge and short. Regardless the adapters design is problematic. Here’s hoping that ATX 3 PSU’s use higher quality whatever’s and manage to correct this. I struggle to understand how these passed standards Q&A it seems like an obvious test case.
 

TheHig

[H]ard|Gawd
Joined
Apr 9, 2016
Messages
1,120
1666649654753.png

From the Buildzoid video.. Here is the shot of the person's install that ended up cooking the connector shown in the pic. I could see plenty of people choosing to install a chonker like the 4090 this way.

Here is the shot he has in the vid for the guidelines of said connecter
1666649818412.png


In any case the connector seems finnicky AF and putting 1600-2 grand on the line with this solution just seems a bit shitty if you ask me. I'm looking for a rev 2.0 here soon maybe?
 

LigTasm

Supreme [H]ardness
Joined
Jul 29, 2011
Messages
5,644
Looking back at the thread title, it seems the exciting part was going to be the rush to check geforce experience 17 times a day to see if you were the "chosen one" for the stock this week.
 

DPI

[H]F Junkie
Joined
Apr 20, 2013
Messages
12,594
Worse: Reddit.
Nothing the internet and reaction-youtubers and rumor tech blog o-sphere loves more than a circlejerk over on an out of context single photo posted on Reddit or twitter. Want 10,000+ upvotes, a shower of karma and instant online sensation? Post a photo of a wrench in your empty 4090 box and send it up- "Got EMPTY 4090 box from Amazon.. Bezos?!"

That's not to say the 12VHPWR doesn't need some scrutiny, but these n=1 anecdotes should probably be taken with a grain of salt -- we don't know what any poster of a photo and a claim did to get to that endpoint. Was he screwing with the connector or cable while the GPU was powered on, like OCD making sure it's in tight or trying to fix his cable management? If so he'd never admit to that in seeking sympathy or an RMA. These claimed cases may well be legit and have happened spontaneously, but there are tens of thousands of these GPU's installed and running yet the internet is in full viral viral contagion mode over 1 or 2 outliers.
 
Last edited:

XenIneX

Gawd
Joined
May 19, 2012
Messages
883
Nothing the internet and reaction-youtubers and rumor tech blog o-sphere loves more than a circlejerk over on an out of context single photo posted on Reddit or twitter. Want 10,000+ upvotes, a shower of karma and instant online sensation? Post a photo of a wrench in your empty 4090 box and send it up- "Got EMPTY 4090 box from Amazon.. Bezos?!"

That's not to say the 12VHPWR doesn't need some scrutiny, but these n=1 anecdotes should probably be taken with a grain of salt -- we don't know what any poster of a photo and a claim did to get to that endpoint. Was he screwing with the connector or cable while the GPU was powered on, like OCD making sure it's in tight or trying to fix his cable management? If so he'd never admit to that in seeking sympathy or an RMA. These claimed cases may well be legit and have happened spontaneously, but there are tens of thousands of these GPU's installed and running yet the internet is in full viral viral contagion mode over 1 or 2 outliers.
Yes. On the one hand, if you make an idiot-proof connector, the world will make a better idiot. On the other hand... they did not make an idiot-proof connector.

Bend radius specifications are for licensed or certified professionals. If your shit's gonna be handled by plebs, you put an appropriate strain relief on that thing.
 
Top