FAQ  •  Register  •  Login

5820k vs 6700k for better multiboxing performance?

Moderator: MiRai

<<

underfl0w

Posts: 5

Joined: Sun Dec 06, 2015 9:17 pm

Post Sun Dec 06, 2015 9:25 pm

5820k vs 6700k for better multiboxing performance?

I'm planning on building a new rig, not specifically for multiboxing but when I game I am generally multiboxing.

My current CPU is an i7 920, the first generation. It's served me well but has been performing worse and worse, and I can't do much of any multitasking outside of boxing, so I figured it's a good time to build a new system.

When I'm running a full group (6 clients) in EQ2, my CPU core utilization fluctuates between 95-100%.

My thought process is that the 6 core 5820k would suit me better since I can assign a client to its own core, but then I thought that if maybe the Skylake platform is an overall better performer and with a little higher clock frequency that it might also work even with only 4 cores.

The rest of the system would roughly be the same, just the CPU and motherboard would vary.

Any suggestions?

Thanks!
<<

MiRai

User avatar

Vibrant Videographer

Posts: 3010

Joined: Fri Nov 20, 2009 3:30 pm

Post Sun Dec 06, 2015 10:14 pm

Re: 5820k vs 6700k for better multiboxing performance?

I don't have a concrete answer, but I'd probably go for more cores when it comes to multiboxing/multi-tasking. I would imagine that having more cores would allow the OS to balance things a bit better.
<<

underfl0w

Posts: 5

Joined: Sun Dec 06, 2015 9:17 pm

Post Mon Dec 07, 2015 12:52 am

Re: 5820k vs 6700k for better multiboxing performance?

Thanks, that's what I am figuring, the more cores the better. I figured I'd throw it out there and see if anyone with familiarity with both platforms could shed some insight into maybe some other aspects of them that would differ performance wise.
<<

bob

User avatar

League of Extraordinary Multiboxers

Posts: 4592

Joined: Sat Feb 15, 2014 11:14 am

Location: In the dining room, with the lead pipe.

Post Mon Dec 07, 2015 8:25 am

Re: 5820k vs 6700k for better multiboxing performance?

http://ark.intel.com/compare/82932,88195

It's really going to come down to whether you need extra cores, or the extra GHz. A rule of thumb (which is sometimes wrong), is that current games will be happy on 2 cores and a 3GHz plus processor. So for multiboxing, more cores is generally better.

There are trade offs though. The main differences that stand out are:
  • Single core GHz
  • PCIE links
  • Memory bandwidth
  • Onboard graphics (which actually depends on whether you use if for offloading encoding)
  • Bus Type - although this will matter little to you, the change in Bus Type is an evolution that Intel are pushing to get ready for the 3D XPoint memory, which is not expecting to be ready for a while.

Do they matter for multiboxing..... probably not. Of all of them the PCIE links if you happen to be running 3 way SLI. Other than that, I'd get a decent cooling system, overclock the 5820K a little, and probably be very happy.

Of the rest, extra memory bandwidth is great if you do video encoding as it will help speed it up. Single core GHz is great for anything single threaded, or even for power savings, as your computer can shut down other cores that aren't busy (not useful while multiboxing).
<<

underfl0w

Posts: 5

Joined: Sun Dec 06, 2015 9:17 pm

Post Mon Dec 07, 2015 1:56 pm

Re: 5820k vs 6700k for better multiboxing performance?

The trade offs are mostly what I get hung up on.

From what I've read about the CPUs, the 5820k will see better overclocking gains, and can close the gap more to what the 6700k can achieve.

Since I won't be using SLI, only having 28 PCI lanes doesn't really concern me too much right now, but then in the future that might limit what I could do with the system.

Unfortunately it seems that most of the X99 boards are missing out on the new Turbo M.2 (x4) option. I'm not sure how much of a gain one would see with a x4 M.2 versus a x2 M.2 while multiboxing though.

Since both support DDR4 at roughly the same frequencies, will the new DMI 3 bus actually see the benefits of its double link speed from DMI 2 in X99 systems? (Either of them seem like a step backward compared to QPI's theoretical bandwidth capabilities though.)

Given that my current system is pushing eight years old, anything will give me better performance at this point, and either will surely last me a long time as well. I'll probably sweat the minutiae until Skylake-E is released in 8 months and then have a whole new set of trade offs to worry about, haha.
<<

MiRai

User avatar

Vibrant Videographer

Posts: 3010

Joined: Fri Nov 20, 2009 3:30 pm

Post Mon Dec 07, 2015 4:14 pm

Re: 5820k vs 6700k for better multiboxing performance?

underfl0w wrote:Since I won't be using SLI, only having 28 PCI lanes doesn't really concern me too much right now, but then in the future that might limit what I could do with the system.

It's only an issue if you're expecting 16 lanes out of each PCIe slot because 3-way SLI would eat up 24 PCIe lanes (8x/8x/8x), while 2-way would only eat up 16 (8x/8x).

EDIT: Unless there's a restriction on the CPU in general when it comes to PCIe lane allocation. I'm not entirely sure about that myself.

underfl0w wrote:Unfortunately it seems that most of the X99 boards are missing out on the new Turbo M.2 (x4) option. I'm not sure how much of a gain one would see with a x4 M.2 versus a x2 M.2 while multiboxing though.

Personally, I'd probably not game off of the M.2 drive since they still have some throttling issues when they heat up, and while my new build does have an M.2 drive, I'll be using it strictly for the OS and gaming off of a standard SSD.

underfl0w wrote:Since both support DDR4 at roughly the same frequencies, will the new DMI 3 bus actually see the benefits of its double link speed from DMI 2 in X99 systems? (Either of them seem like a step backward compared to QPI's theoretical bandwidth capabilities though.)

I have no idea.

underfl0w wrote:Given that my current system is pushing eight years old, anything will give me better performance at this point, and either will surely last me a long time as well. I'll probably sweat the minutiae until Skylake-E is released in 8 months and then have a whole new set of trade offs to worry about, haha.

Broadwell-E isn't even out yet, but here are the leaked specs. So, Skylake-E probably won't be out for at least a year after that, which would mean you'd be looking at Q2-Q4 2017. You could always wait a few months for BW-E to get a cheaper 6-core, or perhaps even a more-affordable 8-core, but I'll admit that I'm a bit salty that we're going to see a 10-core after I just spent the big bucks on the top-end 8-core. >_>
<<

bob

User avatar

League of Extraordinary Multiboxers

Posts: 4592

Joined: Sat Feb 15, 2014 11:14 am

Location: In the dining room, with the lead pipe.

Post Mon Dec 07, 2015 5:01 pm

Re: 5820k vs 6700k for better multiboxing performance?

underfl0w wrote:Since both support DDR4 at roughly the same frequencies, will the new DMI 3 bus actually see the benefits of its double link speed from DMI 2 in X99 systems? (Either of them seem like a step backward compared to QPI's theoretical bandwidth capabilities though.)

For anything that is using the DMI interface, probably. Is your memory controller using it (the DMI) to talk to the RAM. No. The memory controller communicates directly to the RAM at the frequency set, via DDR4 signalling.

From a Sandy Bridge onwards, consumer processors use derivative of the QPI for the internal interconnect ring, although the data bus was only 32bits wide, rather than the 64bits of the QPI (not sure if this has increased in newer architectures).
QPI is only listed on the 5820K because it is part of the Xeon family so it is using QPI as the internal interconnect ring, rather than the unnamed interconnect that is used in consumer processors. The interconnect is presumably publicly unnamed, because it is not externally exposed. There is no external exposure of the QPI on a 5820K either; it is used internally.
The speed of the interconnect does not automatically affect your memory bandwidth, which is limited by the memory controller, it's clock, and your RAM. It does affect how fast the memory controller communicates with other participants on the interconnect.

DMI3 will definitely matter when communicating with the South Bridge (which is done via the memory controller). A 5820K has DMI2. Would you actually notice a difference between DMI2 and DMI3. Only if you managed to flood the South Bridge. Maybe a SSD RAID would do it..

edit: clarified a couple of specifics.
<<

underfl0w

Posts: 5

Joined: Sun Dec 06, 2015 9:17 pm

Post Mon Dec 07, 2015 6:13 pm

Re: 5820k vs 6700k for better multiboxing performance?

MiRai wrote:
underfl0w wrote:Unfortunately it seems that most of the X99 boards are missing out on the new Turbo M.2 (x4) option. I'm not sure how much of a gain one would see with a x4 M.2 versus a x2 M.2 while multiboxing though.

Personally, I'd probably not game off of the M.2 drive since they still have some throttling issues when they heat up, and while my new build does have an M.2 drive, I'll be using it strictly for the OS and gaming off of a standard SSD.


I'll probably end up doing this as well, I was looking at the new Samsung 950 Pro M.2 as an OS/apps drive and their 850 Pro SSD as a gaming drive.

MiRai wrote:
underfl0w wrote:Given that my current system is pushing eight years old, anything will give me better performance at this point, and either will surely last me a long time as well. I'll probably sweat the minutiae until Skylake-E is released in 8 months and then have a whole new set of trade offs to worry about, haha.

Broadwell-E isn't even out yet, but here are the leaked specs. So, Skylake-E probably won't be out for at least a year after that, which would mean you'd be looking at Q2-Q4 2017. You could always wait a few months for BW-E to get a cheaper 6-core, or perhaps even a more-affordable 8-core, but I'll admit that I'm a bit salty that we're going to see a 10-core after I just spent the big bucks on the top-end 8-core. >_>


I completely forgot about Broadwell. It seemed like it had such a lackluster launch and marketing campaign. At least I didn't see it being promoted as much as when Haswell launched or the new Skylake launch.

If Broadwell-E will use the same socket as Haswell-E, I wonder if the manufacturers will refresh their motherboard stocks with some new features or it'll just be the same boards as they have now. It seems like there's not a whole lot of options for the X99 boards for Haswell-E.

Thanks for the insight, I'll take a look deeper into Broadwell-E maybe.
<<

underfl0w

Posts: 5

Joined: Sun Dec 06, 2015 9:17 pm

Post Mon Dec 07, 2015 6:27 pm

Re: 5820k vs 6700k for better multiboxing performance?

bob wrote:
underfl0w wrote:Since both support DDR4 at roughly the same frequencies, will the new DMI 3 bus actually see the benefits of its double link speed from DMI 2 in X99 systems? (Either of them seem like a step backward compared to QPI's theoretical bandwidth capabilities though.)

For anything that is using the DMI interface, probably. Is your memory controller using it (the DMI) to talk to the RAM. No. The memory controller communicates directly to the RAM at the frequency set, via DDR4 signalling.

From a Sandy Bridge onwards, consumer processors use derivative of the QPI for the internal interconnect ring, although the data bus was only 32bits wide, rather than the 64bits of the QPI (not sure if this has increased in newer architectures).
QPI is only listed on the 5820K because it is part of the Xeon family so it is using QPI as the internal interconnect ring, rather than the unnamed interconnect that is used in consumer processors. The interconnect is presumably publicly unnamed, because it is not externally exposed. There is no external exposure of the QPI on a 5820K either; it is used internally.
The speed of the interconnect does not automatically affect your memory bandwidth, which is limited by the memory controller, it's clock, and your RAM. It does affect how fast the memory controller communicates with other participants on the interconnect.

DMI3 will definitely matter when communicating with the South Bridge (which is done via the memory controller). A 5820K has DMI2. Would you actually notice a difference between DMI2 and DMI3. Only if you managed to flood the South Bridge. Maybe a SSD RAID would do it..

edit: clarified a couple of specifics.


Thanks for clarifying that, I completely forgot that the memory controller is on the CPU now.

I doubt I'll be doing any SSD RAID configuration, and there will probably be very little going on overall. A single video card (980), an M.2 and SATA SSD, and an optical drive.

Thank you very much for your feedback!

Return to Multiboxing Hardware

Who is online

Users browsing this forum: No registered users and 0 guests

cron