[SOLVED] 2011v3 – is it still okay for 2019?

[indigo.root]

Hey, there. How do you guys think, is it okay to continue to use 5820k with x99 system or should I ditch it for something else?

Options:
1. Long live 5820k (overclocked, 4150mhz, the first two cores boost to 4450, 1.2v)
2. Sell CPU to get that top Broadwell-E
3. Switch the socket to something modern and energy effective – at least, it's still possible to sell it for good money.

What would you do? I don't play games a lot, bought it for video editing.

 

[karenjoly]

Look at the cost of updating the cpu :

Core i7-6950X is $1,579.15 US at amazon

i7-6850K is $ 755 US.

The cost of upgrading the system:

The i7-9700K is $ 400 US but you'll need a new board , cooler and memory.

I'd go for the 6850K. 6c/ 12t. I have one and it performs very competitively. Also, a better disk subsystem and graphics card will make some performance improvement for video editing aside from cpu .

 

[indigo.root]

Thanks for your answer, but why 6850? It's roughly the same CPU that I have, a bit outdated too, the same number of cores, slightly faster maybe. It would require some time to resell my CPU and get the new one (even in used condition 6850 would cost twice more).

I was thinking about 6950x at least because it has ≈ same multicore speed as 9900k without changing my motherboard. However, looks like 6950 used costs pretty much the same as 9900k + new mobo (without quad-channel and 128gb of upper limit).

So, should I wait for a new generation (10nm?) or ditch it already?

 

[The Original Ralph]

i upgraded from a i7-5960X (8C/16T) that i had OC'd to 4.3 GHz on an X99 motherboard. I upgraded to the system in my sig below

my main concern, the 5960X, iirc, had 40 PCI lanes, the 9900K has 16 lanes and wasn't certain how that would play out in rendering video files, ie time to render. But i had seen the reduction in render times when i went from 4.0 to 4.3 on the 5960x.

I didn't sell my X99 board or the CPU until i'd seen how the 9900K would run and if i saw any benefit from the even higher clock speed. I could not have been more impressed - reduction in render times ranging 30 to 50%, depending on the codec of the file i'm rendering. I've got the 9900K right now running at 4.9 on all cores, and because of the AVX offset, 4.6 whenever the cpu encounters AVX workloads which is about 80% of the time, depending on the software..

As to PCI lanes, i run three drives active, all NVMe, one for OS, one as a video "worktable" to read a file from, and the 3rd as another worktable to write the file to.

for what it's worth, i paid $560 for the cpu, $268 for the mobo (MSI MEG Z390 Ace, and i also bought new ram which i assumed i'd have to - i could have used the same ram from my X99 board.

After selling the X99 mobo, 5960x my cost to upgrade was slightly over $500.

But wait, i forgot, i'm building a new rig, new case, with custom water cooling (2 radiators) - not counting the case, my custom loop so far has run $780, but the custom loop is only cause i want that 5.0 or 5.1 OC. At 4.9 I'm seeing decent temps with a Noctua D15S, when rendering and cpu showing 97-100% load, high 60Cs to mid 70Cs

If you decide to go with a Z390 mobo, i'd research, especially on this forum, the different brands and the issues they're having with BIOS - they seem to have missed something on the Z390 chipset. I went thru two Gigabyte Z390 boards and an Asrock Taichi Z390 that fried itself and my CPU on a BIOS update

hope that helps some

almost forgot not a biggie but my Noctua D15S moved right over to the 1151 socket

 

[indigo.root]

i upgraded from a i7-5960X (8C/16T) that i had OC'd to 4.3 GHz on an X99 motherboard. I upgraded to the system in my sig below

…

thanks for sharing your experience. could you tell me, how do you run 3 NVMe drives with videocard using only 16 lanes? that drives are SATA or M.2? I'm asking because my current board can utilize PCI lanes for extra M.2 speed, so it consumes 4 lanes additionally. in other words, your videocard eats all the lanes and the drives consume nothing, right?

it's great that there's that noticeable performance gain, but do you think it's time to upgrade for me? my concerns is that it would be harder to sell my mobo+CPU later. btw, how do you think what is the good price for this combo now? mobo is MSI M-Power x99s.

 

[The Original Ralph]

the programs i use don't use the GPU - they;re pushing their workload on the CPU
BUt like i explained above, the OS drive is running the software, and it's reading a video file on one NVME drive, while the file is being written to another NVMe drive. Edit: one program i use does use the GPU but when i'm running it, i'm only using the OS drive and one of the NVMe drives

The Z390 CHipset is also supposed to have, iirc, 24 lanes of it's own - i'm just not literate in chipset lanes and when each come into play. Keep meaning to research the subject but so much to research, and so little time

maybe someone literate in chipset / cpu architecture can step in

 

[karenjoly]

Re lanes, see attached. Intel now calls the quantity " platform lanes". The 9900/ Z390 has 40. Most modern intel chipsets have 24, so the difference is the quantity provided by the processor, in this case 16, so 40 total.

The beauty of the x299 is that the platform total may get to 68, that's 44 from the latest x processors.

The first pcie expansion slot (x16) connects from the cpu. Other lanes will be supported from remaining cpu lanes or through the chipset. Users mostly covet cpu lanes, as if one could tell the difference.

https://www.businesswire.com/news/home/20181008005597/en/Intel-Announces-World%E2%80%99s-Gaming-Processor-New-9th

 

[The Original Ralph]

if i'm reading your response correctly, there's no read advantage to cpu lanes vs chipset lanes?

 

[karenjoly]

Below is a brief description of the pch/cpu bus architecture. Most of this arrangement is well known. But the interesting question can be phrased thusly :

As we know, there are cpu and pch side devices connected by pcie lanes. The chipset devices communicate to the cpu through DMI.Version 3 is now used. If one has, like you have, several M.2 cards and perhaps two gpus connected by chipset lanes, how does all that bandwidth go to the cpu when DMI bandwidth is so much narrower ? You might even conclude that cpu lanes are a priority.

But this article points out :

If you have several chip-side devices communicating to your CPU bus, this may be your bottleneck in your PC.When all PCIe bus lanes are being used on a PC, your PC negotiates which device gets how much PCIe bandwidth.


https://silentpc.com/articles/performance-and-pci-express-bus-lanes

 

[The Original Ralph]

https://ark.intel.com/products/133293/Intel-Z390-Chipset

interesting - thanks for posting that

BTW - i went to intel's side, their indication is Z390 supports a total of 24 lanes, so i assume it's a case of what the motherboard is supplying in addition to the CPU, or does that count the 16 from the CPU?

link to intel's Z390 spec description https://ark.intel.com/products/133293/Intel-Z390-Chipset

 

[karenjoly]

In the Business Wire article, the graph shows the 9900/ Z390 "platform" has 40 lanes. Most modern intel chipsets like the Z390/370, provide 24. The remainder is the quantity provided by the processor, in the case of the 9900/9700 etc.,, 16.

The Z390 provides 24, the cpu provides 16, the platform lane total therefore is 40.

The x299 platform has 68 with the newer x cpus.