6700k vs. 5820k

WarrenX

In Runtime
Messages
199
I am building a brand new PC and finally switching from a life long passion for AMD to see what all the fuss is about with Intel. So I find myself in a deadlock between these two processors. My primary function will be gaming with a few background processes going on (20 tab chrome, music, etc.).

i7 6700K
On the one hand, the Skylake seems to be the intuitive choice as it is the cutting edge generation with the z170 chipset, higher IPC, and greater OC headroom. My reasons for being hesitant to get this CPU: feels like I'm taking a step back coming from a fx6350 in terms of core count, many of the benches I see seem to give the 5820k a healthy lead in full CPU utilization (and some in real world fps), and the APU in the chip may be under-utilized in potential synergy with dedicated GPU in DX12.

i7 5820K
On the other hand, the Haswell-E seems to me to be the wiser choice with its additional serial/logical cores, same features (hyper-threading), competitive OC range (reading people getting 4.5 stable on first prime run), and possible future proofing (increased multi-core support). My reasons against the 5820k: I feel like committing to the x99 would lock me into an older esoteric architecture with a higher overhead (quad channel ddr4, PCIE channels), the single core bandwidth (I am looking to play WoW and other MMORPGS), and I am not really a GPU enthusiast so I will most likely stick with a single r9 290x.

I sort of want to pop my Intel cherry the right way by getting the 6700k, but if I can overcome superficial novelty maybe the 5820k is a gem in the rough. Please tell me to get the skylake :(
 
Hey mate. So I'm kinda of in the same boat. I have a AMD 955 BE and it's just not getting me enough juice for modern games @ a solid frame rate.

So you say 6700k vs 5820k?
I think it's more of a 6700k vs 4790k.

5820 is supposed to be the gamer CPU of the old architecture. However, I find that most games still use only 1-4 cores at the most. 5820k is simply tailored for something different.

However as I am looking at the 6700k vs the 4790k, I lean towards the 4790k. This is mostly because its the top of the line of the "older" generation. It's allows you to use the DDR3 ram you have on your AMD build DDR3 pin 240, and it's still rated as one of the best cpu's out there. The 6700k is good, but I still think 4790k is better.

P.S. I picked up a 6700k, 6600k, 4790k, 4690k within last week and I tested all 4 to see which one would perform the best for the buck.

If money is no concern.
4790k > 6700k > 4690k > 6600K

If money is of concern.
4690k > 4790k > 6600k > 6700K

The new chips are just not worth the money in my opinion.
AT BEST, you get a 5-10% boost. They will run a few watts lower. Create a bit less heat. But why? There is nothing a 6700k can do that a 4790k can't... or a 5820k for that matter.

And by the time intel releases their new CPU's in Q3 of 2016, the socket will change because the 10nm chips will deliver significant speed increases over the old 14nm counterparts.

Just my 2 cents.
 
Hey Ross. Appreciate your input mate :) . I guess with regard to my comparison my reasoning is: do I want to get the latest and reasonably greatest Single Core bandwidth for gaming which has the most capacity speed, core clock weight, and thus IPC, or do I want to get an all-around rock solid foundation CPU like the 5820. So I would say in my mind it is like a scalpel vs. a multi tool.

With regards to the 4790k, don't get me wrong, logic and reason scream overlooking the generation and seeing it for its purely arcane numbers... but I just cant. Viscerally I just dont think I can bring myself to "settle" for a 4xxx on my first real intel build. Just doesn't sit well with the impetus of this build which is sort of a "turning the corner" philosophy.

My rationalization for disregarding the purchasing power parity blunder here is as follows: skylake could have suprise potential in store in the future (dx12 apu integration), the z170s are weathered for the future with skylake-e (so I can just hock the 6700 if I am not satisfied), and I feel like if I get into LGA 1150 I am just going to be chained to a sinking ship with respect to future bottlenecks (sata, pci-e 3.1/4, ddr4 quad channel). I totally get the 2-3 year futureproof rationale, but my thinking is if this build is going to be implemented in say a month, I have already missed the window to fully-utilize objective future proofing for even the conservative range of three years (which coincidentally is sort of the meta problem version of these core bandwidths).
 
Last edited:
Here's the deal guys, X99 is for PCI-E lanes. WIll you use more than 2 GPUs at a time? Will you be doing anything that requires 6 or more cores? If you answered yes to these questions then X99 is for you. Just know that Broadwell-E is right around the corner and will have 4 SKUs making the 8 core cheaper.

Skylake vs Haswell is what this comes down to. Here's the bullet points
- DDR4 makes no performance difference in almost any consumer application
- Skylake IPC is barely 5% over Haswell, you only see FPS difference due to higher stock clocks on Skylake
- If you don't require higher memory capacities that DDR4 offers then you might as well hop on the DDR3 train
- Skylake PCH (chipset) comes with newer features. If you don't need more than 1 M.2 slot, USB3 3.1, or USB Type C then you don't need the newer platform

(dx12 apu integration)
You won't be using the IGP with DX12, namely because it won't help. The only thing the IGP will be good for is QuickSync if you stream but if you have an Nvidia GPU you won't need it. Since I'm assuming you got the 290x then you'll need to stream using the CPU which can be costly. If you don't stream disregard this. DX12 in itself when is first rolling out will only offer lower overhead and higher draw call counts. Most of the advanced features won't be utilized for another few years.

Another thing I'm seeing is a lot of i7s being listed off for what appears to be no good reason. i7s are expensive, especially the 6700k. If you aren't utilizing HT it makes 0 sense to buy the 6700k over the 6600k. The FX 6350 is more like a tri-core, not a hex (AMD is being sued for false advertising this). In any regard to core count the upgrade will absolutely stomp anything AMD has on the market right now anyways, so whatever you decide to get will be an upgrade.

Last but not least, gents I'm running a 4 year old SB-E processor that is within a few FPS difference of the latest and greatest today that's 4 generations ahead. My best friend's 5960x actually scored 2k lower physics score in Firestrike than my processor and I have 0 issues running 3 of the best GPUs on the market. The reality here is no matter what platform you get you won't see a real performance difference and if overclocking is into play each generation will grant you a margin of 0-5fps difference in each scenario. The only thing that something like a 5820k+ will accell in is in high CPU usage scenarios like Handbrake, video encoding, rendering, ect. I have a Xeon 1650v3 in my work machine that games exactly the same as my 3960x at home with 32GB of DDR4 in it. It performs stock to stock equally the same in Premier Pro and Photoshop as my 3960x.

So bottom line here is, buy what you want and be happy with your purchase. If you're coming from AMD it's a big upgrade. You won't see much difference in variance to architectures and performance so if you need cores go with cores, if not go with the latest.

And by the time intel releases their new CPU's in Q3 of 2016, the socket will change because the 10nm chips will deliver significant speed increases over the old 14nm counterparts.
The Tock is usually always on the same platform IE SB to IB, Haswell to Broadwell. Kabylake on the other hand will continue to be on 14nm because Canonlake (10nm) was delayed until 2017 meaning Kabylake will be on the current platform. Kabylake is another tock for 14nm.
 
Last edited:
On another note, The Skylake architecture does support 20 pcie lanes. That really doesn't mean much unless you decide to upgrade at some point to use a M.2 hard drive. Also the Kaby Lake and above cpu's might only be able to run on Windows 10 pc's from what I'm seeing on the report from Engadget:
Microsoft won't support old Windows versions on new processors
This could be a future issue for anyone planing to use Windows 7
 
Thank you PP Mguire for your input. I understand your point about the i5, but I am confused about the "threading" state of games and how well (if at all) games can be benefited from HT (either directly or indirectly). From what I have been reading It all seems really ambiguous about what a "heavy workload" is and what scenarios would qualify as ideal for intelligent process scheduling. I don't understand enough about how intricate processing works to see what extent parallelization would make a difference in real world daily driving. It's 6:00AM im going to bed, have a good night everyone see you tomorrow! :p
 
On another note, The Skylake architecture does support 20 pcie lanes. That really doesn't mean much unless you decide to upgrade at some point to use a M.2 hard drive. Also the Kaby Lake and above cpu's might only be able to run on Windows 10 pc's from what I'm seeing on the report from Engadget:
Microsoft won't support old Windows versions on new processors
This could be a future issue for anyone planing to use Windows 7
That's only for security and support. Windows will still physically run on these processors.

Thank you PP Mguire for your input. I understand your point about the i5, but I am confused about the "threading" state of games and how well (if at all) games can be benefited from HT (either directly or indirectly). From what I have been reading It all seems really ambiguous about what a "heavy workload" is and what scenarios would qualify as ideal for intelligent process scheduling. I don't understand enough about how intricate processing works to see what extent parallelization would make a difference in real world daily driving. It's 6:00AM im going to bed, have a good night everyone see you tomorrow! :p
Ok HT is a virtual thread that utilizes unused resources of a core to process more information. To put it simple, games don't use this and won't. You might see a game using virtual threads but that's only because some games are programmed to utilize as many logical processors that it has available. This can actually sometimes hamper performance depending on the load that is assigned to one given virtual thread.

When people say "heavy workload" it generally means CPU intensive tasks. Video editing, CPU side streaming, transcoding (Handbrake), rendering, CAD, 3D work, and all those qualify as tasks that benefit from HT as well.

Another way to look at gaming and HT is think of the GPGPU side or the parallel side of a GPU. Do games utilize this extra section in GPUs? No, unless you count PhysX. It's only until recently where a tech called Async Compute has come around (generally called HSA from AMD previously) where an API can queue up tasks to be run in parallel for each serial side of a GPU. TL;DR a game can send commands to the compute and gaming side of a GPU. Don't expect something like this to be used for HT and processors though, as there isn't much benefit to it.

Using more cores might become beneficial after the mass adoption of DX12 as they are opening up this use for AI and other things. Anothing thing where HT won't become beneficial, as it just kinda sits there.
 
That's only for security and support. Windows will still physically run on these processors.
Wouldn't that depend on whether or not Intel, AMD and Qualcom decide to make legacy (windows 7) drivers for thier cpu's? I think M$ means not to support native drivers for new chips in the near future to support Windows 7/8/8.1
 
Last edited:
Wouldn't depend on whether or not Intel, AMD and Qualcom decide to make legacy (windows 7) drivers for thier cpu's? I think M$ means not to support native drivers for new chips in the near future to support Windows 7/8/8.1

Was going to reply - but I'll suggest you take it to the Win10 thread so this topic doesn't get derailed.
 
Back
Top Bottom