2600K vs current Intel (OC/stock)

[Vayra86]

Very nice read, and provides confirmation that we're well past the 4 core era for gaming, but also the major benefits extracted from AVX(2). Adjust accordingly

Upgrading from an Intel Core i7-2600K: Testing Sandy Bridge in 2019

www.anandtech.com

Note:

 

[phill]

Still have one, still love it

Nice read though, thanks for the post

 

[XL-R8R]

Looking at the graphs and charts - and considering most "gaming" users do little else but game - I really don't see anything to complain about when it comes to the performance of this 8 year old (already!) chip.

If anything, this article actually shows us what an excellent investment that particular line of CPU's actually were.

 

[phill]

Looking at the graphs and charts - and considering most "gaming" users do little else but game - I really don't see anything to complain about when it comes to the performance of this 8 year old (already!) chip.
If anything, this article actually shows us what an excellent investment that particular line of CPU's actually were.

Agreed very much so If anyone just wishes to be above the 60 fps, these CPUs are still more than enough to pump out decent frame rates

 

[londiste]

With regards to gaming, 18% increase with 9700K or 14% with 7700K at 1440p is not something to disregard. Even more so considering this was tested with GTX1080 that runs fast out of steam at 1440p. 30%+ at 1080p is even more noteworthy difference but I think Anandtech tested 1080p with medium or low.

Yes, OCd 2600K brings the difference down by a lot but when comparing that also consider 7700k and 9700k are no slouches for overclocking either. According to Siliconlottery tests, most (75+%) of 7700K and 9700K OC to 5.0 GHz and practically all to 4.9GHz. When comparing results, these are the frequencies tested CPUs run at - normal/boost, most multithreaded loads including games are more likely running at normal clocks:

• 2600K - 3.4/3.8 GHz
• 7700K - 4.2/4.5 GHz
• 2600K OC - 4.7 GHz
• 9700K - 4.6/4.9 GHz

I really would have wanted to see the same suit of tests at the same clocks - preferably with memory at the same transfer rate too, as they had 2600K with DDR3-2400 that should not be too difficult.

 

[Rahnak]

It's a very interesting article indeed. I hadn't really had a problem with my 2500k (although I do keep away from newer, more demanding titles because of it) until I played Rise of the Tomb Raider recently. In one of the snowy, more open area of the map, it loaded all four cores to 100% and made me drop a good number of frames. Overclocking to 4.2 helped me get some of those back though.
But yeah, it really is the time to move on.

 

[FreedomEclipse]

Well its not like the OC'd 2600k did badly at all (even if it was consistently at the bottom of the pack) but obviously its no spring chicken anymore.

 

[londiste]

I hadn't really had a problem with my 2500k (although I do keep away from newer, more demanding titles because of it) until I played Rise of the Tomb Raider recently. In one of the snowy, more open area of the map, it loaded all four cores to 100% and made me drop a good number of frames. Overclocking to 4.2 helped me get some of those back though.
But yeah, it really is the time to move on.

4 cores/4 threads does get in the way in a lot of newer games. 4 cores/8 threads is OK for all but a handful of games (Battlefield MP, Assassin's Creed: Odyssey and some Far Cries come to mind). From my personal experience (after going from i7-6700K to i5-8400) 6 cores/6 threads is roughly equal to 4 cores/8 threads.

Future-proofing is anyone's guess. More cores and threads is always good but for games, single-thread performance still matters a lot.

 

[Vayra86]

With regards to gaming, 18% increase with 9700K or 14% with 7700K at 1440p is not something to disregard. Even more so considering this was tested with GTX1080 that runs fast out of steam at 1440p. 30%+ at 1080p is even more noteworthy difference but I think Anandtech tested 1080p with medium or low.

Yes, OCd 2600K brings the difference down by a lot but when comparing that also consider 7700k and 9700k are no slouches for overclocking either. According to Siliconlottery tests, most (75+%) of 7700K and 9700K OC to 5.0 GHz and practically all to 4.9GHz. When comparing results, these are the frequencies tested CPUs run at - normal/boost, most multithreaded loads including games are more likely running at normal clocks:

• 2600K - 3.4/3.8 GHz
• 7700K - 4.2/4.5 GHz
• 2600K OC - 4.7 GHz
• 9700K - 4.6/4.9 GHz

I really would have wanted to see the same suit of tests at the same clocks - preferably with memory at the same transfer rate too, as they had 2600K with DDR3-2400 that should not be too difficult.

Its actually quite a bit more if you think of it; that 18%/14% is with the 2600K @ 4.7 all core OC as the baseline. So in fact the 7700K already exceeds the 2600K despite a 500 mhz lower clock. It kind of echoes my own experience going from the 3570K to this 8700K. There is far more to gain than just a core/frequency advantage, there is a considerable IPC gap as well and the newer platform helps too. We're also talking PCIE 3.0 for example. It all adds up.

We also see some pretty nasty 95th percentile results for the 2600K in some games. That's where you'd really notice its age in daily use.

FYI the 1080p testing was at Medium, and yes, good catch on the GTX 1080, I think many people missed that and its vital to understanding these results; a 1080 is considered (upper) midrange now.

 

[MrGRiMv25]

It's kind of cool to see stuff from 8 years ago keeping up fairly well, unlike the early 2000's where something from 1999 would be completely out of steam by 2003.

 

[londiste]

Its actually quite a bit more if you think of it; that 18%/14% is with the 2600K @ 4.7 all core OC as the baseline. So in fact the 7700K already exceeds the 2600K despite a 500 mhz lower clock. It kind of echoes my own experience going from the 3570K to this 8700K. There is far more to gain than just a core/frequency advantage, there is a considerable IPC gap as well and the newer platform helps too. We're also talking PCIE 3.0 for example. It all adds up.

I deliberately took the percentages from stock clocks, not OC. OC to 4.7 (statically on all cores, I assume) will cut that difference to half
But you are right about even 500MHz faster 2600K losing to newer competition, just not by that much.

Newer generations do bring other things to table though, when not from CPU, then from platform. AVX is a factor in some places, DDR4 will bring a boost, PCI-e 3.0 had little impact when Sandy Bridge came out but has impact now, M.2 was not a thing etc.

 

[Vayra86]

I deliberately took the percentages from stock clocks, not OC. OC to 4.7 (statically on all cores, I assume) will cut that difference to half
But you are right about even 500MHz faster 2600K losing to newer competition, just not by that much.

Ah yes I misread, I noticed now you were talking about 1440p results and it all matches up again. Apologies

 

[Vario]

There are a lot of reasons to upgrade the chipset from Z68, beyond just the CPU. M.2 NVMe drives, DDR4 3200+, PCIE 3.0. If you have a Sandy, or Ivy, or even Haswell, it might finally be worth upgrading this year. Motherboards also do not last forever.

 

[Rahnak]

4 cores/4 threads does get in the way in a lot of newer games. 4 cores/8 threads is OK for all but a handful of games (Battlefield MP, Assassin's Creed: Odyssey and some Far Cries come to mind). From my personal experience (after going from i7-6700K to i5-8400) 6 cores/6 threads is roughly equal to 4 cores/8 threads.

Future-proofing is anyone's guess. More cores and threads is always good but for games, single-thread performance still matters a lot.

Yeah, I really want to play the last two ACs but I know better than to try with this CPU. If the new ryzens can deliver on all that hype I'll probably be picking up an 8 core for myself.

 

[SL2]

Yes, OCd 2600K brings the difference down by a lot but when comparing that also consider 7700k and 9700k are no slouches for overclocking either.

They kind of are, if looking at %. A 138 % OC on air isn't really a thing anymore.

Imagine a 9700K running at 138 %, ie 6350 MHz.. These days people are happy for a 600 MHz OC.

 

[Vayra86]

They kind of are, if looking at %. A 138 % OC on air isn't really a thing anymore.

Imagine a 9700K running at 138 %, ie 6350 MHz.. These days people are happy for a 600 MHz OC.

What's really happening though is just that chip design gets tighter and tighter, binning gets more aggressive and safe margins are being slowly reduced over time. You see this on CPU but also GPU and you see it with AMD and Nvidia and Intel. They all clock close to max potential out of the box and adjust clocks constantly and get smarter about it every gen. Its telling that XFR Ryzen results can be better than manual OCs. Another one is Nvidia OC headroom which is completely fixed and there seems to be little if any variation in chip quality. They all do similar clocks unless you can get a major temp reduction going, allowing boost to stretch its legs more. I think that is going to be the trend anyway: overclocking will be much more a case of allowing the chip itself to do its business as well as possible, the real bottlenecks being power and temperature. When you consider Intel pre-programs CPUs with set voltage points for every multiplier...

The reason isn't really that important in the end. Overclocking on air ust isn't that fun anymore.

Amen to that.. For me that trend started with 22nm Ivy already. In hindsight though... it was clear Intel also realized Sandy was way too good. I think that was a key motivator to switching from soldered to paste TIM. I remember Toms' article on how the heat was due to 'higher density' but that could never have been the whole story.

 

[SL2]

Also, compare this with a eight year old P4 single core with HT back when Sandy Bridge was launched in 2011.

The Sandy Bridge is still pretty good for its age today, the P4 was pretty slow eight years ago..

It all show hows what happened with Moores law, lack of competition, 10 nm fail, laptops and phones being more popular, and the need for faster CPU's slowing down among average consumers.
A 4 year old computer today can be pretty good today, but in 2003 it was so last millenium..

What's really happening though is just that chip design gets tighter and tighter, binning gets more aggressive and safe margins are being slowly reduced over time.

The reason isn't really that important in the end. Overclocking on air just isn't that fun anymore.

 

[natr0n]

Using a 2600k currently. It's a nice cpu for almost anything.

 

[SL2]

Using a 2600k currently. It's a nice cpu for almost anything.

Especially for 4K gaming! I couldn't imagine the difference being THAT small. 95th percentile looks similar too.

 

[TheoneandonlyMrK]

Also, compare this with a eight year old P4 single core with HT back when Sandy Bridge was launched in 2011.

The Sandy Bridge is still pretty good for its age today, the P4 was pretty slow eight years ago..

It all show hows what happened with Moores law, lack of competition, 10 nm fail, laptops and phones being more popular, and the need for faster CPU's slowing down among average consumers.
A 4 year old computer today can be pretty good today, but in 2003 it was so last millenium..


The reason isn't really that important in the end. Overclocking on air ust isn't that fun anymore.

Indeed but remember , it's very use case specific, just for gaming at 4k 60hz , little difference.
1080-1440p 60hz , little difference, but,
High fps big difference.
Office or simulation or rendering or many other tasks , massive difference.

My old Fx8350 had the same reasonable performance at 4k, 60hz , there's nothing in it in a lot of games, making an upgrade quite subjective.

 

[Vayra86]

Beware and take note its tested on a GTX1080! I will do this with a picture in the OP...

 

[dirtyferret]

Very nice read, and provides confirmation that we're well past the 4 core era for gaming, but also the major benefits extracted from AVX(2). Adjust accordingly

Upgrading from an Intel Core i7-2600K: Testing Sandy Bridge in 2019

www.anandtech.com

Note:

View attachment 122934

not really following your conclusion as you listed a test on an 4c/8T chip from eight years ago (not a quad chip) offering close to 100FPS+ in every gaming test when OC

TPU, gamers nexus and Toms reviewed an OC 8350 that offered better gaming performance then the Ryzen 1600/x. I'm not saying go out and buy one but 4 core CPUs are still capable gaming chips especially when OC.

 

[Vayra86]

not really following your conclusion as you listed a test on an 4c/8T chip from eight years ago (not a quad chip) offering close to 100FPS+ in every gaming test when OC

TPU, gamers nexus and Toms reviewed an OC 8350 that offered better gaming performance then the Ryzen 1600/x. I'm not saying go out and buy one but 4 core CPUs are still capable gaming chips especially when OC.

The trend is what my conclusion is about - I'm not saying a 2600K has become useless or anything - but you can see pretty sweet gains when more than 4 physical cores are present. Also, this test unfortunately lacks a 4c/4t CPU but you'd see a bigger hit in 95th percentile frame times than you see here; even in this test with a 4c/8t CPU the loss is slightly greater than the loss in average FPS.

When you zoom in on a per-game basis, you can see bigger gaps here and there, and major gains from an 8c/16t CPU. In addition, and that is why I'm stressing it in the OP now, this was tested on a GTX 1080, hardly the fastest GPU out there.

You have to consider as well that these are very high clocked CPUs. That is not the mainstream; the mainstream is actually moving towards higher core counts and not necessarily higher clocks, especially on mobile, you see very low baseclocks.

 

[dirtyferret]

The trend is what my conclusion is about - I'm not saying a 2600K has become useless or anything - but you can see pretty sweet gains when more than 4 physical cores are present. Also, this test unfortunately lacks a 4c/4t CPU but you'd see a bigger hit in 95th percentile frame times than you see here; even in this test with a 4c/8t CPU the loss is slightly greater than the loss in average FPS.

When you zoom in on a per-game basis, you can see bigger gaps here and there, and major gains from an 8c/16t CPU. In addition, and that is why I'm stressing it in the OP now, this was tested on a GTX 1080, hardly the fastest GPU out there.

You have to consider as well that these are very high clocked CPUs. That is not the mainstream; the mainstream is actually moving towards higher core counts and not necessarily higher clocks, especially on mobile, you see very low baseclocks.

I'm still not following you completely; yes the trend is towards six core CPUs but I don't see any major gains (or sweet gains) going to 8/16 CPUs as opposed to six cores CPUs.

The Nvidia 1080 is still very fast and cost $600+ at launch, I think most people with a Nvidia 1080 in their system probably have moved on from the 2500k or 2600k since they obviously are willing to put money into their gaming PC.

Anyone holding onto a 2500k or 2600k (or similar set up) is not too concerned with 95th percentile. Most likely they want to milk the platform for all it's worth while getting playable performance at solid 1080p settings otherwise they would have upgraded.

My personal guess is anyone still hanging onto those Sandy Bridge CPUs probaly has nothing more powerful then a Nvidia GTX1060 or AMD 580 and simply looks for close to 60FPS @ 1080p.

It's an interesting article none the less.

 

[GorbazTheDragon]

I find it really amusing that they test the 7700k with 2400 and 9700k with 2666, intentional or not this is hugely biased towards the older chip.

Basically everyone buying in the last year or so should have 3000/3200MT/s memory, especially for the i7s, and also agree that the newer chips should be run with realistic overclocks too.

I only moved from a 4790k to a 7700k and the difference in smoothness is extremely noticeable on a lot of titles, among them R6 Siege and ME Andromeda, even stuff like CSGO and BF3/4 is noticeably more consistent, though those are consistently above my 144Hz monitor even on the haswell chip.

While I do think that Sandy Bridge (especially at its time) was criminally underrated, I don't really think the graphs really show how big the disparity actually is. Though I do overall agree with the sentiments echoed in the conclusion...