Call me a noob, but what makes the Quadro cards better at CAD than the much cheaper Geforce solution? Is it a hardware difference, software difference, Or both? I have owned 4 different Quadro cards and I never knew the answer to this
Emphasis on a different kind of 2D/3D calculation.
When you buy a workstation card you get workstation drivers, when you buy a gaming card you get game drivers. The card is basically the same.
>>54629446
I did some research, what you say is true. Can you use geforce drivers on the relative quadro?
>>54629458
I've seen it done on older cards, newer drivers most likely have a software level lockout which makes it unlikely.
Big time high end payroll cgi artist here, not completely sure about heavy cad files but absolutly no one uses quadro or anything that isn't geforce for 3D.
>>54629408
That's bullshit. The chip is the same, and most of the calculations, which are a bunch of matrix multiplication can be turned into a series of fused multiply add instructions.
>>54630143
I have a Quadro 2000 and a Quadro Fx 4800.....why do both of them suck at gaymes?
>>54629385
>Quadro cards better at CAD than the much cheaper Geforce solution?
Hi Anon, I used to manage desktop deployment for a CAD department. We used autodesk's suite along with solidworks, catia, masterCAM, and a bunch of niche architecture software. We used both Quadros (desktop and mobile) and the AMD FirePro line.
Most of the contacts I've known in CAD and GPU hardware sales tell me this:
- "Professional"/"Workstation" Class GPU platform uses and supports buffered/ECC memory
- Comes with guaranteed first-class support - (this is what you're said to "pay for" in service subscriptions - such as faster access to bug reporting and resolution tools, quicker upgrades to new OPENGL/etc versions, ...)
- Typically, supports stronger coupling with CAD software (for example, SolidWorks will not run with full hardware acceleration with a "gaming" card, only a FirePro/Quadro). This is because the CAD software developers are also testing with the Pro hardware, almost exclusively. Thus some edge case performance tweaks and API/features which might not be present or workable on gaming-tier cards 100% of the time, might sneak through.
Since there is a much vider variety of (and more importantly, variety in the quality of) gaming-class cards across all OEMs, vs. the (generally first-party) pro cards, it is much easier to maintain reactive support for the latter.
>>54629385
jewish voodoo
>>54630143
You don't get it.
Matrix multiplication or not, you have to dedicate a certain amount of die space on the ALU to FP32 operations or FP64 operations.
With stream processors, these ALUs are multiplied by a shitload.
So the area taken up by either FP32 or FP64 operations in the ALU take up much more area.
Now, if you were a video game developer, you would realize that FP32 calculations are the only thing you need. Nothing needs to be that precise. Nvidia listens to its customers. To speed things up for their game devs they reduce the area dedicated to FP64 calculations. FP64 becomes slower, more die space is dedicated to FP32.
Same happens for FP64 for workstation applications.
Do you understand now?
As this guy said
>>54630240
workstation cards have ECC memory
ECC memory has an extra parity chip used for error detection and correction, so that long running calculations (over a month long) don't get corrupted by background radiation which has a tendency to flip bits over a long period of time. This error correction has overhead which reduces the memory access speed, which is critical to realtime applications like games.
So there is definitely a hardware aspect to it.
There may also be a driver optimization aspect, which is very real because for the longest time, only Workstation cards worked on Autodesk Maya and 3ds Max. Nowadays Solidworks still ``requires" a workstation card, but Maya and 3ds Max released a viewport which operates on the same style of engine that video games do, so Nvidia and AMD can't jew you in the driver by requiring you buy a workstation card for workstation applications.
>>54629385
hey I have the same netbook as that kid, too bad mine doesn't work
a fuck load of textures
>>54630240
I recall one article (kill me I don't remember which one) where a GeForce was hacked to present itself as an equivalently-spec'ed Quadro to the computer, they tried to enable full hardware acceleration in SolidWorks and it started glitching while offering less viewport performance, which would indicate that there's low-level differences between the cards besides just the drivers.
It should be noted that not all Quadros natively support ECC, Maxwell isn't capable of supporting it, so the functionality is achieved through over-provisioning the VRAM, since cards like the M4000 and M6000 have 8/12GB, it's not a problem to have less effective memory.
In this respect, I find it strange that people went crazy over the 3.5GB fiasco of the 970, when all the Maxwell Quadros lose some 10~20% of their RAM to provisioning.
Either way, with each generation, Quadros are losing distinguishing features one-by-one as they become mainstream. FP64 got gimped for more FP32, VR and multi-display is now consumer-level, 10-bit and even 12-bit color are now on the consumer Pascal cards, and RAM is now in the range of 8GB+ for new cards, which is enough for heavy-duty work. ECC is the only thing that's not available, and it's use outside of anything that's not scientific calculations being performed for days on end with hundreds of cards is pointless. It really does come down to support and program compatibility at this point, but even most pros I reckon can easily get away with just using regular GeForce cards, especially when you can fill out a computer with four of them for the price of one top-end Quadro.
>>54629385
Driver stability and optimizations, stable GPU and memory operating frequencies, 3D output long before 3D Vision, etc.
as far as i know, it is about precision, licensing, and bullshit
they charge you more because they can, and know business will spend mad dosh regardless
>>54629385
Having a Quadro card is mainly just to unlock extra features
Nvidia/AMD pays companies to disable these features on consumer cards
>TFW GTX can't 30bit output
