Adding to Cart…
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2025 Daz Productions Inc. All Rights Reserved.You currently have no notifications.
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2025 Daz Productions Inc. All Rights Reserved.
Comments
I think it's biggest impact on DS would be decreasing the likelihood of needing to use virtual memory which is much slower than physical RAM.
...or possibly having the application crash on rare instances. Most likely though during rendering the process would go into swap mode.
I'm running with 12 GB which used to be more than enough, but lately with the higher resolution textures and meshes, I find my system drops into virtual memory mode with somewhat annoying frequency.
...The Xeons retail for 621$ each, The biggest i7 currently on the market is the 8 core Core 5960X. which runs at 3.0 GHz, not all that much faster than the Xeons I am looking at, and retails for 1,049$. I saw a recent article indicating that a 10 core i7 is on the way but is expected to carry a fairly hefty price tag and still only run at 3.0 GHz.
Granted there are single socket X99 boards with the LGA 2011-V3 socket that support up to 128 GB of DDR4 in quad Channel configuration However most have only three PCI 3.0 X16 slots. My design calls for four GPUs. Only MSI offers a 128 GB board with four to five PCI 3.0 x16 slots. Why 128 GB with the power of multiple Titan Xs? Well, should a process exceed the video memory and drop out of GPU mode, the system would have more than enough resources to handle the load without choking. Basically, that is the "plan B" so to say. I have read accounts about intense rendering jobs which have brought 64 GB machines to their knees.
Daz Studio is rapidly growing beyond being a "3D toy" anymore. When I showed a portfolio of my work to a panel of people in the CG and graphic arts industry recently, they were surprised and amazed when they learned that I created the images they viewed in a "hobbyist's" programme rather than one of the expensive pro grade apps.
Thanks both of you. I'll keep an eye on that.
I was planning on building a new machine, and just found out I will have even more $ to spend on it, I was getting a 960, but if I could swing it the 970, but now that I have more $, I will be getting this 980Ti I am building a gaming machine, and this was the only part that was towards this hobby
Sorry to respond to this so late...
Are you talking about the Xeon E5 2630 V3? If so, I'd highly recommend giving this article from PUget Systems a read before investing in a dual processor system:
https://www.pugetsystems.com/labs/articles/Adobe-Photoshop-CC-Multi-Core-Performance-625/#MultipleCPUPerformance
Granted, it's for photoshop. But the only thing in DS that would benefit from two processors would be rendering in 3Delight. The ten core i7 is expected to be about a grand, which isn't too shappy considering that 5960x is about that much now. In a lot of those tests in the article, performance increases stop after six cores, and in some cases it gets worse.
If you feel you must have Xeons, save up for a single 2667 v3 or 2687 v3WS.
I'm trying to figure out if I can build the system but maybe spread the cost by maybe installing 8 gigabytes of memory and then upgrade another eight gigabytes to give me 16 down the road, and maybe use my hard drive for my from a dead PC (I think it's 1Tb) and use that and then upgrade to a SSD drive later and switch the other drive for storage
If you buy matched memory, that *may* result in less hassle. That's what I do, but it costs more up front.
IMHO matched memory is still the way to go but as technology has moved forward RAM has become far more temperamental if you can't do it at the time. 10 Years ago I'd say "for the love of great Atheismo! DO NOT MIX AND MATCH RAM!", now I would just tell you try and buy it all now, or if you buy later get the same type and brand or at the least the same speed so it does not slow down the rest of your RAM, which it will do to ALL your RAM if it's slower than what you had in there before but it won't necessarily destabilize anything.
Mismatched memory will result in the Northbridge chip on the motherboard running all the RAM at the lowest speed and CAS. Furthermore, because of this, Dual/Quad channel benefits will be disabled. This means the machine will run slower than it could.
ALWAYS try to buy matched pairs of memory (and make sure they are installed correctly!) If your machine supports Dual/Triple/Quad memory, always by matched. Dual channel boards almost always have 4 slots, so buy one pair, then later upgrade with another pair (the pairs don't have to match each other!) Triple channel (the old 1366 boards) buy triplets and Quad Channel buy sets of 4. It's slightly pricier, but the performance gains with memory are across the board. The improvements are small, but consistent and continual.
Whoa, entry level for build it yourself is now as expensive as Apple HW. I think I'll stick to Wal*Mart, Amazon, & eBay. :-)
Mismatched brand or mismatched speed? I already stated mismatched speed will lower everything to that lower speed but OC's have reported mismatched brands are not as problematic as they were years ago. I've set up several Xeon systems that were attained a level of stability I didn't think possible with mismatched batches of RAM could provide a decade ago. I'm not calling you out but if you can provide me with some context I'd be curious to read it for my own benefit on the subject.
http://arstechnica.com/civis/viewtopic.php?f=8&t=1272285
thanks for working this up, JCThomas. it's just the startingpoint i needed.
very nifty to have a range of assemblies with explanations and comparisons...and prices, lord knows...
not to mention the contribution of so many other knowledgeable DAZ folk.
If you are making this box for Iray? I would go with older i7 3960x and a Intel x79 chipset motherboard with 8 ram slots of DDR3 and get a cheap 64gigs of ram. I would also get a board with at least 4 full PCIe slots for your Iray Nvidia graphics. Next get yourself a cheap card to run your monitors for primary slot and get the expensive cards just for rendering. With Pascal just around the corner I would wait for them to comeout so they will drive the cost of a titan x down a little before purchasing them for this system.
Mismatched speed is the most critical, but even brand often has 'different' characteristics. While the base clock speed for the RAM is the primary concern, followed by CAS level, and then voltage required (it varies from manufacturer to manufacturer and from speed to other speeds) there are also quite a few other parameters that are inherent in a given stick of RAM.
It is best to buy actual MATCHED PAIRS (they come together in the single package). They'll come from the same silicon wafer batch, and they should have virtually ALL characteristics at the exact same values (within a tiny fraction of variance.) Using two sticks of the same 'speed' but different manufacture can end up with some of those 'less known' parameters not being the same....and can end up with instability in the system, since if speed, voltage requirements, and CAS level are equivalent, the Northbridge chip on the motherboard will try to run them identically in dual-channel mode. And if they aren't the same, you may get VERY intermittant issues.
If you must run 'equivalent' mismatched pairs, be sure to run a burn-in using MemTest86 for at least 24 hours.....any issues at all, and you'll need to run the two in different slottings so dual channel gets disabled.....
Hello JCThomas and friends,
Could you please help me with this 3D render/gaming/VR/music build? It's based off of JCThomas's high-end 1151 build. https://pcpartpicker.com/user/uhlek/saved/#view=VxHwrH
I am building it for:
* FPS gaming at 1080p 144Hz
* 3D rendering in Octane Render and Iray
* VR & Oculus Rift
* FL Studio music production
I already own an old nVidia GTX 780 that I'll add to this build to boost its performance in Octane and Iray.
Am I overdoing it on the power supply? I plan to upgrade with 1-2 more nVidia GTX 980 Ti once their price drops in several months.
Thank you!
...the 10 core E5-2687W v3 is 2,200$ whereas the dual 8 core E5 2630 V3s would be under 1,400$. I also do use 3DL as well.
Ideally I would like 4 Titan-Xs for the 12 GB of Video memory, but most likely it would be four 980 TIs so there would still be a fair chance of the render process dumping to CPU and memory. I do pretty "big" scenes so I need all the "overhead" I can get.
The commercial HBM 2 Pascal GPUs (16 GB) won't be out for at the very least, another year, maybe longer (the only model to be released this year will be a Tesla compute unit). Granted, with the calculation speed advancement of the Pascal architecture, I could probably get away with only two of those instead of four 980 TIs/Titan-Xs.
I'd say it's a great build. If you're planning on adding 2 more 980 Tis for a total of 4 gpus, then, no, you're not overdoing it on the PSU. However, unless you have one of the coveted 6GB models of the 780, your system will only be able to use 3GB of VRAM. Either let the 780 sit it out, or let it drive your monitors. Also, don't expect too much of a price drop. Once stock becomes limited, the prices don't fall all that much. That was the case with Kepler cards after maxwell came out at least.
Wrong - and right, at the same time.Iray will use all the cards in a system, providing there is enough memory to hold the scene. If you have 3 cards - 3 GB, 4 GB, and 6 GB, all tagged for use by Iray, then Iray will load the scene into all three cards. If the result is 3.2 GB, the 3 GB card drops out and the scene will render on the 4 and 6 GB cards. Hit 4+ GB and the 4 GB card also drops out, leaving the 6 GB card to do all the work. And if the scene blows past 6 GB you end up with all the work done by the CPU.
Now where it is right is using the 780 to drive your monitors. You can lose anywhere from 50 MB to nearly a GB of Vram just driving your monitors - and a lower-end card with the monitors attached can get soooo busy doing cuda render calculations that screen updates get real slow. Like so slow that windows solitaire is unplayable. :-) So plug the 780 into the box, but don't try running renders on it - use it for the monitors.
Other comment - I'm running much the same cpu cooler (I've got the 12S, with only one fan on it) and am running the 3.5 GHz 6-core I7 at 100% for very long times with the core temps never getting above 55 or 56 C. The big question - what kind of box are you going to put this in? I've an Antec with 3 120 mm intake fans in front of the drive bays, 2 120 mm fans in the back as exhaust fans, and a 210 mm fan on the top, also as an exhaust fan. And if I ever get around to adding more GPU cards, there's a slot for an intake fan on the side, over the PCI slots. Lots of ventilation leads to lower fan speeds and noise and happier components. :-)
Yes, it is the 6GB 780 GTX. I have a 3rd video card, an older AMD GPU, card I will connect to my monitor when using Octane. That will permit me to use all the CUDA cores on the both 980 Ti and 780 for rendering. Or should I just sell my 780 GTX, and instead go for 980 Ti x 2 EVGA hybrids?
I plan on switching my display to the GTX 980 Ti when gaming. :)
I'll be housing the system in a Phanteks Ethoos Pro case. There should be a lot of air space and room for the large Noctua GPU fan. I think I will install an EVGA GTX 980 Ti hybrid with liquid cooling, with plan to overclock the GPU for faster Octane/Iray renders.
Thank you for the feedback!
if you plan on getting more than one 980 hybrid, you need to think about where you will mount the radiators for them. Especially if you have the AMD card and/or 780 in there. The pipe length is generous, but not unlimited, and you dont want to go installing the radiators somewhere where the pipes have to bend around in some crazy S shape. That will put unneccessary strain on them, and you will probably develop leaks eventually.
I almost got twin GTX titan X hybrids for my build, but went with upgraded air coolers instead. Everyone i spoke to said you either have all of your GPUs cooled by a custom water loop, or you air cool them.
...yeah air cooling saves interior space especially if you are running 4 cards. The drawback is finding a case that has superior built in cooling and still is large enough to house a quartet of Titan/980 TI sized cards as well as drives and other components. The best type of case would be one with not only multiple fans front, top, back and bottom, but a large fan and vent in the left panel right were the GPUS are. Sadly, the popular trend in case design seems to be having a window there instead.
I was wondering if one could actually build a custom external "VCA" using four hybrid Titan-Xs. What would be involved in the design and how would it connect/"talk" with the main box?
12288 cores would come close to providing almost "realtime" rendering speeds (as Mec4D illustrated, even two Titan-Xs are pretty bleedin' fast).
I saw this a while ago. It uses PCIe riser cables to connect the gpus. Looks awesome, but it would be extremely loud.
http://tomglimps.com/custom-pc-for-otoys-octane-render-by-polder-animation/
Several Cooler Master HAF cases have side fans. I have a Lian-Li PC A75 with two 140mm slots on the side panel. Several corsair cases have side panel fan mounts also.
I just purchased the Cooler Master HAF case as they are great for cooling as the side panel not only have fan but the rest is a grid , I run 10 fans now in the case , one side, one bottom , 2 front , 2 double back, and 4 x for my water cooling of my 2 Titans and yes everything is very quiet controlled by Asus mother board , and fast fans don't mean better cooling but the right pressure . I had to switch to water cooling for my GPUs and that was the best choice I ever made , now idle 25 C, full load 37 max with 21 C ambient room temperature before it was around 67C on full load and without CM HAF case and side fan 80C
I was for plan to build custom VCA using PCIs riser cables but many people not recommended doing it unless you have 100% good PCIs riser cable as many of them can simple damage the card as they are not created for this kind of work .
with my power supply 1300W I can run 3 Titans X without worries as they need 500W total and 1300W can handle 3 titans as well but don't think that 4 x Titans will do the job like 4 x Titan X since you have just one processor so the result will be slower as you expect for real time .. If I put 1 genesis or 20 they still render at the same speed in real time or while rotating the scene , actually viewport is slower than Iray viewport when rotating , however rendering with one card or 2 there is not 100% improvement and I don't think Iray will use 4 cards to render in realtime even if it show it does the other card is half as warm and for the system I use HD intel graphic so my cards are used only for rendering .
I may go with one more card but not now , I waiting for Nivida to make better card before I do as I am pretty happy with my 2 x super clocked Twins as there is no waiting time anymore and 4 titans will not make it faster in the viewport , better CPU will do as it need to process the data to the cards before it render so things like instant does not exist in iray view port
with the next update of my rig I go for mother board that support 4 way SLI , my MB support now 2 ways and the water cooling radiators of my GPUs going outside the case and in place my future new CPU will get 240 water cooling radiator on top.
The CM HAF have a lot of space to do cool things and after I changed the side panel to window with just a side fan it looks much better with better pressure and less less dust , I can use magnet filters on top and problem solved and all is very quiet thanks to closed water system cooling .
Below my Hulk .. still with the original side panel now it is plexy glass with smaller side fan as I don't need so much air to flow anymore thanks to water cooling
I run 2 x SC Titan X with 2 Titan X and manually added 2 Hybrids water cooling sets for Titan X with double fans on both sides for each radiator can't be more happy with .
Curious MEC4D - I run 2x Titans as well - 1 Titan is quite as a mouse - the other sounds like an Indy car shifting gears... What do yours sound like on long renders?
Thermaltake Suppressor F51 E-ATX Mid-Tower Chassis (ultra quite but cramped for space - requires extra small hands to operate)
MSI x99a MPOWER LGA2011v3 MB
Intel Core i7-5960x (3.0 Ghz / 8 core / 20MB Cache)
64 GB Kingston DDR-4 (2133 mhz)
1300 Watt EVGA Power Supply
Kraken x61 Low Pro - Liquid Cooling Kit (CPU)
Win 7 Pro (Test Build / Designed for Professional Software Developers)
4x 250 GB SSD Raid 1 (5.25” hot swap carriage)
2x 4 TB HDD (Mounted to Case)
2x EVGA GForce GTX Titan (12GB)
2x Msi GTX 970’s
That is a lot of cards for so small case , my 2 x SC Titan X don't make noise since they are water cooled , plus I have plates to cool Vram so you don't know they are even working , the temperature get max 37 C after hours of rendering usually 32C . Sometimes I place a hand in the back to see if they are working lol as no sound at all and not very hot air escaping the radiators
But when I got them last year and placed in mid-tower without water cooling I could dry my hair in the morning while rendering and the internal fan blow like crazy especially the one that was first as the second one below did block a little the air flow .
I suspect that the card that make more noises is the card that get less air flow so the fan run faster to cool it down and they can be noisy , I use Precision X to control my cards fans on the radiators , actually replaced the original with Thermaltake fans as they are efficient and so quiet
Thanks MEC4D - I'll have to look into that Case + liquid cooling - what did you mean by "plus I have plates to cool Vram" - I'm not familiar with "plates".
Also – my 970’s are now out of the case – got real hot with those in… I copied the PC specs off my DA account page and I guess it still lists them as functioning cards. So only 2 x Titans now – even still, neck breaking rendering speeds at ultra-high rez.
EVGA Backplates for Titan X series , they cool down the memory couple degree lower also protecting the cards
I got them together with my Hybrid water cooling
here separate EVGA Backplate
http://www.evga.com/Products/Product.aspx?pn=100-BP-2999-B9
and Titan X Hybrid water cooling manual install + Backplate
http://www.evga.com/Products/Product.aspx?pn=400-HY-0990-B1
I hunted both for 6 months before I got them always out of stock but man , the difference is shocking ! how lower temperature of the GPU how faster it process
...hmmm, good point about the CPU. I may need to rethink my design a bit. Sadly most 8 and 10 core Xeons are slow, in the 2.3 - 2.6 GHz range. I don't bother overclocking as I am not a gamer and it shortens the lifespan of the CPU. i7s offer better CPU speeds, but not sure I can find an MB that would support two of those and 128GB of DDR4 memory in quad channel mode.
looked at the HAF series, just not into the window, lights, & all. Nice that it can fit several Hybrid Titan-X's though
-----
The first Pascal GPUs are to be released this year (June target date). From what I could gather, for us enthusiasts they will still have the same form factor and use GDDR5 (GTX 1070) or GDDR5X (GTX 1080) instead of HBM 2 but have 8 GB memory and a GPU chip with 8 billion transistors. CUDA core count is projected at 2,560 (less than the Titan X's 3072) but the GP104 processor is rumoured to have a faster calculation rate so possibly needing a boatload of CUDA cores for rendering speed may be a thing of the past. It is tempting, though 8 GB is about borderline for most of my scenes (hence the dual CPU and high memory setup above). Still if they are faster, and cost less than a Titan-X, I may consider them. Finally an 8 GB GPU that isn't a Quadro.
The first HBM 2 Psacal GPUS to hit the market will be Tesla compute units only. Really hope this technology becomes available to us soon. 16 GB of Video memory, for me that is a dream come true. It would mean 99% of my scenes could render in GPU mode.
No such board exists because i7 does not have any link communication to run in dual socket , there are plenty of MB that support other CPU and the memory you looking for , my old MCI supported 128 GB .
I am not so sure to the end if Iray will use all 4 cards in the system for rendering , I know Octane do as it is software related .. Blender does not so you can have 10 titans x and not improvement in speed as it use only what it need skipping the rest . It may load scene for all cards but process less with other cards and I do not see difference in the view port if I render in real time with one Titan or 2 only when render in window . I am not sure about the new Nvidia cards , I may wait a little longer to see them all in action but once released Titan X will get lower in price so I may snap one more for the collection , I have still one more spare water cooling hybrid I got from EVGA so I can use it as I still have place inside for one more in the front panel and with over 9000 cores I will be good for a long time still I am fine with what I have right now .
Regarding the HAF case , it is design for water cooling and have one of the best air flow if you do not use water cooling and highly recommended by many shops , I did not believed thinking someone try to sell me old case but I saw the difference in performance and temperatures before I installed water cooling for my cards , I did not cared too much about the military look just wanted to have better air flow side fan and a lot of space for future updates and improvements and this thing is huge and heavy .
I need light to see if my fans are working since not much noises with water cooling and if one of them fail I would not notice that and damage easy my cards while rendering so not much of decoration here, I set the light on full light power for the pic , but they just glow slightly on regular basic . It also indicate the power of the system if it is on Auto, Performance or Quiet mode so handy plus I have double side fans just in case as you never know ..so my twins are protected lol