Memory bandwidth with Dual Channel on Skylake versus Quad Channel with an 5930K

Hi skitszo.

Thanks very much for your reply.

The computer will be used for 'professional use'. I have close to $40k worth of legitimate software on there with about another $5k a year in annual charges. Only mention that to indicate that spending a bit more isn't the end of the world (DDR3 versus DDR4).

Software primarily is Autodesk Building Design Suite Premium. Within that suite I primarily use AutoCAD, Navisworks & Revit. The other main program I use is called Autosprink - which is what I use for 80% of my work.

While there are features with Building Suite that use multiple cores, unfortunately a lot (most) of the features that I use are limited to single core (not high end rendering). Same goes for Autosprink. Building Design Suite is incorporating more multiple core support with each release (every year), however Autosprink is less so.
So, single core performance is really important to me. There are the thousands of moves I can make in a day that are annoying being 3 seconds. Super annoying as back in the day when everything was done in 2D I could make similar moves instantly. Also have situations where I have to do an import or such that can take 20 minutes or so.

When I’m doing an import or such when I look at a desktop ‘Process Monitor’ gadget (kind of like resource monitor); it shows the CPU at 12.5% (4 cores with hyperthreading – shows as 8 cores, so the program is maxing out only one of those cores).
What I’m wondering is if the CPU, though showing at 12.5%, is being limited by the memory bandwidth when doing those imports.

On the project I’m currently on (trying to finish up), I have had at times all my 32GB of RAM used as well as using another 16GB page file.
However, it seemed to put the Navis file into page file and I didn’t find the performance difference to be that bad for that program compared to when I had less programs open and the Navis was in RAM only (rather than page file).

I want my cake and eat it too… I want the maximum clock speed for single threaded but also access to memory.
I know all pre-NDA reviews, etc. have to be taken with a large grain of salt, however there seems to be some consistency in the rumours of late that the Skylake processors are anticipated to be very much overclockable. If the rumours of 5.5Ghz being stable overclock turn out to be true, it would be a bit of a bummer to be shelling out for an expensive X99 system knowing that for maybe 80% of my work, someone (my competition) are able to out perform my system by maybe 20% even though they maybe spent considerably less (i.e. by going with an overclocked Skylake).

I have been a bit concerned even with the idea of the Skylake being limited to 4 RAM slots for a total maximum of 64GB (4*16GB), since I have already seen 48GB being used (between RAM & Page file), and expect the next project to be even more demanding.
However, since it wouldn’t be all the time that I would have all those files open at the same time; that I would be able to go with the Skylake to get the maximum clock speed along with an Intel 750SSD, so if I do run out of RAM, that I can have fast page file.
I don’t ask for much. All I want is a 5.5Ghz Skylake with 128GB Quad DDR4 (2017 maybe lol).
Any thoughts would be appreciated!!

Thank you both very much for your replies.

I think you're probably right Tradesman1, in that the X99 is a safer bet. And as you mentioned, if I go that route I would go with the 128GB option.

The Skylake consideration was being understood to likely last considerably less time; anticipating the RAM to be the item driving next upgrade (usually is for me). However I was also considering that (a) it would also cost a bit less and (b) if the higher clock speed would speed up work, then was looking at shorter lifespan as being the cost of getting the higher clocks. However if the memory bandwidth was being bottlenecked then the higher clocks aren’t as attractive.


Thanks for the suggestions skitszo.

I don’t think the dual CPU wouldn’t really be of benefit to me. Those programs I mentioned do have capabilities to do ‘realistic rendering’, which does take full use of multi cores (those cores being on a single CPU or Dual CPU). However those are more for presentations, etc.

For my end, I am designing mechanical services (finding places for pipes, etc. to go that meet codes and don’t clash with other services…). While doing this the graphically representation of the objects are more ‘blocky’ and shadows, etc. are likewise far more basic.

Where I do most of my work (calculations, etc.) though (Autosprink), it is mostly purely single threaded.

It seems the X99 i7 boards have 8 RAM slots, so could do that without going to a dual cpu board.

As you mentioned, I am figuring on going with a PCIe SSD (Intel 750).

If going with the X99 system I would have liked to go with an overclocked Xeon (E5-1660v3), so that I could populate with 4*32GB Ram, allowing 4 remaining slots for future use.
Not as crazy as it first sounds but it doesn’t seem 'that long ago' that I swapped 128MB instead of 64MB into a laptop at work (I think it was about 1999-2000), and it was like a new computer So maybe in 2030 I’ll be looking to upgrade to 128,000GB RAM (JK).


However I’m not comfortable putting the computer together myself and given the price the machine would be costing I would want it put together by a reputable company. The ones I have seen don’t seem to offer overclocking on the Xeons; even though I have seen users reporting doing so on several forums. I contacted Intel support in this regards as I hadn’t seen anything official supporting overclocking with unlocked Xeons. I was told “… After confirming this once again, I can only repeat my previous statement that XEON CPU’s are not unlocked. In General Intel provides very limited support on overclocking but does not provide any advice or assistance regarding overclocking. That does not mean that overclocking is not possible, however it means that we are unable to provide any advice or assistance.”
I think the X99 i7 5930K overclocked with 128GB RAM might be the balance between the Skylake i7 and an overclocked Xeon.

Thanks again for all your input!