4C/4T vs 4C/8T vs 6C/6T

[newtekie1]

For the typical gamer type the 6 core 6 thread 8600K is perfectly adequate and should be for several years. It also can clock very high. Plenty of people on Overclock.net are in the 5.0Ghz+ range with it.

A typical gamer should be fine with an 4c/4t processor. The reality is, if you aren't doing a lot of multi-threading work, which games still really aren't, then a 3850K is still more than enough for gaming.

Pretty much all of those have delidded. 5.0 is easy with delid and proper paste - It can be archieved with very cheap air cooling. It's much harder and impossible on some without delid - This is why many settles with 4.7-4.8 GHz.

I've seen several 8700K's and 8600K's hit TJ MAX using 240-280mm AIO (H110's + H115i with PERFORMANCE preset + a few Noctua NH-D14/D15) ... Same CPU's did 5.2 and 5.3 after, with much much lower temps (around 75-85C avg load during burn in vs 100C+ aka throttling before)

This is how bad Intel's TIM is (the GAP does not help). USELESS for serious OC.

Delidding is a must if you want decent overclocks with good temps. Sad but true. Some people on this forum does not agree tho (LOL) ... Ask if Intel's TIM is good on overclock.net and see what people will say. 90% or more have delidded. And for good reason.

I'm one that doesn't agree that Intel's TIM is bad. From what I can tell they use some pretty good paste actually, the problem is the gap.

But the reality is, the difference between 4.8GHz and 5.2GHz is going to be negligible to most people that aren't going for benchmarking. Even the theoretical performance increase is less than 10%, normal use isn't gong to show much difference. And 99% of people will be happy at 4.8GHz. No one has to delid the processors, what they gain from it really isn't worth it despite how much overclockers like to whine about it.

 

[Vario]

A typical gamer should be fine with an 4c/4t processor. The reality is, if you aren't doing a lot of multi-threading work, which games still really aren't, then a 3850K is still more than enough for gaming.



I'm one that doesn't agree that Intel's TIM is bad. From what I can tell they use some pretty good paste actually, the problem is the gap.

But the reality is, the difference between 4.8GHz and 5.2GHz is going to be negligible to most people that aren't going for benchmarking. Even the theoretical performance increase is less than 10%, normal use isn't gong to show much difference. And 99% of people will be happy at 4.8GHz. No one has to delid the processors, what they gain from it really isn't worth it despite how much overclockers like to whine about it.

Agreed 100%. I think it is the gap and the 100-200 mhz you might gain does not matter in the slightest. Nor does the temperature reduction really matter, these processors can handle heat pretty well.

 

[SamirD]

I really don't think the 4c/8t is close enough to the 6c/6t results to threaten it's marketshare. There is still a pretty significant gap between the two. If we assume the 4c/8t part was a little more expensive than the 8350k and cost about $200, the 8600k is $250 and I think the extra $50 is worth it if you do a lot of multi-threaded work.

It's interesting to hear you say that considering the graphs show a more significant gap between 4c8t and 4c4t than 6c6t. In fact, aside from the Intel test, the 4c8t seems right on the 6c6t's heels in most applications--that is, unless I read the graphs wrong.

 

[Hockster]

Supposedly Intel's choice of TIM is based on longevity and not pure thermal properties.

 

[R-T-B]

Supposedly Intel's choice of TIM is based on longevity and not pure thermal properties.

I had a tube of the same "Dow Corning" stuff I suspect Intel uses. It came with one of their "Extreme" heatsinks for LGA1366, one of their few tower style coolers.

The stuff was opened circa 2015 and was manufactered in 2012...

...and was pure, total, sand like powder in the package. The syringe would barely push it out.

I have serious doubts about its longevity. Hopefully it's not the same stuff, but I also have my doubts there.

 

[jboydgolfer]

For the typical gamer type the 6 core 6 thread 8600K is perfectly adequate and should be for several years. It also can clock very high. Plenty of people on Overclock.net are in the 5.0Ghz+ range with it.

I agree. I got mine into 5.2ghz, on aio with no prob right out of the box. Pretty much anyone gets them the basic close to 5ghz just as easy. Great chip.

what they gain from it really isn't worth it despite how much overclockers like to whine about it.

100% correct. i think the "i have nothing better to do than cry about it" comes out in the anal retentive "it should be perfect for what I want" people. Fact of the matter is, I will always trust that the Multi BILLION dollar/ year company has a bit more knowledge in what is or isnt right for a Good CPU over some nub on a PC site who cries endlessly about how his CPU needed this or that to get it here or there.

additionally:

it always blows my mind how blind a person can be, saying how this or that is wrong with a CPU, all while oblivious to how arrogant they come off thinking that a entire R&D team missed something while testing thousands of CPU's, that a single kid caught in his moms basement while testing his new CPU in a $800 PC which happens to be his 1st PC

One example of a miserable ocd grump.
Their not happy for the 40% they can overclock their new cpu....theyre pissed about the 2% they cant overclock it without preparations, & proper equipment.

 

[Vayra86]

It's interesting to hear you say that considering the graphs show a more significant gap between 4c8t and 4c4t than 6c6t. In fact, aside from the Intel test, the 4c8t seems right on the 6c6t's heels in most applications--that is, unless I read the graphs wrong.

4c8t is right on 6c6t's heels in a 'best case scenario' where you actually use a workload that benefits from HT and is well threaded. In the most common use cases however, it falls down to being 'worth' about half a real core or fár less, up to the point of not having a benefit at all. For CPUs, a 25% performance drop is significant; that's more than the last three Intel gens combined.

6c6t on the other hand is 100% consistent in comparison. It'll do everything equally fast, alongside any number of other tasks. Also consider the marketplace where a 6c6t i5 is now available at same price as 4c8t of last gen.

 

[jabbadap]

I agree. I got mine into 5.2ghz, on aio with no prob right out of the box. Pretty much anyone gets them the basic close to 5ghz just as easy. Great chip.


100% correct. i think the "i have nothing better to do than cry about it" comes out in the anal retentive "it should be perfect for what I want" people. Fact of the matter is, I will always trust that the Multi BILLION dollar/ year company has a bit more knowledge in what is or isnt right for a Good CPU over some nub on a PC site who cries endlessly about how his CPU needed this or that to get it here or there.

additionally:

it always blows my mind how blind a person can be, saying how this or that is wrong with a CPU, all while oblivious to how arrogant they come off thinking that a entire R&D team missed something while testing thousands of CPU's, that a single kid caught in his moms basement while testing his new CPU in a $800 PC which happens to be his 1st PC

One example of a miserable ocd grump.
Their not happy for the 40% they can overclock their new cpu....theyre pissed about the 2% they cant overclock it without preparations, & proper equipment.

Are you using AVX offset? For fluidity one should use AVX stable clock speeds rather than maximum obtainable AVX offsets clock speeds...

On the topic though great work @newtekie1 . For the gametest you could make gtx1060 to not bottlenecking that much by running tests at 1080p at low settings or use 720p resolution. But when 6c/6t beats 4c/8t on all cpu intensive tests the gap will most probably just widen on games tests.

 

[jboydgolfer]

Are you using AVX offset? For fluidity one should use AVX stable clock speeds rather than maximum obtainable AVX offsets clock speeds...

I'm not at my PC so I'm going to have to go off of memory but off the top my head I'd say I have AVX set to auto . It holds the overclock frequency even under sustained 100% load

I had to increase short duration and long duration a small amount but once I got it tuned in right (130 & 160 respectively) iirc , it held its clocks well. Set to offset +50,loadline1

 

[Hayder_Master]

very nice work dear, i am asking your permission to share this awesome work on social media (i will translated to Arabic) and sure i will mention you name.

 

[jabbadap]

I'm not at my PC so I'm going to have to go off of memory but off the top my head I'd say I have AVX set to auto . It holds the overclock frequency even under sustained 100% load

I had to increase short duration and long duration a small amount but once I got it tuned in right (130 & 160 respectively) iirc , it held its clocks well. Set to offset +50,loadline1

Yeah if it truly is 5.2GHz AVX stable, then you have won the silicon lottery.

Asrock AVX auto offset can be 4 or 0 depending if Multi Core Enhancement is enabled or not. So pour some AVX load on it, if it drops to 4.8GHz all core then auto=4.

 

[jboydgolfer]

very nice work dear, i am asking your permission to share this awesome work on social media (i will translated to Arabic) and sure i will mention you name.

^^^^^^^^
@newtekie1 should get his attention in case he has thread notifications off.

https://www.techpowerup.com/forums/threads/4c-4t-vs-4c-8t-vs-6c-6t.241983/post-3819258

 

[bug]

Meh, flamebait.

Everybody knows the choice of more cores vs more GHz is down to each users workflow.

 

[dirtyferret]

The 8600k is "perfectly adequate" for gaming , looks like all those ryzen 1600 owners are screwed if they want more then adequate gaming performance...

 

[jboydgolfer]

Any cpu made in the last few years with 4 or more cores is capable of great gaming, regardless of wether its intel of or amd. Any cpu feom the last 10 years with 4 or more cores is fine for gaming. Extra cores are great, but not required. Atleast thats my experience.

 

[newtekie1]

It's interesting to hear you say that considering the graphs show a more significant gap between 4c8t and 4c4t than 6c6t. In fact, aside from the Intel test, the 4c8t seems right on the 6c6t's heels in most applications--that is, unless I read the graphs wrong.

15-20% is not right on the heels.

very nice work dear, i am asking your permission to share this awesome work on social media (i will translated to Arabic) and sure i will mention you name.

Sure. As long as you credit me.

 

[Bill_Bright]

I somehow missed this thread but I am glad dirtyferret pointed it out here.
IMO, better late than never to say, excellent analysis! Well done newtekie1!

I have to say I didn't learn anything new but that's fine because what this comparison did was (1) confirm what I already knew. But more importantly, it (2) explained in a very succinct and easy to understand manner how hyperthreading (HT) impacts performance on certain tasks. It also (3), gives insight to the difference between HT and the number of actual cores.

I would like to add one thing to the discussion. It should be remembered that Windows, the program itself, supports hyperthreading. What that means is when the OS is performing various housekeeping or other tasks (perhaps in the background), that is, when the OS is "multitasking" with a HT capable processor, the OS is likely able to complete some or all of those tasks more quickly. That in turn, frees up system resources more quickly - resources that can then be allocated to other tasks sooner, even towards tasks that do not support HT.

Conclusion? Leave the defaults alone. That is, keep HT enabled. In the very rare scenario where HT might hinder performance, the impact is typically negligible and insignificant and only noticeable in benchmark tests, not in real world. The benefits for keeping it enabled, however, are often significant and very noticeable, depending on the tasks being being performed.

As a side note,

I'm one that doesn't agree that Intel's TIM is bad. From what I can tell they use some pretty good paste actually, the problem is the gap.

I was so happy to see this comment. I cringe every time I see someone criticize OEM TIM from Intel or AMD.

TIM is incredibly critical for thermal coupling and surely both Intel and AMD know and appreciate this. Yet it is incredibly inexpensive - especially in the quantities Intel and AMD use. Those thermal pads add barely a couple pennies to the cost of the CPU and heat sink fan assembly. Considering the cost vs benefits, it makes no sense for either company to supply OEM TIM that could not perform effectively. So they don't!

The most effective transfer of heat occurs with direct metal to metal contact so I also agree with newtekie1 that it is the gap that is the problem. That gap is inevitable when using "pads". BUT folks need to realize and remember that the bulk of those pads consists of purified "paraffin", typically a petroleum distillate very similar to kerosene. It takes a couple heat up and cool down cycles for that paraffin to thoroughly melt, spread out completely, then evaporate away to leave only the very thin layer of thermal interface materials behind.

So unless you are doing extreme overclocking (and in that case, you should be buying a CPU that does not come with an OEM supplied cooler in the first place), give the OEM TIM a little time and a few heat up and cool down cycles to reach its peak effectiveness.

 

[xorbe]

I can literally answer this from memory, having benchmarked compiling Liunx on 8086K very recently. 4C/8T HT is about 5C/5T at best. Also, HT threading can suffer up to about 8% performance loss per thread compared to non-HT (this could be a variety of things, cpu or Linux scheduler). 6C/6T will beat 4C/8T in multi-threaded loads, it will be about 20% faster.

 

[trparky]

I'm putting this post in this thread as versus the other thread that linked to this thread.

Basically under normal circumstances a core remains not fully utilized most of the time

Yes, very much so. Most of the time your processor core is doing nothing. It's tapping its virtual foot and saying "Hey, will someone give me something to do here. I'm bored!" Yes, this is true even in today's systems with SSDs because even though SSDs can deliver data faster than ever before your processor is still sitting around bored out of its silicon mind most of the time. Why? It has to do a lot with a great many things including branch prediction failures, cache misses, and various other situations that can result in performance bottlenecks. While a processor core is waiting for something (either because of a branch prediction failure or cache miss) that processor can't do anything; it's got to wait for whatever resources that it needs to continue doing work. Meanwhile that core is wasted. HT introduces the opportunity to run other tasks on that same core while the first "virtual" core is still waiting for whatever resources it needs to continue doing it's work.

 

[Ruru]

Went from 2c/2t (Pentium G4400) -> 2c/4t (Pentium G4560) -> 4c/4t (i5-7600K @ 4.7GHz) -> 4c/8t (i7-7700K @ 5GHz) 5c/10t (i7-5820K @ 4.5GHz with a defective core) and every step has given a nice boost.

 

[Vayra86]

Went from 2c/2t (Pentium G4400) -> 2c/4t (Pentium G4560) -> 4c/4t (i5-7600K @ 4.7GHz) -> 4c/8t (i7-7700K @ 5GHz) 5c/10t (i7-5820K @ 4.5GHz with a defective core) and every step has given a nice boost.

Absolutely, the key is always in timing vs. the applications you use. And more is always better until its too much

 

[GorbazTheDragon]

Surprised you didn't answer the original question of whether the 4/8 would cannibalise the 6/6 market, I think the answer would definitely be yes.

The thing it comes down to is that an unlocked 4/8 cpu would be able to more than match stuff like the 8400 and other mid/low clocked 6/6 chips.

If the new Xeon E3s 4/8s are well priced they could be a good alternative to the i3 and i5 4/4s and 6/6s.

 

[Melvis]

For those people that want to know about Gaming results with cores disabled etc etc

 

[xorbe]

The only reason SMT disabled is pushed as an SKU is to prop up prices of SMT enabled parts.

 

[Totally]

Stuff

If you throw out the Intel burn test and cinebench, all the 4c/8t and 6c/6t results are within 15% (13% at widest, 6% at narrowest) of each other. We your co-worker was saying has merit as going from 4c/8t to 6c/6t doesn't yield enough benefit to justify it.