I would personally prefer the 9900K for its out of the box top single threaded performance.
With 5K$ I wouldn’t compromise with anything else than the 2080Ti, with the hope that some day Cycles gpu rendering will be able to use nvlink, which is the greatest new feature of the RTX series, i.e. the capability to add Vram via nvlink bridge and use it as one greater memory pool (2x11=22gb Vram).
The EVGA 1600W T2 is one of the best psus out there (based on a Superflower platform) and will be able to handle up to 4 x 2080ti’s if needed.
The MSI MEG Z390 Godlike is the only Z390 that’s suitable for gpu rendering with its plx chips. Pcie bandwidth is really needed only just before rendering begins, during the transferring process. Usually even x4 is adequate for this job, but with this motherboard every gpu slot will have the bandwidth it needs the moment it needs it.
Finally, as an alternative build you could choose a 2950X-Z399 combo. It would be slightly faster in final renderings (with hybrid rendering), but for anything else, where single threaded performance is paramount, it would be slightly slower compared to the 9900K.
the 1080tis and threadripper are a solid choice, quadros will be of no benefit over the 1080tis (except maybe power draw) and they cost a lot more, you can see in this benchmark the 1080ti is equal or faster than the quadro p6000 which is about 6 times more expensive at aud$6999 compared to the 1080ti at aud$1100.
I’v heard people say they haven’t had good luck with asrock in the past but their current lineup of mother-boards are pretty decent, also while evga and corsair do make some cheap low-end power supplies their higher-end power supplies are usually quite good you just have to read some reviews, seasonic are a good brand as well though abit expensive.
I don’t think the number of pcie lanes matter too much with rendering,
the scene data is sent to the gpu at the start of the render then all the work is done on the gpu without much data going over the pcie lanes while rendering,
you could probably get away with running the cards at x4 or maybe even x2.
Finally was accepted by Auto-Notify that EVGA has availability and that I was eligible to purchase two 11G-P4-2281-KR 11GB GPUs!!! Set me back ~$2200.00
I’m working on getting the I9 and mobo now.
Thanks everyone, can’t wait to see how this build goes!!
My suggestion for the specific motherboard was not well documented. I would change it to this one for multiple GPU support https://pcpartpicker.com/product/m9YLrH/asus-ws-z390-pro-atx-lga1151-motherboard-ws-z390-pro
The MSI MEG Godlike doesn’t have a plx chip eventually. Most people thought that it would. The Asus WS is a better option for this kind of use, and it’s much cheaper too.
I’m really sorry for my poor advice on that. I hope it’s not too late.
Thank God you managed to cancel the order. Hope the new system is lightning fast. I’d like to see some photos when you finally build it. Keep us informed, if that’s ok with you.
Just joined and my first post; I hope the OP doesn’t mind a wee thread-hijack!
Due to a modest inheritance I am in the position of a “once-in-a-lifetime, money no object” PC upgrade.
However, I am budgeting for £5000 (UK).
My question is really about i9 vs threadripper.
I plan to get a GTX2080TI to handle the rendering (although I run 2.79 nightly with CPU+GPU enabled) so I am wondering which would benefit Blender more in a day-to-day editing scenario?
(I do spend much more time editing than rendering so that is where there will be most benefit for me)
The main problem with my current system (i7-6700K, GTX1080, Win 10 Pro 64bit, 32G RAM) is the PC’s response during editing.
Would an i9 make more improvement to this rather than the multi-core of the threadripper?
I have read conflicting comments about Blender and its multi-core optimisations. I suspect during the editing process, multi-core has minimal benefit?
Most of the stuff we do while editing are single threaded. Very few tasks can scale to more cores and just a handful of them can really absorb the full power of more than 6-8 cores at the same time (mainly rendering and file exporting/compression/encoding etc.). So, getting the best STP (Single Threaded Performance) would give you the fastest editing power you can get. The debate in you case is how much could you gain in this field if you upgrade to 9900K coming from a 6700K based system. At stock frequencies the 9900K is almost 20% faster in STP tasks compared to 6700K, but then, the 6700K could be easily OC’ed to a stable 4.6-4.7GHz and come closer to the 9900K in STP, roughly 10% slower. The 9900K is already stretched enough in clock speeds, so I wouldn’t bother OC’ing it by an insignificant 100-200MHz in order to get 2-3% better STP.
Another thing about your current rig is your motherboard and if it’s capable of having more than 1 gpu installed for rendering, if you plan to do so, which is the logical thing to do with this budget and for gpu rendering with Cycles.
One thing is sure, that with 5K£ you can get the best hardware for your workflow. Choices are infinite.
I’m a little bit concerned, though, about what you pointed out in regards to your system’s responsiveness while editing. I don’t think it’s the cpu. Could you be more specific on that? I’m asking, because the reason behind this sluggishness could be something totally irrelevant with the cpu and of another nature, and just upgrading the hardware wouldn’t solve the problem.
I put the i7 system together 2 years back and I have always felt it was sluggish.
(I did actually have a lot of trouble with the build and went through 2 x mobo, 2 x RAM 2 x CPU!)
Mobo is Gigabyte GA-Z270X
In an effort to improve things I now have my Blender files on a 2T NVMe M.2 drive.
I was never planning a second GPU in the i7 system nor in the proposed i9.
Most of my time is toying around editing with the occasional render so STP, not render, is where I am focusing.
My plan was not to upgrade the i7 but build a whole new system.
The performance issues are most noticeable if I have a few Makehuman figures and get bad latency when selecting armatures. It is actually very reminiscent of ye olde memory page swapping, taking several seconds to respond!
Hmm. The only part that was constantly running through this process without changing was the gpu.
Do you have Nvidia’s latest drivers? Always check clean uninstall when updating drivers.
Also, could you post a screenshot of your System’s preferences (inside User’s Preferences menu)? This might be an OpenGL thing. I’m almost certain that it’s not your cpu which causes these issues.
My cpu is much slower than yours in STP (5960X), and my gpu is also a bit slower (1070) and I’ve never felt Blender to be so sluggish while editing stuff. So, there must be an optimization issue with your gpu or OpenGL preferences.
Not had much time to play this w/e but the selection certainly appears more responsive.
Thank you.
I still plan to order a new system, but not in such a rush now.
I expect to go the i9 route but the i9-99xxX’s are just not around.
Only i9-99xx CPU in EBuyer (who I always use) is the big 9980XE at £2500, which is too much!
(I don’t mind spending a little over the “bang-per-buck” point, but I don’t think an extra £1k will bring that much performance increase!)
I am tempted to get an i9-7940X, but I don’t know what I am missing from an i9-99xx series.
A list on Wikipedia shows cores and clocks which I can compare but I think there’s some other improvement.
Is there a simple answer? A link?
Honestly, for your workflow you don’t need so many cores. For mostly editing and for a single gpu system, I think the i9 9900K is the best cpu for you right now. It’s currently the STP king, and will be until the new Ryzen cpus come out. They are said to be close enough in STP to the 9900K, but the 8c/16t chip’s price will certainly be much lower compared to the Intel’s flagship. The bad thing is that they aren’t coming out before summer, so, it’s a long time to wait.
If you need an opinion for your new rig, I’ll be glad to help.
Sure, I build all my rigs by hand myself. BUT there is a level of money (Over £3,000) i would rather have it built, just for being able to send it back if don’t work. The company above does custom builds to order - just phone them up, tell the exactly want and build it they will.
But of course, if you are no stranger to rolling up your sleeves, you can create a monster for a lot less then £5k…
I have put together plenty of PCs over the previous 10 years or so. My current i7 was the first I have had REAL trouble with. I suspect I accidentally damaged something and perhaps, as I feared, a bad part damaged good parts giving a crazy CPU<>MOBO<>RAM vicious circle! (Although I reserve the right to wrong )
This is part of the reason I want to build a total new system; it is still possible that I have over-stressed some parts and something is not playing ball. I have 32GB installed but it sometimes feels like 4GB…
I did consider getting it pre-built. I looked at Scan but I find their site difficult to use. Not tried OverClockers, maybe I will.
And returning briefly to the suggested i9-9900K, at “only” £500 I could afford to stick in a second GPU…
I am grateful for all input but I think I would rather move into newer CPU format, LGA2066 rather than staying with LGA1151. I have no technical argument for this other than what I see as progress!
And looking at a similar core spec to the suggested i9-9900K, I am leaning more to the i9-9900X, although it is twice the price.
And while not wishing to start a flame war, does anyone else get a little miffed when checking mobo specs that you have to wade through 3 pages of data on addressable RGB LEDs?
The 9900X isn’t a wise choice for a system oriented towards gpu rendering, imho. But, it’s your money, you can do whatever you want with them.
And don’t be mislead. There is no such thing as “newer cpu format” with s2066. LGA 2066 is already 2 years old. The 99xxX series is Skylake X at 14nm. Nothing new here compared to 79xxXs, just a better binning for the chips and slightly higher clocks.
If you really wanted something more progressive, then wait for the 7nm AMD Zen 2 processors coming later this year.
My last piece of advice would be this one: A 9800X would be a much more rational pick, if you have to go to Intel’s HEDT and need the extras it brings with it. Since you’re rendering with the gpu, there is no need for a 1000$ 10c/20t cpu. The 9800X costs much less, has a 4.5GHz turbo clock, and you could also try to OC it to a 4.6-4.7 for 24/7 in order to extract some some additional STP juice out of it. You still have the 44 pcie lanes and headroom for more devices with this platform, but you can shave off some money and be able to add a 2nd gpu when you decide to.
)
)