Question What is a similar spec'd alternative to the Quadro P4000?

Aeacus

Titan
Ambassador
Performance wise, Quadro M6000, (but older than P4000)
comparison: https://gpu.userbenchmark.com/Compare/Nvidia-Quadro-P4000-vs-Nvidia-Quadro-M6000/m250990vsm36278

Price wise, Quadro M5000, (far weaker than P4000)
pcpp: https://pcpartpicker.com/products/compare/dDqbt6,7ms8TW/

But if you want to newer GPU, which isn't EOL, then next in line from Pascal architecture would be Turing (RTX series, e.g Quadro RTX 4000), while Ada Lovelace is the newest (RTX A series, e.g Quadro RTX A4000). But those it will cost accordingly,
Turing, pcpp: https://pcpartpicker.com/products/compare/sLNv6h,sChKHx,k866Mp,8kF48d/
Ada Lovelace, nvidia: https://www.nvidia.com/en-us/design-visualization/desktop-graphics/

Something similar and a little better would be a Gtx 1660 super.
Gaming GPU is not designed for CAD work, as Quadro GPUs are. So, why suggest completely different use case GPU?

Gaming GPUs are only good for 1 thing - gaming. For everything else, except gaming - Quadro does better.
 

IDProG

Distinguished
Gaming GPU is not designed for CAD work, as Quadro GPUs are. So, why suggest completely different use case GPU?

Gaming GPUs are only good for 1 thing - gaming. For everything else, except gaming - Quadro does better.
This is absolutely false. There is a reason why Nvidia keeps skimping on VRAM on GeForce cards. There is a reason why Nvidia really, really hated having to launch the 3060 12GB version.

Because they don't want the cheap GeForce cards to cannibalize the expensive Quadro cards.

They have always tried to maximize profit margin by ripping off the generally-richer professional 3D artist.

"Look, guys, this card has twice the VRAM (that only costs $4/GB to produce) for twice the price".
The Quadro cards do not even use HBM2 memory. That's how you know it's absolute bollocks.
 
  • Like
Reactions: bit_user

Aeacus

Titan
Ambassador
There is a reason why Nvidia really, really hated having to launch the 3060 12GB version.
Unless you are Nvidia, you can not say what Nvidia felt or didn't, when releasing RTX 3060 12GB GPU.

No-one forces Nvidia to do something they do not want. Not the AIBs and not especially the consumers. So, if Nvidia wanted to produce and release RTX 3060 12GB, it's due to their own accord.

As far as GeForce vs Quadro goes, both GPUs use completely different drivers. Also, what goes into the GPU itself is completely different. Only vague similarity is the chip architecture.

Because they don't want the cheap GeForce cards to cannibalize the expensive Quadro cards.

Completely different use cases for both GPUs. Sure, you may CAD on GeForce GPU but compared to purpose built Quadro, render times on GeForce are huge. Not to mention the lack of error correction (ECC VRAM), among other things.

It's like comparing super car with semi-truck, two completely different purpose machines, while trying to justify why use super car to haul stuff, compared using semi-truck for cargo haulage. These two can't be compared. Just like you can not compare GeForce and Quadro.

They have always tried to maximize profit margin by ripping off the generally-richer professional 3D artist.

Yet, we have consumer GPU with a price tag of $1599. :rolleyes: Heck, even RTX A4000 doesn't cost that much, ~$1000.
 

bit_user

Titan
Ambassador
I know the P4000 is now pretty much end of life, but what would a similarly spec'd (and cost) model for CAD work be today?
Which CAD software? Is there a user forum for it, where you can inquire what others are having success with?

I'm skeptical the additional price of a workstation-level card is justified, but I'm speaking from zero experience with modern CAD software. In terms of hardware differences, Nvidia's workstation cards mainly differ in the amount of VRAM (more), clock speeds (less), power requirements (less), cooling solution (slimmer), and warranty (longer). At their core, they use the exact same GPU chips as the gaming cards, as you can see from this page:

Maybe the OpenGL performance of their drivers is crippled on Windows, if you don't use the workstation boards, but they don't seem to differentiate on Linux.

Anyway, with so many unknowns, I think the best advice you can get is from someone using the same software for similar tasks.
 
Thanks for all the replies! End PC user will be using Archicad 25/26 (and by the end of the year, 27). Ideally, he'd also like to use Redshift for rendering, but the recommended specs for that, GPU-wise, are $$$. As he doesn't want to go the route of "gaming specific" Geforce cards, and he doesn't have an endless budget, I'm now considering the RTX A2000. Not a beast of a card by any means, but I'm hoping it'll do the trick. Pairing with i7-12700.

The A2000 looks like it's comparable (and superior) to the P4000 which was his "basic" GPU wish list option that he gave me - he didn't realise that the card is pretty much end of life (and can't be had at all over here in NZ anymore). His current machine, which I built him 7 years ago, is running a 7700K paired with a P1000, so I believe this is a solid upgrade.
 
  • Like
Reactions: bit_user
The older NVIDIA "Titan" devices have more VRAM than the average gaming card, but cost less than Quadro alternatives. I don't know how much VRAM your Quadro has, but for CAD, this is one of the selling points of the more expensive cards...the ability to hold a lot in memory. I know early on the Quadros had line antialiasing enabled in hardware, but the regular gaming cards did not...I do not know if that is still true or not, I think that difference may have gone away with newer cards. You might want to see this search and see if anything for the price you are interested in has sufficient VRAM:
Google: price nvidia "titan"
 
The older NVIDIA "Titan" devices have more VRAM than the average gaming card, but cost less than Quadro alternatives. I don't know how much VRAM your Quadro has, but for CAD, this is one of the selling points of the more expensive cards...the ability to hold a lot in memory. I know early on the Quadros had line antialiasing enabled in hardware, but the regular gaming cards did not...I do not know if that is still true or not, I think that difference may have gone away with newer cards. You might want to see this search and see if anything for the price you are interested in has sufficient VRAM:
Google: price nvidia "titan"

Thanks for the tip - sadly, over here in New Zealand, pretty much a no go. One or two of the monster Titans still knocking about, but way out of budget.

The A2000 I'm looking at only has 6GB of RAM. The person I'm building for was asking for a P4000 (8GB) as his basic option, and a comparison of the two in the likes of Passmark, suggests that due to it's newer age and architecture, even though the A2000 has less RAM, it still 'out does' the P4000. I don't want to under spec him for his needs, but nevertheless, he's a small town architect, so budget-wise the difference between a just sub-$1000 card and the next step up, which is +$2000 is quite a bit (obviously not taking in the cost of the rest of the system).
 

Aeacus

Titan
Ambassador
Thanks for the tip - sadly, over here in New Zealand, pretty much a no go. One or two of the monster Titans still knocking about, but way out of budget.

The A2000 I'm looking at only has 6GB of RAM. The person I'm building for was asking for a P4000 (8GB) as his basic option, and a comparison of the two in the likes of Passmark, suggests that due to it's newer age and architecture, even though the A2000 has less RAM, it still 'out does' the P4000. I don't want to under spec him for his needs, but nevertheless, he's a small town architect, so budget-wise the difference between a just sub-$1000 card and the next step up, which is +$2000 is quite a bit (obviously not taking in the cost of the rest of the system).
There are some newer gaming GPUs from NVIDIA which have more than 8 GB of VRAM, and probably outperform the spec'd GPU by quite a bit. The down side is that they might also cost significantly more, but it is worth looking. Anyone doing CAD will benefit from more VRAM (and more regular RAM and a faster NVMe). Perhaps the customer does not understand the importance of VRAM; you can't really go wrong by getting what he/she wants, but it is likely he/she will be happier with more VRAM. Perhaps a model with 12 GB would be available...that certainly beats 8 GB of VRAM.

Note that if regular CPU RAM is insufficient, then virtual (swap) memory can be used. It slows things down a lot, but still works. VRAM is a different beast though: Virtual (swap) memory cannot be used if VRAM is insufficient. VRAM requires actual physical RAM.
 
  • Like
Reactions: woodson75
There is the 12GB version of A2000 as well,
amazon: https://www.amazon.com/PNY-VCNRTXA200012GB-PB-NVIDIA-A2000-12GB/dp/B09SJ2BZ85

The 6GB version costs ~$200 less,
amazon: https://www.amazon.com/PNY-NVIDIA-A2000-Graphic-VCNRTXA2000-PB/dp/B09KMMQW6M

I suggest getting RTX A2000 12GB since it's better than P4000 and price wise, should be below $1000 as well, even in New Zealand.

Mostly out of stock over here it seems, but the price seems to be well over $1K. Have made enquiries into re-stocking all the same. Thanks for the tip.