What is the current sweet spot for a CPU?

[AtenRa]

What is the current sweet spot for a CPU, value/performance? I'm helping one of my son's friends build a rig & yes, I am lazy/short-of-time, & I've not been following the latest hardware. My rig still runs great with a 2600k at 4Ghz so I avoid the hardware forums so I don't catch "upgrade fever". Although I think a new video card is in my future.

He of course has limited funds & needs everything, case, PS, everything. Just seeing the prices Intel is charging, I think I will be recommending AMD, but then again, that is why I am posting this here, to hopefully find some folks to be find enough to provide some guidance.

We may do some overclocking, if the new CPU is capable.

Thanks in advance!
Joe S.

Tell us what budget he has and then we could give you a better advise.

Also, what games he will play, if it is for DOTA 2/LOL etc the recommendations will be different than playing BF 4/ Witcher 3 etc.

 

[CHADBOGA]

YES! Sounds like we got a winner right there. Thanks!

& thanks everyone for your input, seriously, this forum is the best tech forum around. Once we decide on the build, I will post it here. 4690k sounds right though.

That is a pretty good choice, but also keep your eye out for the 4790K, IMHO, the best value CPU Intel has released for years.

 

[MiddleOfTheRoad]

wow thanks because I had no clue how those cpus performed at all. maybe if you had a clue about Skylake then you would know it scales great with faster memory and they only used 2133MHz in that test you linked to.

Wow, thanks because you clearly can't run faster ram in a Haswell (insane sarcasm here, dude).... Haswell also scales up performance similarly......

BTW Toyota, they make DDR3 speeds up to 3000+ Mhz memory......

 

[bruinwar]

Sorry for the delay on my response. His budget? That he doesn't seem to know or I would have shared it already. I put together a build that hit $800 & he kina blanched at that, so I replaced the SSD with a spinner, cut the RAM from 16 to 8 gig & got it to below $700.

He wants to play Fallout 4 & Witcher 3, that much I know. A $200 CPU & a $200 video card is pretty much a minimum IMO. Case, PS, MB, RAM, HD... I don't see building it for much less that $700. I've got a Win7 disk that will work if he goes to Win10 after patching, or he can add the cost of an OS to the pile.

Thanks everyone for your input.

 

[MiddleOfTheRoad]

1-2 fps difference could also be the statistical error,it's pretty obvious that the GPU is at the limit in these tests and that's why the cpus are so close together.

Well, the synthetics show the older, considerably cheaper CPU is a bit stronger. To be advocating the more expensive CPU based on the DDR4 is silly. Especially when the OP is looking for the gaming CPU sweet spot. Skylake doesn't justify its current price tag, although that will likely change in the future.

 

[TeknoBug]

Stick with the 2600K for now til you save up enough to build a 6700K setup, it'll be a fairly good jump in performance.

 

[toyota]

Wow, thanks because you clearly can't run faster ram in a Haswell (insane sarcasm here, dude).... Haswell also scales up performance similarly......

BTW Toyota, they make DDR3 speeds up to 3000+ Mhz memory......

once again showing how much you know. Haswell with faster memory does NOT scale anything like Skylake.

 

[Madpacket]

once again showing how much you know. Haswell with faster memory does NOT scale anything like Skylake.

Umm, why do you think this is the case?

http://www.gamespot.com/forums/pc-m...ulusly-high-performance-gained-by-f-32754964/

Note they incorrectly label the RAM in the graph as DDR4 when in fact they're using DDR3. These are huge differences though using DDR3

 

[bystander36]

Umm, why do you think this is the case?

http://www.gamespot.com/forums/pc-m...ulusly-high-performance-gained-by-f-32754964/

Note they incorrectly label the RAM in the graph as DDR4 when in fact they're using DDR3. These are huge differences though using DDR3

That is true, they were talking about the DDR3 on this Haswell system in the paragraph before the chart. Some mislabeling most likely happened.

 

[Magic Carpet]

@bystander36

Interesting, so what's up with that, a trend or just one outlier?

 

[bystander36]

@bystander36

Interesting, so what's up with that, a trend or just one outlier?

We'll have to wait and see. It's an outlier atm, but it could become a trend. Who knows.

 

[MiddleOfTheRoad]

once again showing how much you know. Haswell with faster memory does NOT scale anything like Skylake.

Yeah, sure it doesn't.....

The minor advantage for Skylake's scaling is arbitrary in every benchmark I've seen. You're talking about 3 or 4 frames per second on the majority of modern games.
Certainly not worth the $120 price premium. When you factor in that the Haswell at stock clock starts above a stock clock 6700K in performance, even the diminishing
performance returns for DDR3 versus DDR4 -- it largely end up with a wash unless you are buying ridiculously expensive DDR4 (thus expanding the already considerable
price penalty over Haswell). Skylake needs a major price drop to be considered a "sweet spot."

 

[toyota]

Yeah, sure it doesn't..... Keep digging your hole.

if you had a clue then you would know darn well that Skylake scales much better with faster memory than Haswell. It is not even close.

AGAIN I agree that its a poor value compared to the 4790k and have said that multiple times.

 

[jihe]

E5-2670

 

[Magic Carpet]

if you had a clue then you would know darn well that Skylake scales much better with faster memory than Haswell. It is not even close.

Perhaps, I haven't compared them head-to-head with fast DDR sticks. But, Haswell can also shine with faster memory. I'd like to see memory scaling of Haswell-E vs Skylake-S with the same memory type, however. It's obvious, that even 2-3 year old Haswell could benefit from faster memory in CPU limited situations.

 

[Sweepr]

When you factor in that the Haswell at stock clock starts above a stock clock 6700K in performance

Not sure if trolling.

Core i5 6600K vs Core i5 4690K at fixed 4.5GHz, Skylake is:
16.2% faster @ Battlefield 4 MP
10.9% faster @ Counter Strife Global Offensive
6% faster @ Crysis 3
22.7% faster @ Far Cry 4
13% faster @ GTA V
12.5% faster @ The Witcher 3
9.2% faster @ Watch Dogs
14.7% faster @ Project Cars
14.8% faster @ Starcraft 2
24.5% faster @ Total War Attila

Overall gaming performance per clock (IPC) from 14 games/tests @ 1080p Ultra:

Sandy Bridge to Haswell: 13.5%
Haswell to Skylake: 12.8%
Skylake is faster than Broadwell with eDRAM.

Another review, Eurogamer compared 4 generations of Core i5 running the latest titles at 1080p.

...a fully maxed GTA 5 is CPU-limited even on an overclocked Core i7 4790K, while certain scenes in the Witcher 3 on ultra settings can also cause difficulties for even the most powerful processor. Our contention is that more games like this will come along, and in demanding CPU titles, the Core i5 6600K will make a noticeable difference.

But in the here and now, the Core i5 6600K is a strong product - at worst it provides mildly enhanced performance over its predecessors, at best it's noticeably faster and should help to reduce CPU bottlenecking during gameplay. For those looking to buy or construct a new, capable gaming PC, the i5 remains the default choice and the 6600K is the best iteration yet - AMD's older eight-core FX chips are cheaper, but our tests clearly demonstrate that despite the low-power many-core set-up of the new console CPUs, modern games generally prefer the high per-core performance that Intel provides.

In the games tested by them Core i5 6600K (Skylake-S) provided a larger performance per clock performance gain than Core i5 4690K (Haswell) compared to Core i5 2500K (Sandy Bridge).

Core i5 6600K vs Core i5 4690K (same 3.5-3.9GHz base/turbo)
- 17% faster @ The Witcher 3
- 11% faster @ GTA V
- 10% faster @ Battlefield 4

And both reviews used bread and butter DDR4-2666, not some crazy expensive kit. Same holds true for Core i7-6700K, it is faster than the Haswell equivalent at stock (both apps and games). Not a big difference and maybe not worth the price difference right now, but it is faster.

 

[MiddleOfTheRoad]

Not sure if trolling.
And both reviews used bread and butter DDR4-2666, not some crazy expensive kit. Same holds true for Core i7-6700K, it is faster than the Haswell equivalent at stock (both apps and games). Not a big difference and maybe not worth the price difference right now, but it is faster.

Sorry, but Passmark has a lot more credibility than the cherrypicking users from the Anandtech forums.

The i7 4790K earns the higher score on Passmark.

 

[MiddleOfTheRoad]

Perhaps, I haven't compared them head-to-head with fast DDR sticks. But, Haswell can also shine with faster memory. I'd like to see memory scaling of Haswell-E vs Skylake-S with the same memory type, however. It's obvious, that even 2-3 year old Haswell could benefit from faster memory in CPU limited situations.

+1

Agreed, Haswell scales pretty darn well with faster memory. DDR4 does seem to have a modest edge, but not enough to justify the added cost for the CPU. To put it modestly, Toyota is greatly exaggerating the performance advantage.

 

[Durp]

if you had a clue then you would know darn well that Skylake scales much better with faster memory than Haswell. It is not even close.

Haswell scales very well with faster memory, similar to Skylake but people seem to overlook it. It's probably overlooked because of so many completely idiotic CPU benchmarks out there with settings that cause the GPU to bottleneck and hide any advantages of faster memory. Faster ram can show impressive gains when the processor is the bottleneck.

Look at this for example, especially the low resolution results.

http://www.xbitlabs.com/articles/memory/display/haswell-ddr3_7.html#sect0

 

[Sweepr]

Sorry, but Passmark has a lot more credibility than the cherrypicking users from the Anandtech forums.

The i7 4790K earns the higher score on Passmark.

Sure, that's why all reputable websites include it in their reviews. Passmark scores that can't even differenciate between stock and OCed are the ultimate metric of CPU performance. LOL.

 

[ShintaiDK]

Sorry, but Passmark has a lot more credibility than the cherrypicking users from the Anandtech forums.

The i7 4790K earns the higher score on Passmark.

Passmark is a horrible metric to use for any compare. Its been pointed out over and over again. But the benchmark suits your purpose, hence you use it.

 

[Headfoot]

Used gear is always going to be better performance for the same price at the cost of the warranty. Given that you are going to have a really hard time getting some brands to honor RMAs (Asus) this is more or less valuable depending.

You can now overclock locked Skylake processors. IMO the "Sweet Spot" is an i5-6400 that can be overclocked. You get 4 cores, and you dont have to pay the "K" unlocked price tax. This way you still get the most modern core, the most modern motherboard features (nvme/pci-e+m.2 ssd support). http://overclocking.guide/overclocking-non-k-intel-skylake-cpus-performance-tests/. i5-6400 is $185 on Newegg right now, with z170 mobos at ~$110. EDIT: just checked, basically all of Asrocks lineup can OC locked Skylakes: http://www.asrock.com/news/events/SkyOC/. Boards as cheap as $100.

I see absolutely no reason to get a 4690k + z97 when OC'd i5-6400 + z170 is available. If you are willing to overclock, the 6400 should overclock quite well. If you are not willing to overclock, then a 4790k is the best bet as it has the highest out-of-the-box clocks at a reasonable price.

 

[MiddleOfTheRoad]

Passmark is a horrible metric to use for any compare. Its been pointed out over and over again. But the benchmark suits your purpose, hence you use it.

You mean a benchmark that hasn't been rigged by the Intel Compiler? Yep.... Passmark doesn't appear to rig the results for a specific manufacturer. That's why you hate it.

For synthetics, I automatically discount PCMark and 3DMark because their hands have been already caught in the "rigged" cookie jar. While Passmark doesn't appear to have been caught cooking the books yet.

http://www.zdnet.com/article/pcmark-2005-memory-benchmark-favors-intel-cpus/
http://www.geek.com/games/nvidia-still-cheating-even-with-latest-3dmark-build-552713/

 

[bystander36]

You mean a benchmark that hasn't been rigged by the Intel Compiler? Yep.... Passmark doesn't appear to rig the results for a specific manufacturer. That's why you hate it.

For synthetics, I automatically discount PCMark and 3DMark because their hands have been already caught in the "rigged" cookie jar. Passmark hasn't been caught cooking the books yet.

http://www.zdnet.com/article/pcmark-2005-memory-benchmark-favors-intel-cpus/
http://www.geek.com/games/nvidia-still-cheating-even-with-latest-3dmark-build-552713/

All "synthetic benchmarks" are questionable. They only give you an idea of what could be reality. Real world games and applications are what matter. Some of these synthetic benchmarks test very specific features, but in the real world, there are lots of other things happening at the same time.

You don't build a PC for synthetic benchmarks. At least most people don't. You build them for specific applications, gaming and what ever you actually do with your PC. Unless you compete in OC contests, synthetic benchmarks mean very little.

 

[MiddleOfTheRoad]

Sure, that's why all reputable websites include it in their reviews. Passmark scores that can't even differenciate between stock and OCed are the ultimate metric of CPU performance. LOL.

Not to be the bearer of bad news -- But I'm pretty sure Tom's Hardware is a more popular web site than Anandtech. They use Passmark quite a bit.

Passmark scores are used a lot in PC World magazine, too.... The magazine with the largest circulation of any covering computers. Seems pretty legit to me.

Considering virtually every other synthetic benchmark excluding Phoronix has been caught cheating..... I generally stick to the 2 that don't seem to do it.