News Nvidia Be Damned, RTX 4070 is Getting Illicit Blower-Style Cooler Treatment

IMHO if I was an AIB i'd take this to court.
They (nvidia) already started to shaft AIB by selling their own GPU at the MSRP forcing AIB to have to raise prices just to make profit.

They don't let AIB do a lot of the stuff in old days to make their stuff unique (again limiting profit they can make off em)

Telling em what cooler design they can or cant use? thats just stupid.

No company should have that much control or say about what a company does with product they purchase & build around.
 

ottonis

Reputable
Jun 10, 2020
166
133
4,760
Nvidia cannot "ban" any third party GPU cards. There's just no legal basis for that.
They could only threaten the vendors not to sell them any GPUs anymore if they dare making those blower style cards.
In that case, the top 5 vendors would just need to ignore nVidia's threats and go on making such cards - nVidia would singlehandedly destroy its own business should they still "ban" those companies.
 
  • Like
Reactions: artk2219
Illicit is a bit of an exaggeration. It may be against Nvidia's supply license, which could see Leadtek's chip supply reduced, axed or have penalties attached if Leadtek didn't ask for permission first and Nvidia decides to enforce it.

Yeah that kind of makes sense, but Nvidia should not be in a dominating position, just because they are scared that their own sales/profits are going to be affected if Leadtek or any other AIB starts selling blower-style GPUs at a lesser amount.

I mean we know Nvidia wants to protect it's own HPC and professional Quadro/A-series graphics cards, it makes sense, but they shouldn't have FULL control over what other other vendors/AIBs do with their GPUs, in my opinion.

That is, unless Nvidia has issued a serious warning before, that NO blower-type GPUs are allowed to be made by AIBs, without their permission , regardless of the GPU tier/class, 90/80/70/60. And not just xx90 class blower-style GPUs.

I think in this case, Leadtek might have already taken prior permission from Nvidia, imo. Else, they won't do such a bold move just to know that their products are going to get banned later on, when Nvidia finds it in the market.
 
  • Like
Reactions: artk2219

abufrejoval

Reputable
Jun 19, 2020
333
231
5,060
Yeah that kind of makes sense, but Nvidia should not be in a dominating position, just because they are scared that their own sales/profits are going to be affected if Leadtek or any other AIB starts selling blower-style GPUs at a lesser amount.

I mean we know Nvidia wants to protect it's own HPC and professional Quadro/A-series graphics cards, it makes sense, but they shouldn't have FULL control over what other other vendors/AIBs do with their GPUs, in my opinion.

That is, unless Nvidia has issued a serious warning before, that NO blower-type GPUs are allowed to be made by AIBs, without their permission , regardless of the GPU tier/class, 90/80/70/60. And not just xx90 class blower-style GPUs.

I think in this case, Leadtek might have already taken prior permission from Nvidia, imo. Else, they won't do such a bold move just to know that their products are going to get banned later on, when Nvidia finds it in the market.
Anti trust regulation fails to work in these near monopoly situations. Sure Nvidia shouldn't have that type of control, but they do and just like Intel/Apple/Microsoft etc. did and do, they understand the risk of not exercising it even beyond the strictly legal, too.

Actually, I'm quite surprised they lifted the restrictions on using CUDA in pass-through VMs, which was originally aimed against VDI use. I guess their motivation there was a bit like with Microsoft, where a too tough restriction on illegal (home-use) office only risked LibreOffice et. al. gaining to much market share and bleeding into corporate use.

So they want to ensure that CUDA developers have easy access to ML hardware, but that ML operations be done only after paying the "green" tax.

Perhaps after-market blower kits like the ones for water cooling would help bust their practise or a serious AMD offer with working pytorch support: Had AMD put their faith into the courts alone during the x86 patent disputes, they would have never survived even if Intel was finally proven guilty.

Personally, I certainly dislike loosing all these slots to cards far too wide, but I also dislike water in my computers.

A ribbon cable and upright mount may be advisable for physical stability in a tower chassis anyway, but the idea of signal integrity issues as a result isn't appealing.

I guess there just isn't any good solution as we climb beyond 500 Watts on personal computers, especially in summer and without AC.
 

crostini

Commendable
Sep 26, 2020
6
5
1,515
"If it comes out, it will be a very attractive professional SKU with 16GB of memory and (probably) a 200W TGP."

WAIT, WHAT? We have planning for 16GB 4070? I've heard the specifics on how this would be possible for a 4070, but I haven't heard if, or when this will happen. I am ready to get a new GPU so bad, and I'm still waiting due to the 4070 price and 12GB.
 
  • Like
Reactions: bit_user

dipique

Distinguished
Nov 25, 2013
15
19
18,515
"If it comes out, it will be a very attractive professional SKU with 16GB of memory and (probably) a 200W TGP."

WAIT, WHAT? We have planning for 16GB 4070? I've heard the specifics on how this would be possible for a 4070, but I haven't heard if, or when this will happen. I am ready to get a new GPU so bad, and I'm still waiting due to the 4070 price and 12GB.
Given that the 4060 is coming out with a 16gb version it'd be shocking if the 4070 didn't.
 

Eximo

Titan
Ambassador
One of two things, it is a very crippled 4080 die and has 16GB of memory (doubt it)or it is a very crippled 4070 die with a third of its memory channel missing, with double the ram density.

The latter seems more likely because the TDP is more in line with a 4070. So a 4070 with a 128 bit bus.

4060Ti is coming with a 16GB version, not the 4060.
 

abufrejoval

Reputable
Jun 19, 2020
333
231
5,060
There is a simple way to get around the "blower ban": slap on a straight-finned HSF like those found in datacenter GPUs and have the customer slap his own fans on the end (or setup stupidly high positive static pressure) to shove air through.
That's essentially what SuperMicro does when they sell you data-center GPUs in a workstation case: backmount fans sucking air through the shrouds.

Now we just need someone to sell those including the shrouds at 1/10 SuperMicro prices on major e-tailers.

P.S. well, this worked with V100s in dual slots. With 450 Watts and a dual slot form-factor, that design may need some not so trivial rework, say a Titan shroud and jet fans?
 

Eximo

Titan
Ambassador
I'm actually surprised no one makes desktop/tower chassis designed for datacenter cards. I wouldn't mind an all heatsink card, a shroud and a 120mm/140mm fan pushing air all the way from the front to the back of the case.
 

InvalidError

Titan
Moderator
I'm actually surprised no one makes desktop/tower chassis designed for datacenter cards. I wouldn't mind an all heatsink card, a shroud and a 120mm/140mm fan pushing air all the way from the front to the back of the case.
You'd probably want a standardized form factor specifically for that where all of the heatsinks and internal components are neatly aligned and packed like it is on servers where almost everything is custom-made for it.

For typical PC cases where most of the space is wasted, you'd need filler blocks to send air where it is actually needed instead of wasting airflow over empty PCIe slots, empty drive cages and empty space around other components like some servers have blanks to block off unused areas and not waste airflow there.
 

Eximo

Titan
Ambassador
You'd probably want a standardized form factor specifically for that where all of the heatsinks and internal components are neatly aligned and packed like it is on servers where almost everything is custom-made for it.

For typical PC cases where most of the space is wasted, you'd need filler blocks to send air where it is actually needed instead of wasting airflow over empty PCIe slots, empty drive cages and empty space around other components like some servers have blanks to block off unused areas and not waste airflow there.

That is what I mean. The datacenter GPU form factor has been very consistent, just need a chassis with the proper airflow package. Little silicone bushings for a tight seal and little flaps or doors to cover up slots not in use.

Some of the designs you get in OEM systems are actually really good for internal airflow. But they are generally not that nice to look at on the outside. (Though I am partial to Dell's workstation chassis)