The Core i7-8086K Review: 40 Years Of x86

[AgentLozen]

I also won one, but I am using a 4590, so I will happily use it.

Gratz on that btw. I could use an 8086K, but I didn't enter the giveaway so whose fault is that?

 

[Guest]

I want to drink the kool-aid too!

 

[AgentLozen]

I want to drink the kool-aid too!

It doesn't matter where the kool aid comes from if you're thirsty enough.

 

[War00Spawn]

I too won one of the i7-8086 CPU. But the big win is I got one of the 40 AMD Threadripper 1950X CPU's! With luck maybe AMD will send a Gen 2 version Time to upgrade my i7-4790K (But dam X399 MB are $$)

 

[Gam3r01]

I too won one of the i7-8086 CPU. But the big win is I got one of the 40 AMD Threadripper 1950X CPU's! With luck maybe AMD will send a Gen 2 version Time to upgrade my i7-4790K (But dam X399 MB are $$)

It was specified its a 1950x

 

[BryanFRitt]

How about some benchmarks for the June 8, 1978 Intel 8086 CPU ?

 

[BryanFRitt]

How about some benchmarks for the June 8, 1978 Intel 8086 CPU ?

 

[xpusostomos]

All the commenters above who don't know what's going on. "delidded" means you take the top off the CPU to add better thermal paste for better cooling and performance. This ruins your warranty. The "silicon lottery" is a cute way of saying some chips are better than others. These 8086k are chosen to be the best. If you get a cheaper chip, you're buying a lottery, some are better than others.

 

[jimmysmitty]

All the commenters above who don't know what's going on. "delidded" means you take the top off the CPU to add better thermal paste for better cooling and performance. This ruins your warranty. The "silicon lottery" is a cute way of saying some chips are better than others. These 8086k are chosen to be the best. If you get a cheaper chip, you're buying a lottery, some are better than others.

Pretty much true. I got a Q6600 G0 which alone was a great chip. Pretty much every one would OC to 3GHz just by setting the FSB to 333MHz and you didn't even have to touch the voltage. I was able to actually run mine at a lower than spec voltage. Was able to push it beyond 3GHz on stock voltage.

 

[cat1092]

The 8060K is indeed a very cool marketing effort by Intel, I give them credit on that end.

No one seems to be commenting on the couple of tests where an i5 kicks both the 8086K & 8700K. I've always had the question, and this (sort of) justifies it, isn't physical cores overall better than virtual? In other words, kick the virtual to the curb & have all physical.

How would that go over in the auto industry, a 4 cylinder engine with 4 virtual ones & call it an 8 cylinder engine? In a cloud of smoke, fast!

Am still on both of my Devil's Canyon builds (i7-4790K & i5-4690K), the first being delidded & IHS replaced for a slick, machined copper replacement from Rockit 88. Has shaved (on varied workloads) 15-20C from peak temps with the 4790K, haven't got around to doing the 4690K (as well as a i7-4770). All of my Haswell systems handles everything I throw their way, to include overclocking the i5-4690K to i7-4790K specs, very underrated chip.

Furthermore, the i7-8086 isn't a 5.0GHz CPU out of the box, Intel still has (a lot of) work to do to get there, if ever.

What Intel should had done was to celebrate their 1st true 4.0GHz (stock) CPU in the 4790K, just as the 8086K, in the same manner. Because for over 12 years after fumbling the ball, with Dell/HP as 'partners' to cool a single core 3.8GHz chip with H/T on an AIO CPU & exhaust fan with finned aluminum heatsink, 4.0GHz looked as though it never would happen. Intel would later regret this, and openly stated so a decade after the fact, as 'Moore's Law' (which isn't a true recognized scientific law by any means), lost meaning at Intel. NVIDIA has taken that 'law' & ran away with it on their GPU lineup, with real results between generations, while using less power to get there, usually a 30-35% power increase every couple of years, while doubling the memory. Intel can't do that anymore, if so, we'd have a true 5.0Ghz chip with close to 6.0GHz Turbo.

Thankfully, in 2014 history was written, as 4.0GHz was released to the masses. Some binned 4790K chips were also able to hit the 5.0GHz mark, although 4.6 through 4.8GHz were more typical. Once past 4.6GHz, voltages jumps fast (from 1.280 to over 1.4V if auto overclocking in the UEFI).

Passmark still has the i7-4790K slightly above the 6700K (part of the Skylake family), despite the latter running DDR4 RAM. I still believe based on the couple of i5 tests performed, just like my 4690K, physical cores performs better than virtual (need to test on my 4790K). With both Intel & AMD's vast resources, why are we still into the H/T & HyperTransport hype?

Cat

 

[ingtar33]

I've always had the question, and this (sort of) justifies it, isn't physical cores overall better than virtual? In other words, kick the virtual to the curb & have all physical.

yes they are. to be more specific; HT on average in tasks where it improves performance (it doesn't always) improves it an average of 20%-25%. So 4c/8t would perform roughly on par with a true 5 core cpu. AMD's version of hyperthreading (SMT) seems to perform slightly better then intel's hovering around 25%-35%, likely due to the fact it's a newer design and intel hasn't had any competition for hyperthreading... ever... to spur on development and improvement; but its still no where near the performance of a full core. And remember, these are in tasks where hyperthreading/smt works. Not all tasks benefit from hyperthreading/smt, in fact some actually see a performance hit.

So the short answer to your question is "yes" real cores are a lot better then simulated ones.

 

[AgentLozen]

I've read before that hyper threaded cores are roughly 10% as effective as a real, physical core. I can't recall the source for this though.

Hyper threading does have a positive impact in some applications. It's not like its a determent. Ingtar33 is right in that there are minor performance hits sometimes with HT enabled though.

What I want you to take away from my post is that hyper threading helps in some places and is good to have enabled. Its benefits out weigh its problems.

 

[jimmysmitty]

I've read before that hyper threaded cores are roughly 10% as effective as a real, physical core. I can't recall the source for this though.

Hyper threading does have a positive impact in some applications. It's not like its a determent. Ingtar33 is right in that there are minor performance hits sometimes with HT enabled though.

What I want you to take away from my post is that hyper threading helps in some places and is good to have enabled. Its benefits out weigh its problems.

Its normally 20% in highly threaded work loads.

I have not seen a program lose performance from HT since the original HT implementation in the Pentium 4 back in the early 2000s.

 

[ingtar33]

I have not seen a program lose performance from HT since the original HT implementation in the Pentium 4 back in the early 2000s.

It happens. usually only around a - 1%-3% hit on performance but usually the more intense and single cored a process is the more likely you'll see a small hit from hyperthreading.

 

[jimmysmitty]

I have not seen a program lose performance from HT since the original HT implementation in the Pentium 4 back in the early 2000s.

It happens. usually only around a - 1%-3% hit on performance but usually the more intense and single cored a process is the more likely you'll see a small hit from hyperthreading.

1-3% is too small to be considered a real performance hit. Its like a 1-3% performance gain. Normally its chalked up to margin of error as some sites will repeat the same test multiple times and end up with a 1-3% range of results that they average out.

 

[AgentLozen]

1-3% is too small to be considered a real performance hit.

I don't think anyone is saying that the performance hit is considerable.
I don't have a HT enabled system and software suite to test this on, but I recall reading in the past that it sometimes has a minor negative effect on performance in some workloads.
Even so, I stand by my original statement that the benefits of HT outweigh the detriments.

 

[xpusostomos]

I still believe based on the couple of i5 tests performed, just like my 4690K, physical cores performs better than virtual (need to test on my 4790K). With both Intel & AMD's vast resources, why are we still into the H/T & HyperTransport hype?

Does anybody outside of Intel actually understand what virtual cores are? Of course real cores are better than virtual ones. My guess is that some Intel engineer found that they could give half of the benefits of real cores without adding much to the silicon. Maybe when one thread is busy the other thread can do things, so they can have 2 threads that only partially interfere with each other without adding much complexity. The benefits are probably pretty specific to the actual silicon they were designing, and might not necessarily carry over to every CPU someone might want to build.