workstation quadro p4000 vs gtx 1080 ti

[Immitem]

I am trying to multi-quote here and I really do not know how so bear with me.

15 hours ago, O_and_N said:

As a final,which gpu do you prefer to keep on going in maya?

 

 I will be keeping the w9100 as I often work with hardsurface models consisting of thousands of objects and millions of polygons and the constant pausing with the other cards is a deal breaker. Since I use Redshift I actually have two 1070s in addition to the Titan XM and will be keeping all of them for GPU rendering. Despite the differences in the viewport performance the 1070 and Titan XM render roughly at the same speed. 

 

2 hours ago, silent_sight said:

It looks like your test is with a pretty huge scene and I'm curious if there's much of a difference with a smaller scene (like an average sized character, for example).  I'm mainly just curious... I wonder if a lot of people who do not have any issues with gtx cards simple haven't pushed it to the point where the quadro really shine?  If that's the case, it would be helpful info for people when choosing a graphics card based on their maya usage (if you're not likely to work with such huge files, then maybe a gtx card really will be enough for you).

 

Does anyone know how the AMD w9100 compares to the quadro p4000?  From very briefly looking into it, it seems the w9100 is really old, so I'm guessing the p4000 would be a lot better, right?

 

 With a single character under a million polygons viewport interaction should not be affected if you are using a geforce card. On the other hand 100 10,000 polygon objects most certainly would. It depends on what you are working on and if it is only a single character with a few objects (character plus clothes plus effects) then the Geforce is the way to go.

 

 The w9100 has 5.2 Teraflops of raw processing power and games just as well as a 290x whereas the p4000 has 5.3 and thus would likely perform the same though there is a significant amount of hearsay that AMD cards, even their consumer Radeon cards, perform better than Quadros in Maya. This was true as far back as Maya 2013 but I do not know if there is any truth to this today. I will say that the Firepro suffers the same problems that my Geforce cards experience in Maya in 3DS Max 2014. Lastly, unlike the AMD drivers the Quadro drivers tend to be utter ass for gaming so if you intend to do any the P4000 would probably get roughly the same performance as a 1050 with an inconsistent FPS. 

[scottyseng]

2 hours ago, silent_sight said:

-snip-

You might also consider the Radeon Pro Duo (Yes, it's a workstation GPU...they replaced the FirePro branding with Radeon Pro). I hear it's a killer value for the price.

 

It's a bit tough though, most GPU Renderers are biased to CUDA cores (Though you'd want to use a GeForce card for this since they have more bruteforce power)

[O_and_N]

10 hours ago, silent_sight said:

immitem - thanks for taking the time to do your testing; hope you didn't pull your hair out with all that lagging with the gtx cards!  It's really interesting how much better the pro card worked than the gtx cards - seems like the only real exception was that the gtx cards had slightly higher fps in VP2.0, yet they lagged a lot.  I think I would rather have slightly slower fps with little-to-no lag.

 

It looks like your test is with a pretty huge scene and I'm curious if there's much of a difference with a smaller scene (like an average sized character, for example).  I'm mainly just curious... I wonder if a lot of people who do not have any issues with gtx cards simple haven't pushed it to the point where the quadro really shine?  If that's the case, it would be helpful info for people when choosing a graphics card based on their maya usage (if you're not likely to work with such huge files, then maybe a gtx card really will be enough for you).

 

Does anyone know how the AMD w9100 compares to the quadro p4000?  From very briefly looking into it, it seems the w9100 is really old, so I'm guessing the p4000 would be a lot better, right?

Yeah,the w9100 is old( 290x game gpu) but still good.In my opinion the p4000 shoud be even better(the core of a 1070 but litle cripled to i hope no less than a stok gtx 980).I only wish it had more vram like 10 or 12.

 

https://www.techpowerup.com/gpudb/2930/quadro-p4000

 

Carefull as some of the specs of the gpu on this page are not correct and i suspect they are a copy/paste of the p4000 model ment for laptops as the share the same brand name.The page say 4.3 Tflops but nvidias page say 5.3.Or we can see on the link that bandwich is 192 gb/sec but yet again on nvidia it say 243 gb/sec.I supose we need a gpuz screenshot of the real specs of the p4000.Honestly,for this ammount of cash it woudnt hurt Nvidia to show full specs on their page!!

[O_and_N]

11 hours ago, Immitem said:

 

 The w9100 has 5.2 Teraflops of raw processing power and games just as well as a 290x whereas the p4000 has 5.3 and thus would likely perform the same though there is a significant amount of hearsay that AMD cards, even their consumer Radeon cards, perform better than Quadros in Maya. This was true as far back as Maya 2013 but I do not know if there is any truth to this today. I will say that the Firepro suffers the same problems that my Geforce cards experience in Maya in 3DS Max 2014. Lastly, unlike the AMD drivers the Quadro drivers tend to be utter ass for gaming so if you intend to do any the P4000 would probably get roughly the same performance as a 1050 with an inconsistent FPS. 

That was somehow true about quadro gpu(but only until the start of the k4000 k5000 and k6000 and for the cheap ones)The mid-high end ones starting from the number 4000 and up game/and game dev just fine)It autodetects the app similar to amd but there are also profiles that you can select from nvidias panel.I still usi a k4200(had a old v8800 before that-a real champ) and it performs a litle better than a gtx 660 which it actually represents.

 

The real thing beween quadros and firepros is the the firepros do tend to keep the same clock speeds as the gaming gpu they represent while quadros lower the clock to a degree so that its more power efficient.Never liked that move from nvidia.Still i see that the p4000 is vr sertified and i still hope its the power of a gtx1060/980.

Lets hope wee see some real benchmarks soon!I saw a few one that show 3d mark orange room in action but that doesnt tell me the full story.A 3d mark  11/firestrike test scores woud tell the truth.Hope someone does it.Loud hate to be the first one to do it lol.

 

[silent_sight]

just checking if anyone has any new information yet...

 

I've been doing a bit more research to try and determine the relative performance of the p4000.  Since there doesn't seem to be much (if any) benchmarking between the different quadro series (Kepler, Maxwell, Pascal) I don't really have anything concrete... but here's what I did find out.

 

-The p4000 is 1.7 times more powerful/faster than the m4000 in relative performance which you can see here: http://www.anandtech.com/show/11103/nvidia-announces-quadro-p4000-p2000-p1000-p600-p400-finishing-the-quadro-pascal-refresh

 

- something I read somewhere stated the m4000 being equivalent/similar in performance to the k6000... I can't remember exactly what it was, but I think it was a speculative review where the person writing it was looking at stats of the upcoming m4000 release (at the time, of course).  I could not find any graphs or anything actually verifying this - not even anything comparing relative performance between the k6000 and m4000.  I really wish I could find the website where I saw that again... but I'm not even sure how deep into google searching and cross searching I was... bleh

 

- I also found this http://hothardware.com/reviews/workstation-war-amds-firepro-w9100-vs-nvidias-quadro-k6000 which shows the k6000 performing slightly better than the w9100

 

so... provided the second item above is actually true, the p4000 would be a little more than 1.7 times better than the w9100 (which Immitem showed to have superior legacy viewport performance than the gtx 1070 and titan XM, as well as no lag in manipulation while the gtx cards all had some).  Of course, the other thing to note is that the 1.7 times relative performance of the p4000 to the m4000 does not necessarily mean that maya's viewport performance will be 1.7 times faster...

 

It still doesn't really help with the question of the gtx 1080 ti... I'm really just assuming it would perform better than the 1070 and titan XM from Immitem's test results

[O_and_N]

From what i know the m4000 is as powerfull as the k6000 in stuff as maya but in brute force gpu mode(lets say gaming or other similar task)it isnt.From all of the reviews i have seen>the k6000 is a gtx 780ti and the m4000 in brute force is no more powerfull as a gtx 680 thanks to its wayyy to low clocks.A m4000 has the core of a gtx970 but when nvdia decide to lower the clocks(which is stupid excuse for power managment) the m4000 renders sadly as a gtx 680.

 

Same with the p4000(but not that bad as the m4000).but is Vr sertified so it must be a good one.

It has the core of a gtx 1070 but with litle cuts.

 

 

If they are not lying,than in the 3d mark tyme spy score of 5134 for the p4000 shoud indicate that is similar in raw power to a stock gtx 980?

http://hothardware.com/reviews/nvidia-quadro-p4000-and-p2000-pro-workstation-gpu-review?page=6

 

 

 

 

 

 

 

 

[silent_sight]

I found this pretty interesting: http://www.tomshardware.com/reviews/nvidia-geforce-gtx-1080-pascal,4572-9.html

 

It's the gtx 1080 (not the 1080 ti) being compared to Maxwell quadro cards and some firepro cards.  Granted, they're testing the direct X performance, which we could already see performing well with gtx cards from Immitem's tests.  I'm also not really sure how they're testing it because I can't seem to see it specified, but my guess is some heavy poly models in a turn table...  it would have been awesome if they tested open cl with the same cards... oh well.

[DUS x SSJ4 x]

Been looking into this subject too. Found this thread through my googling and googling trying to find an answer.

I just started a new job as an Industrial Designer and work in Adobe and 3DS Max.

The machine they have me on is an old Lenovo c20x thinkstation. Which was fully spec'd when it was ordered in 2011 or something.

CPU- Dual Xeon X5650 (6 cores 12 threads) 2.66ghz
RAM- 24gb DDR3? but the MB is full.
Boot - 512gb SSD
Storage - 2 tb hhd
GPU - Quadro 4000 (2gb GDDR5)

the gpu is what has been causing me problems. Well at least I think. Basically unless I use a basic reference setting with no lighting, shadows etc. I could not even get one render out over night with decent mental ray render settings.

Looking at the system a GPU upgrade seems to be all it would need. Sure a new system would be better, but at huge cost.

A Quadro P4000 and GTX 1080 ti are almost the same price. So its been hard to decide what I should go to them with as my proposed hardware change.

Would the surrounding hardware bottle neck either of these gpu's? doesn't seem like it would. RAM isn't amazing, but its still a decent amount. I also might add, im not 100% sure a 1080 would fit in the case surely it would, but the P4000 is the same size as the 4000 in it now.

So yeah wrestling with the same problem. Would love some extra input from people. Keep in mind, creating some 3ds max renders is not the main part of the job. So its not like the system will be running 24/7 at 100%.







 

[silent_sight]

On ‎5‎/‎29‎/‎2017 at 1:12 AM, DUS x SSJ4 x said:

Been looking into this subject too. Found this thread through my googling and googling trying to find an answer.

I just started a new job as an Industrial Designer and work in Adobe and 3DS Max.

The machine they have me on is an old Lenovo c20x thinkstation. Which was fully spec'd when it was ordered in 2011 or something.

CPU- Dual Xeon X5650 (6 cores 12 threads) 2.66ghz
RAM- 24gb DDR3? but the MB is full.
Boot - 512gb SSD
Storage - 2 tb hhd
GPU - Quadro 4000 (2gb GDDR5)

the gpu is what has been causing me problems. Well at least I think. Basically unless I use a basic reference setting with no lighting, shadows etc. I could not even get one render out over night with decent mental ray render settings.

Looking at the system a GPU upgrade seems to be all it would need. Sure a new system would be better, but at huge cost.

A Quadro P4000 and GTX 1080 ti are almost the same price. So its been hard to decide what I should go to them with as my proposed hardware change.

Would the surrounding hardware bottle neck either of these gpu's? doesn't seem like it would. RAM isn't amazing, but its still a decent amount. I also might add, im not 100% sure a 1080 would fit in the case surely it would, but the P4000 is the same size as the 4000 in it now.

So yeah wrestling with the same problem. Would love some extra input from people. Keep in mind, creating some 3ds max renders is not the main part of the job. So its not like the system will be running 24/7 at 100%.







 

I think it's going to depend on what you're doing...  You've really only mentioned the rendering and how you've not been able to get a decent render overnight.  I personally don't do much (if any) rendering; however, I do know that it is predominately cpu based - dependent on the speed and number of cpu cores as well as the amount of ram.  But in recent years, gpu rendering has been coming along pretty strong, which also performs best with multiple cores and ram (as in cuda cores and vram on the card/cards).  So, if you are doing gpu rendering, then you'll most likely want to go with the 1080 ti since it has twice the cuda cores and almost half again as much vram as the p4000.

 

I actually ended up deciding to go with the quadro p4000 card.  One of the main reasons for this had to do with the graphics card testing that was done in Immitem's post at the autodesk forum thread started by O_and_N (which he/she linked to).  All of the gtx cards would hang when it came to interaction within the scene (either consistently, or intermittently), and that would be excruciatingly aggravating for me.  Also, the legacy viewport speed was consistently slow with the gtx cards and the workstation cards just performed better there.  Even though the gtx cards did have higher fps in viewport 2.0 for maya, the workstation card's fps was still well over 30 (or even 60 for that matter - anything over 24 is pretty much gravy anyway).  Between the p4000 and p5000, the rather minimal relative performance boost between them in maya didn't really warrant the $1,500ish difference in price.

 

Now, having said all that, you are also using some adobe programs and will probably want interactivity within 3ds max.  I don't know if what you will be doing there would need quadro drivers or not... I would guess the quadro card would perform well in 3ds max for interactivity, but I'm really not well versed enough in 3ds max's interactivity to know for sure.  It's possible the 1080 ti's interactivity could be acceptable if the primary focus of the machine is for rendering even if it lags here and there when manipulating your scene.

 

Either card will most likely perform MUCH better than the quadro 4000

 

I hope this helps, but I also hope there are some other people who could give you some different/supplemental info that ends up being more helpful than mine...

[Plug]

old but a good test to see
gtx vs kepler arc

 

[LG25]

On 3/24/2017 at 12:49 PM, Princess Cadence said:

 

Kind of funny. He laughs at the DVI port as something for 10 year old equipment, but...

[Abounding05]

I'd call Nvidia. Often these class of products have a sales team dedicated to helping you figure out which is the right one.