Graphics card query

[kev030981]

Hi - I'm about to upgrade my system,, and would love some help with a few graphics card questions?

My Current system is i7-2600K 16GB with a gtx 570. I run a 27" display at 2560 x 1440. I use the system for a mix of both gaming and architectural visualisations (3DS max, vray, Revit, Adobe Creative suite)

From advice I had decided to go with an upgrade to gtx 780, which looks good for my needs. It now seems I am to get a bit more £ for a job I recently finished , so I'm considering upgrading (my upgrade) to a Titan. Rather than ask if its better than a 780 (I've read countless threads on this) -I was wondering if someone could help me understand, as simply as possible - what each of the specs means in terms of real life use?

What I mean is
* What will increase viewport speed on complex scenes (3DS max / Revit) - memory? Cuda cores? etc
* What will increase rendering speeds?
* What will allow me to run more complex (high poly) scenes - i suppose this is an extension of viewport speed? (I think?).
* What does this double precision point thing actually mean for the above apps (if anything)? It seems to be the main dif (other than ram) between the two cards.

I'm quite technical - but am never 100% sure what each of the specs mean when I'm comparing the cards. There doesn't seem to be a clear explanation from my searching on the web? If anyone has any thoughts (or a link) I'd be most grateful!

Then with all that said ^ - Does anyone think I'd be wasting the money on the titan? Or will it make a real difference, if so - i'd gladly spend the extra on it but don't want to waste the money!

Many thanks!

PS: I've looked in Quadro cards, but want to stick with GTX due to gaming. Ive also been (dreaming for at least a few months) about putting a quadro in and running a duel boot. (although this is later - much later). Any thoughts on this - superflous with titan? Which quadro would make a difference? K4000 i would assume?

[RaoulM]

Hello,

> * What will increase viewport speed on complex scenes (3DS max / Revit) - memory? Cuda cores? etc

In my meaning both would be great.

> * What will increase rendering speeds?

This is pending on what kind of render engine you are using!

Normal terms and conditions I would prefer to build first a stable workstation and then make this fast in the second

step! 

1. Intel Xeon E5 CPU + ECC/REG RAM

2. nVIDIA Maximus setup

> * What will allow me to run more complex (high poly) scenes - i suppose this is an extension of viewport speed?

(I think?).

Powerful CPU

much memory (RAM)

Powerful Graphic cards with also much RAM

> * What does this double precision point thing actually mean for the above apps (if anything)? It seems to be the main dif (other than ram) between the two cards.

- One graphic card is made for professional artists to create something!

- The other graphic cards are made to play inside the scenes or art work!

And in my pore opinion is this not so far the same thing!

> I've looked in Quadro cards, but want to stick with GTX due to gaming. 

Then build one PC for gaming or buy a gaming console and build one machine for do professional 3D work.

I mean really that all the autodesk software was made for highly professionals and  for real work!

For sure many peoples are madking their money with a normal job and this is only their hobby or second job, for sure.

But for all peoples you are working with this software and must earn their money with it, should´nt ask about

the graphic cards or what kind to take, because the nVIDIA Quadro option is abel to serve them also a second

funktione or option, you are able to add one or more Tesla card(s) and speed the entire system up.

Something at last please, the most peoples are only, I repead only are looking to the tech specs. from each

graphic card, but this is even only the half truth, because the drivers and/or software support is more imprtant then 

pure tech specs. but this is also even differ from each of us because of the differend usage of the autodesk software.

[braudpat]

Hello from France

1) The GTX 780 is a "small affordable" TItan, so it's the right choice (for me) ...

2) Don't hesitate to buy the fastest Xeon E3/E5 or Core I7, you can ...

The "standard" Core Speed is very important ! and must be at leat 3.0 Ghz or more ...

The Turbo-Boost Mode will adapt in real time the core speed depending on the demand of the Windows program : ACAD, REVIT, 3DS Max, etc ...

If you are going to run during many hours/days continuously, you must have a Xeon and ECC Memory (ECC means reliability and error correction)

For example, I have a HP Z1 Workstation with 32 Gb ECC (and a modest Quadro 1000 - 2 Gb), I never get any "strange" error possibly coming from memory error ...

The HP Z1 has a Xeon E3 (3.2-3.7 Ghz - 4 cores - 8 threads), it is not the best Xeon but it's quite enought for me !?

If you plan a lot of Rendering, you must consider a Xeon E5 with 6 cores - 12 threads ...

3) Don't hesitate to buy a lot of memory : 24/32 Gb is the right choice if you will use large 2D/3D Models ...

4) A fast SSD as primary HD and a fast Sata as secondary HD ...

5) Win Seven Pro and NOT (if possible) Win Eight ...

Bye, Pat

[kev030981]

Hello! Thanks both of you for your posts, very helpful advice.

May I summarise what I've gathered from your answers to see if I'm right?

 *Processor, memory and gfx card are all important - it is probably better to have a dedicated workstation, and a separate dedicated gaming pc, rather than try to mix the two (if money allows of course)

*Workstation components are better suited to reliability and prolonged use (professional hours)

* Rendering is handled by the processor - the std core speed is the determining factor (?) (I woulod assume the no. of threads makes a huge difference here too?)

*Large amounts of memory allow you to run larger and more complex scenes (?)

Can I just ask in addition to this - what specifically does the memory on a gfx card allow you to do then? Does the 6GB of the titan allow for some identifiable increase (rendering / viewport / no. of polys etc) - as I think this would determine whether I spedn the extra money. (I ask this question directly to any application - but if there is a bonus to 3DS max, revit, ACAD, or Photoshop - this is what I'm interested in) 

Also - interesting what you say about the SSD and Win 7 - that is what I will be running (SSD might be next upgrade though depedning on which card I chose) 

Thanks again to both - really helpful advice 🙂

Kev

[RaoulM]

Hello again,

>  *Processor, memory and gfx card are all important

The Whole system is working together like an engine and on top with the installed software too!

Its a must be to set up many things or assemble a workstation this is able to do the entire work.

But there are some main things you should be looking for.

- under powered ist not bringing the power you need and let you not reach the goal.

- right powered is like the name tells right sorted and should do the job for a certain time.

- over powered or over sorted is like step 2 (right powered) but for a longer time!

- All spare parts should wor fine together and don´t build a so callef bootleneck in side the system

- The hardware should be right oriented for the things or plain the work the machine is made for!

How more you will be spezialized the hardware to the used software it will you bring the most efford.

One example:

My workstation for video cutting and edeting is like the follow:

- ASUS X79 WS Board

- Intel Core i7 3970x

- 64 GB RAM DDR3-2133

- SSD Samsung840 Pro

- WD VelocyRaptor 1 TB

- Adaptec RAID controller 6805

- nVIDIA Quadro 6000 + SDI in/out boards

- Tesla C2075

This is only for speed and power, faster more powerful ist better!

My Workstation for 3D art work is like the following:

ASUS DUAL CPU WS Board

- 2 x Intel Xeon E5 2950

- 128 GB EEC/REG

- SSD Samsung Pro

- WD VelocyRaptor 1 TB

- Adaptec RAID Controller 6805

- nVIDIA Quadro 6000 > will be changed to Qudro k6000

- Tesla C2075 > will be changed to Tesla k20

This is only to have a stable basis, that is not even crashing and serve a good performed, smooth and liquid workflow.

> - it is probably better to have a dedicated workstation, and a separate dedicated gaming pc, rather than try to mix the two (if money allows of course)

For sure if you can do it or plain realize it, I would guessing you to do so.

> *Workstation components are better suited to reliability and prolonged use (professional hours)

it is also more orientated what kind of work you are doing and what is the entire goal of your work.

If I earn only 25k in one year I cant buy hardware for more money! But if this is my only job and I earn ~150.000k

each year and the competition or rivalry is so great it should bring me perhaps to beeing a little bit faster not better

but faster then the others and perhaps I can get some more work related to the fast hardware you are using.

> * Rendering is handled by the processor - the std core speed is the determining factor (?) (I woulod assume the no. of threads makes a huge difference here too?)

In normal terms and conditions GHz is pointing out Cores, this is it, short and easy to remember, but the entire 

system stability and then on top the by yours used render engine is also a point that should not forgetten.

> *Large amounts of memory allow you to run larger and more complex scenes (?)

More or many RAM is in the most cases only good for one thing, it is bringing you liquid and smooth workflow

and if you are working with many programs on your machine, it will be the best to have much of RAM for sure

it is not a "must be" but more a good to have. All code will be "called" and proceeded through the CPU in then 

it will be placed inside of the RAM and if then there is not enough space, the process or data will be swapped

over from the RAM into the pagefile on the HDD/SSD and this is even slower then the RAM, in isial cases.

> Can I just ask in addition to this - what specifically does the memory on a gfx card allow you to do then? Does the 6GB of the titan allow for some identifiable increase (rendering / viewport / no. of polys etc)

Not really, but it allows you to load greater scens inside of the RAM but this is also pointed to the render engine

you are using and on top, the entire work must be placed inside the RAM of the graphic card, otherwise it will be swapped over and will be placed inside the slower system RAM, so if you have more RAM on the graphic card

you are abelt to load greater scens.

> I ask this question directly to any application - but if there is a bonus to 3DS max, revit, ACAD, or Photoshop - this is what I'm interested in

Not even so easy to answer, but if you are yousing some well known ways you will be paticipate from much more 

and on the other side fast RAM.

- 3Ds max > Maya > Adobe Premiere > AfterEffects = big files that should be placed inside of the RAM

- 3Ds max > Adobe Photoshop > Plugins = big files that should be paced comlete inside of the RAM

For sure everybody will even know when he is running out of RAM and this dosen´t matter if the 

system RAM is out od spcae or the memory of the graphic card is running out of RAM

It even slows down the entire Process and worklflow.

> Also - interesting what you say about the SSD and Win 7 - that is what I will be running (SSD might be next upgrade though depedning on which card I chose) 

For sure you should be looking before you buy a SSD what kind of SATA Chips is soldered on your mainboard!

On the both examples above (Workstations) are the badest SATA Chips ever where soldered on the boards!

The y where from Marvell and brings you only SATA II speed, on a brand new workstation with a SATA III SSD from

Samsung (840 Pro Series) so I was very disapointed and very angry but with the Adaptec RAID controllers 

I was getting right sorted and served, and the speed now is impressive against the lame **** Marvell ones.

And another thing is verry importand I really think, please go with a larger modell of SSD please.

The wear leveling algorithm is take from each SSD a little bit space for caching propurses and activities, so

if you are owning a 256 GB or 512 GB SSD, it can be taken more space for this action as from a so far smaller

SSD and the best it is also faster then a SSD from the same Series, again from a 64 GB modell it will be perhaps taken 4 GB for cache, but from a 256 GB modell it will be taken perhaps 12 GB - 24 GB so you can see it will be 

faster then the smaller one!

At last the wear leveling algorithm is unsing evem the name tell, in an algorithm from time to time another free block

on the SSD and that means that the blocks are not used onlly onesided but more then this the free blocks will be written with data more allocated, and this is highing up the entire lifetime of the SSD also.

Cheers 

RaoulM

[dgorsman]

Don't get hung up on the term "workstation" - it doesn't mean what it once did, thanks to the rise of hardcore computer gamers conventional PCs can easily match workstation capabilities.  I've got a couple of long posts on this in here already.

[braudpat]

Hello

+1 with Dgorsman for a "good" PC gamer with a gamer graphic card !

I know Gamer PC with Core i5/i7 overclocked a 4.0-4.5 Ghz which are faster (with ACAD, Inventor, 3DS ax, etc) than any Xeon !

But the stability could be a problem !?

So if the PC is running many hours/days continuously for Rendering/Shading or any other "long" program

you must consider a Xeon Processor + ECC Ram !

Which is used by the Workstation : HP, Dell, etc ...

Bye, Pat

[dgorsman]

"Stability" is not an issue, even with modest rendering - check with the 3DS MAX and Maya folks, many of them are using non-Xeon processors with great success.  Those heavily into full-time rendering work are either using Xeon motherboards and processors for the multi-processor (dual/quad processors @ quad/hex cores each) and larger RAM support (more than 32 GB), or distributed renderfarm set-ups.

[kev030981]

Thanks again all - really appreciate the help. I suppose my last post wasn't very clear, but what I'm essentially asking is will the extra 3GB of VRAM and the double precision point make it worth getting the Titan over the 780 (with specific reference to the programs I use - 3DS MAX, Revit, Photoshop (oh and sometimes sketchup or Rhino) = main ones)

My renders are done in Vray and are often simple clay renders with multiple light sources - a few hours at most.  They are however often quite complex models (high poly) - and this slows my viewport screen down.  I am also paranoid of the project I emabark upon that eventually won't open due to complexity.  So - I'm told that increasing the power of my Gfx card will be my best bet.

My curiousity with the Quadro cards is more of a future dream, right now - a pc that does both gaming and visuals / modelling well is what I'm after.

From the advice given so far I'm leaning towards to 780 and then spending the extra on a SSD.  However - I prefer to get the best possible components one at a time, and that is why I continue to be re drawn towards the titan... But £300 more?? Dunno if its worth it?

Hope you can all bear with me here - sorry if i'm not being clear.  Once again - all help very much aprpeciated 🙂

[dgorsman]

The extra video RAM would be helpful for 3DS MAX for general operations.  Not sure if VRAY is a GPU render engine or not - you'll have to ask more educated folks in the Media and Entertainment forum (they've also got a dedicated hardware board).  From what I've seen, pretty much every small to mid-size user is going with CPU rendering simply because its cheaper to get a faster multi-core processor and lots of RAM which has a number of side benefits for other software.

[braudpat]

Hello

+1 with Dgorsman about Vray : does it use GPU ?

As I said a Core i5 (like the i5-4670K) or a Core i7 (like the i7-4770K) with very good ventirads ou better eventually with water Cool can run (Overclocked) at 4.0-5.0 Ghz, so they will be faster than any Workstation (from HP, Dell, etc)

Core i5-4670K - Socket LGA-1150 - Standard Core speed: 3.4-3.8 Ghz - 6 Gb L3 cache - 4 cores / 4 threads

Core i7-4770K - Socket LGA-1150 - Standard Core speed: 3.5-3.9 Ghz - 8 Gb L3 cache - 4 cores / 8 threads

With a good Gamer graphic card and fast memory, it could be the right choice !?

But the hardware has to be fine tuned and the cooling has to be tuned TOO !

Bye, Pat

[RaoulM]

> Not sure if VRAY is a GPU render engine or not

The VRAY it selfs is using the CPU but the VRAY R/T is using the CPU + GPU and GPGPU setups like the nVIDIA

Maximus one.

The benfit from using the nVIDIA Maximus setp is for me personally even to get more time by saving time on

rendering jobs and also fpr smaller corrections and preview things also at last when the entire project will be 

rendered to finish a job or projekt and this is the real gaol getting me time.

I was beeing 30% all in all faster then without a nVIDIA Maximus setup and I really think it will be 20% - 30% on top

of all with the Maxiumus 2 setup if the Quadro k6000 will be out and combined with the Tesla k20 even also for the postproduction with Adobe CC6 I am saving time against the old days I don´t want to miss it any more.

I  must serching an older thread on this theme, I was reading in June this year, perhaps it is also usefull for

some of yours it is answering http://boxxblogs.blogspot.de/2013/05/nvidia-maximus-and-gpu-rendering.html

Their will be a really good story about nVIDIA GTX cards and the GTX 690 will be shown in action!

Cheers 

RaoulM

[kev030981]

Thanks for replies everyone - This has really helped my research, and influenced my decision. I've posted this question in a few locations, and most of the time it sort of degenerates into a GTX vs Workstation card debate, all of you have been most helpful!  

what I've basically found from research and asking around is that the Titan is faster than a std factory 780, but only by 5% (ish). The OC 780s match and even outperform the Titan - but its important to note this is only for gaming.

When it comes to workstation use - for the types of visualisation software I've mentioned above things become much trickier.  No one seems to have compared a 780 to a titan for professional use (I suppose why would they? Not a common review...) - but there is a very handy review here on tomshardware (http://www.tomshardware.com/reviews/best-workstation-graphics-card,3493.html)

This and a few other sources (http://www.tomshardware.com/reviews/geforce-gtx-titan-opencl-cuda-workstation,3474-23.html) essentially state that the titan is an excellent card, but no match for a professional nvidia quaddro or Firepro when it comes to professional software (no surprise there I expect). This isn't just drivers and reliable memory etc (that help for sure) - but from a viewport speed in maya or 3ds max, what I am hearing is gtx cards are faster with low complexity, and then drop off the face of the earth when a model becomes high poly (say 500 000 polys or more) - where to pro cards don't lose steam until much further on complexity wise.

Even though I focused on several cards (the 780 the titan, and then I may have wandered into murky territory by even mentioning by future dream of quadro cards!) - what I really was trying to find out was what the specs (i.e. extra cuda cores, extra vram) will actually do for my 3DS max (or other software).

In case anyone has the same question I had in future months: all of the tech specs are difficult to define - as each software application differs greatly.

GPU rendering seems to benefit from faster core speed, no. of cores, and amount of vram (basically everything on card).

Viewport speed again benefits from just about everything. 

Complexity of models - or the ability to open a more complex model is largely a vram issue.  So the more the better…

And all of the above can be greatly influenced by drivers and software updates (i.e. 3DS max 2014 has faster viewport speed than 2013, and pro drivers increase gpu rendering times and viewport speed)

To end my long rambling post 🙂 : The solution I have decided to go with is this:

I am getting a 780 (probably an OC one) to satisfy gaming upgrade.  I will then save and buy either a new Quadro K4000 or a used K5000 when I can afford £900, add it to my system and run a duel boot (one having gtx drivers, one quadro). I imagine I'll have to watch the cooling a bit and sapce the cards apart, but otherwise all should be well. Think its a case of right tool for the job here... rather than the titan's’ bridge a gap’ philosophy, I think I;d rather have two cards that specialise in different areas 🙂

Special thanks to RaoulM whose helpful links made my decision more difficult - but ultimately better informed! And to Dgorsman - whose posts I am still going through (you have so many!).

Oh and Vray is a CPU render - VRAY RT is a GPU (From what I've researched)