upgrading from a 2600k to 6700k?

[poofyhairguy]

I think there is a very real possibility that we may not seeing anything "significantly" better than a 2600K for 2 or 3 more years. I doubt IPC improvements get us to a 50% improvement over a 2600K in that timeframe. Yeah, extra cores CAN help, but will software take advantage of them?

I think the software is moving in the right direction for sure. Heck the only reason I even need the 2600k today is for gaming, and games are getting more multi-core compatible all the time (so much so that just moving from a 2500k to a 2600k was a noticeable upgrade last year). Honestly I have no need to even move past the 2600k today, which is why I wouldn't mind a strategy that spreads the CPU upgrade costs over a few years.

The question to me is this:

Will the increase in IPC or core count for future Intel consumer CPUs outpace the depreciation of Haswell/Broadwell Xeon CPUs?

In the old days the answer would be hell yes, but today we see eight core Sandy Xeons on ebay cheaper than Skylake i7s. As a gamer I have to assume the PS5 will have an eight core CPU that is at best equal to Ivy/Haswell IPC (at lower clocks than the chips we are talking about today), so I think it is safe to assume a six or eight core Haswell chip could easily be viable in a gaming machine until like 2022 or something crazy like that.

So more and more, it looks like IDC was probably correct - just buy a CPU now and enjoy it for the next 5 years. In 5 years, the top CPU at the time may only be 30% ahead of the CPU you buy today so you can rinse and repeat at that time.

So basically that boils down to "lower your expectations and settle." It sounds more like bad dating advice than technology enthusiast advice. I personally refuse to admit defeat, I can hear my old Celeron 300A calling at me from its place in history- "don't let them win, find a way to get the value they don't want you to have!"

Outside of used Xeons where I see value in the near future is when the tick-tick-tock strategy creates a glut in the middle tick that will have to be cleared off at great prices.

We saw that some with Haswell, people were waiting for DDR4 and Skylake so at the end of Haswell's life you could pick up a six core Haswell at Frys for cheaper than many people bought their 6700k once Skylake released. The only reason I didn't was the cost of X99 mobos (but I should have).

Skylake was obviously a success so there won't be a surplus of Skylake CPUs to clear out, but Kaby is starting to look like the most boring generation in Intel's recent history. If that is the case by the time Kaby is replaced places like Frys will still have a ton of them, and the savvy buyer would hold onto their current CPU to get to that point if they can to help Frys clear that inventory at negative margins for them. By then DDR4 RAM will be a lot cheaper too.

Hell, half the fun of this computer hobby for me is getting more value than companies like Intel intend for me to have. Without overclocking/unlocking cores I would have moved on to something else over a decade ago.

 

[BonzaiDuck]

I'm probably the one you saw mentioning the 30% gap. I believe benchmarks were showing 25-30% performance increase going between SB and SKL on average. There were some individual tests and applications where the gap was much larger IIRC - I'll have to review the benchmarks again to see what those specifics are.

So even if we assume the low number (25%), you're probably not going to find a 2600K which will clock high enough to outperform a 6700K with an average OC on equivalent forms of cooling. I think that is probably also generally true of SB vs. Haswell and even Broadwell. A 6800K clocked at 4.4 Ghz is going to be tough for even a stellar 2600K to outperform on average.

BTW, I'm similar to you based on your earlier post - got my first computer in 1981 as a kid and then started building my own PCs in 95, after Commodore croaked. I'm chomping at the bit to upgrade my 5 year-old rig, but I want to feel like I get good value doing it and it is hard to see value outside of the 6-core HEDTs to me personally. However, that assumes you can hit 4.4 Ghz on a 6800 or 6850 and that is a tricky proposition based on what I've seen so far.

. . . And that's why I waffle on my own plans between the mainstream 4-core and the HEDTs that don't come with delidding temptations. But either way, I'm probably going to pull the string in the next 5 months or so.

There are all the practical reasons, but I ultimately break down because of the fun of it.

 

[poofyhairguy]

So wait, the X Intels have soldered CPUs like in the good ole days?

 

[Arachnotronic]

So wait, the X Intels have soldered CPUs like in the good ole days?

Yes, HEDT is all solder.

 

[VirtualLarry]

Yes, HEDT is all solder.

Yep. 140W TDP (and higher, in some specific cases), will do that to you.

 

[poofyhairguy]

Yes, HEDT is all solder.

Wow that's awesome. I need to go check if my local Frys has any 5820k's left!

 

[MrGuru]

Well, I've almost completely gone back on wanting to upgrade now. Maybe the 5820 or 68xx would potentially be worth it, but yikes.

 

[IndyColtsFan]

Wow that's awesome. I need to go check if my local Frys has any 5820k's left!

If you have a Microcenter close by, the 5820K is $320 right now.

 

[IndyColtsFan]

Well, I've almost completely gone back on wanting to upgrade now. Maybe the 5820 or 68xx would potentially be worth it, but yikes.

I'm thinking I may just wait and see if I can catch a deal on a 5930k or 5960x at this stage. Hoping maybe someone will clear the 5930K out at $399 or so.

 

[RussianSensation]

Some of you guys are massively downplaying how crucial IPC and single core performance really is now, especially since by far the vast majority of games don't use more than 4 cores. As GPUs continue to get faster by approximately 32-33% per annum or 2X roughly every 3 years, the CPU bottlenecks are going to become even more noticeable.

2560x 1440
5960X @ 4.4Ghz with 1080 SLI = 82 FPS
5960X @ 4.4Ghz with 1080 = 80 FPS

4K
5960X @ 4.4Ghz with 1080 SLI = 72 FPS
5960X @ 4.4Ghz with 1080 = 44 FPS
http://www.guru3d.com/articles_pages/geforce_gtx_1080_2_way_sli_review,13.html

Fast forward to 2017, Big Pascal will be at least 30-35% faster than a 1080.

Skylake i7 6700K at even just 4.6Ghz will beat every single 4.9Ghz 5820K/4790K/6800K:
https://m.youtube.com/watch?v=f5lfMogcrPU

Since 6700K platform upgrade costs less than 5820K/6800K platform, that $ saved is better used towards a 1440p 144Hz/4K monitor to try and move as far away from CPU limited land as possible. Every extra 10% in IPC helps with top end GPUs.

At the same time, there is no need to overthink it. Get 1070/SLI or 1080/SLI on a 2600K OC and see what performance you actually get. If the FPS one gets is satisfactory (i.e., you feel gaming is smooth without looking at the FPS counter) just keep using the Sandy Bridge platform. However, if already decided that OK I will upgrade anyway, the extra cores today do almost nothing in 2016 games other than few examples such as Total War Warhammer.

If I had a 2600K @ 4.6-4.8Ghz, I would probably just keep running it and upgrade to Skylake-X. This way, you'd get the 6-8 cores + 1 true generation increase in IPC. The issue with buying a 5820K now is that CPU was not only $289.99 at MicroCenter last year, but the CPU itself came out almost 2 years ago. In that case one could have enjoyed it for the last 2 years and buying it now feels like "Why didn't I just buy it in August 2014?!"

For games, how can an argument be made that 6700K OC isn't a satisfactory upgrade but a 5820K/6800K OC is when the later are slower in 99% of games?

Another way to look at it: focus on what matters the most for the end experience.

Forget the hype of 6-10 core CPUs. Work backwards instead. 4K 120Hz > 1440p 144Hz > 5K 60Hz > 4K 60Hz > 1440p 60Hz. 1080 SLI > 1070 SLI > 1080 > 1070.

That means you'd have a better gaming experience with a 4K OLED 120Hz + 2600K @ 4.8Ghz + 1080 SLI than with a 6950X @ 4.4Ghz + 1440p 60Hz + 1070.

Just allocate the budget towards the monitor and GPU and you will actually feel the $ was spent wisely. Even in the FC Primal example I linked, even though 5960X 4.4Ghz completely bottlenecks the 1080 SLI at 1440p, if the monitor is only 60Hz, technically, you won't ever know.

DX12 could be the game changer 6-10 core CPUs, we won't know until more 2016 games use DX12. Chances are by the time DX12 games are more widespread SKL-X will be out.

Logically, we can even make the argument that 2016 is just a bad time to upgrade period.
- no HDR monitors
- no affordable 28-32" 4K 120Hz monitors
- overpriced mid-range Pascal masquerading as high-end
- no Skylake-X

If you waited 5 years to upgrade, why not wait 1 more year and build a Big Pascal, 1440p 144Hz/4K HDR, 6-8 core SK-X rig in 2017.

 

[IndyColtsFan]

RS, we get it. However, many of us don't care about a couple hundred here or there and want the extra cores. The 6700k is a great CPU but many of us just don't see the value proposition. $290 makes it a tough decision for sure though.

Also, the "wait until next year argument" means we would be waiting forever to upgrade. We all heard "Haswell will be a huge performance jump" followed by "Skylake will be a huge performance jump," neither of which turned out to be true. And now we're supposed to believe SKL-X, Cannonlake, or another platform will be the big leap we've been looking for? Sorry, not buying it - I suppose by law of averages, Intel is due for a home run but I'm not buying it until I see it. I had fully intended on waiting for SKL-X (and who knows, maybe I still will), but after seeing the BW-E debacle in terms of pricing, I'm not confident it will be worth the wait.

 

[poofyhairguy]

For games, how can an argument be made that 6700K OC isn't a satisfactory upgrade but a 5820K/6800K OC is when the later are slower in 99% of games?

Because if the 5820K/6800K OC is faster than the 6700k OC in games say four years from now (due to Directx 12 and better use of multithreading in games) you can go THAT much longer without an upgrade. It is pretty much the same argument for buying AMD GPU hardware instead of Nvidia really.

The secret is that very few of us need more than Sandy today, but some feel the upgrade itch for platform reasons (USB 3.1, M2 drives,etc). Something 30% better than Sandy today is an upgrade, but it isn't the kind of upgrade that will get you through the NEXT generation of consoles (aka PS5 and Xbox 2). An 8 core Haswell that theoretically is 100% faster than OC Sandy if all the cores are used might just find a way to be relevant that long. Just like Sandy was relevant for way longer than anyone expected.

The point isn't to keep pushing the envelope trying to go beyond what we are doing today (like 120hz) just to justify the 6700k because Intel refuses to give us better. The point is to make the experience we have today last as long as possible without having to stay on Intel's new low value tick tick tock treadmill. I personally think 8 cores of Haswell or Skylake can be relevant for gaming in say 2021 at 60hz, but the 6700k will be having you itching for an upgrade to the consumer 8 cores machine Intel is forced to sell at that point.

 

[lupi]

The egg just dropped the 6700k to 315. As one who really is in need to upgrade one of his machines, that seems a rather good deal short of living next to a microcenter.

 

[poofyhairguy]

Another way to put it:

In 2005 no one could build a single GPU PC that would play console ports at the resolution and hz of the 360 throughout its life. It wasn't until around 2008 or so you could build that PC.

But in December 2011 someone that combined a brand new 7970 with a 2600k built a machine that could play console ports from consoles released in 2013 at the same hz and resolution throughout the life of those 2013 consoles.

Given that trend, I am fairly confident a person will be able to build a single GPU PC in early 2017 that will play ports from 2018/2019 consoles throughout the life of those consoles. But that PC won't have a 6700k and a GTX 1080 in it- both those parts will be obsolete somewhere in the middle of the life of 2018/2019 consoles. But an eight real core CPU plus the best Vega AMD will sell? If history holds that is the system that will match the PS5 for console ports at the same resolution and refresh rate until the day it is replaced with a PS6.

 

[RobertR1]

If I had a 2600K @ 4.6-4.8Ghz, I would probably just keep running it and upgrade to Skylake-X. This way, you'd get the 6-8 cores + 1 true generation increase in IPC. The issue with buying a 5820K now is that CPU was not only $289.99 at MicroCenter last year, but the CPU itself came out almost 2 years ago. In that case one could have enjoyed it for the last 2 years and buying it now feels like "Why didn't I just buy it in August 2014?!"

That seems to be where I'm headed. I don't feel like my 2600k @ 4.8 is slowing down my oc'd 980ti.

I do plan to get the 1080ti so I'm hopeful it aligns closely with Skylake-X. My SSD, case and PSU won't need replacing.

 

[IndyColtsFan]

poofyhairguy is 100% correct. I'll be honest - I miss upgrading every 2 to 3 years, but I just can't justify it any longer when I'm not seeing at least a 50% performance improvement. Sure, there are platform improvements that I'd get today that might also improve my experience - NVMe being one of them. But we're talking about going with a strict 30% improvement with the same number of cores (6700K) versus a 20-25% improvement with additional cores (HEDT) which may push that improvement into the 50%+ range for apps in the future when the cores are taken advantage of.

 

[IEC]

Unless you have specific reasons to be needing more than 4C/8T *now* it's silly to be buying HEDT.

Especially since for gaming an OC'd 6700K beats... well, everything.

 

[sze5003]

I have a i5 2500k now. In the process of going to a i7 6700k. Micro center is a good 30 min drive from my place and maybe 15 if I go from my office.

I think since I've kept my 2500 so long the 6700 will be a good upgrade for me paired with some g skill 3200 16gb.

I wish newegg would price match micro center. It's currently $289 over there. I better get it soon before they run out but I am still selecting a new board and I would like to keep it all on one bill. Oh well, then I have to deal with getting a new windows license which will probably cost me around $100 or so.

 

[poofyhairguy]

Especially since for gaming an OC'd 6700K beats... well, everything.

At one point in 2014 the GTX 980 beat everything, it was the fastest consumer 1080p GPU in the world at the time. Today that same GPU is often falling behind the 2013 Hawaii GPUs at 1080p in modern games (especially Directx 12 ones).

Technology doesn't always age at the same rate and in the same way. The guy who told me his Athlon 64 3800+ was a better buy for gaming than my AMD 3800 X2 made the wrong decision in the long run. The person who bought a GTX 780 instead of a AMD R9 290 probably made the wrong decision in the long run.

I am not going to go so far to say that anyone buying a 6700k is making the "wrong decision," it's a monster chip. But if you can't see a 5820k beating it for gaming one day then you lack perspective of where we have been in this industry.

If you need the CPU difference today between a 2600k and a 6700k (say you have a 120hz screen) then really the 6700k is your only option. But for a lot of us who are pretty much upgrading out of old habit, the 5820k is still an upgrade over the 2600k once OCed and probably one day it will be significantly better than the 6700k in games as well. I can definitely see a reason to go in that direction, or to wait for a 6800k.

 

[swilli89]

Hi there.

I've had my trusty Sandy Bridge 2600k for going on 5 years now, and have almost pulled the trigger on upgrading to the latest Intel chip a few times, but ultimately, the many many reviews basically saying "if your gpu is fine, don't bother" has always stopped me.

The thing that caught my eye with Skylake however, is the pretty decent jump when using SLI that makes me really want to do a quick upgrade.
I'm currently running a previously mentioned 2600k @ 4.8ghz, 16gb 2133 mhz ddr3, 850 pro ssd, 2x 980 Ti @ roughly a 10% OC, and a Dell U3011 (1600p).

If I were to upgrade, I'd get a 6700k (+ an evo 212), MSI x99 SLI PLUS, and 32 gb of ~2400 ddr4 ram. Something like $700 for hopefully another 5+ years.
The new 950 pro ssds caught my eye too, but that's icing.

Thoughts?

You're getting solid IPC improvements but think about it, do you really want to get another 4-core CPU 5 years later? You've waited half a decade and now less than 6 months from now you have Zen coming out with 16 threads. If not that, then I would at least check out the 6 core Broadwell-E chips.

 

[Arachnotronic]

You're getting solid IPC improvements but think about it, do you really want to get another 4-core CPU 5 years later? You've waited half a decade and now less than 6 months from now you have Zen coming out with 16 threads. If not that, then I would at least check out the 6 core Broadwell-E chips.

Intel has 16 thread CPUs available today. 20 thread models, even.

 

[swilli89]

Intel has 16 thread CPUs available today. 20 thread models, even.

Dude how is that relevant? He's talking about buying a $350 chip and you mention the availability of a $1099 and $1499 chip.

We don't know the pricing of Zen's 16 thread chip and you can grab a 6-core Intel cpu (albeit Haswell-e) for about the same as the 6700k.

 

[poofyhairguy]

Intel has 16 thread CPUs available today. 20 thread models, even.

At the same price he bought a 2600k for? Where? I want one.

 

[Arachnotronic]

We don't know the pricing of Zen's 16 thread chip and you can grab a 6-core Intel cpu (albeit Haswell-e) for about the same as the 6700k.

Right, we don't know the pricing of the Zen chip, so why would you suggest it as an alternative to the 6700K?

 

[IndyColtsFan]

I can get a new 6700K for $299 and I'd be upgrading the system in my sig. I'm a bit frustrated with some USB issues I'm having and hope a new system would help that, along with giving me a decent boost. Thought honestly, I can work around the USB issues.

Thoughts?