5930k @ 4.6 GHz

[Shmee]

So I now have my 5930k at 4.6 GHz, currently testing for stability. I upgraded my CPU cooler from the 212 Evo to a Noctua D15, which is helping a lot with temps. I may push for 4.7, time will tell. On the Asus X99 Deluxe.

I suspect this will give similar performance to a lightly OCed 8700k, so I think I am good for now, plus I get quad channel RAM and extra PCIE lanes

I will likely put the 212 Evo in my Ryzen 1700 rig.

 

[Boris Morozov]

Nice! I'm running my 5930k at 4.25ghz with ddr4 3000mhz. I probably could run a bit higher but with my cooling I think it's ok where it's at.. I play most my games at 4k and could benefit with a higher clock speed.

 

[ob3y]

I have that same Mobo with a 6850k @ 4.4 24/7 stable and occasionally 4.5. what voltages are you running to hit 4.6?

Sent from my SM-G955U using Tapatalk

 

[Shmee]

Well, I am at about 1.3 vcore, but I backed off to 4.5 after a BSOD in SWBF. More testing needed.

 

[moonbogg]

I'm wondering if I should stick with my 6800K or grab an 8700K. I'll likely stick with what I got, but I do like the sound of something new and shiny that's faster. I at least must wait for reviews...probably.

 

[Carfax83]

A 5930K at 4.6ghz will still be behind even a stock 8700K. The difference in IPC is enough to make the 8700K practically untouchable. Even Broadwell-E has a significant IPC advantage over Haswell-E to make the latter need a 300-400mhz advantage in clock speed to pull even in general workloads, and about a 600 to 700mhz advantage in clock speed for floating point and SIMD workloads.

 

[WhoBeDaPlaya]

Bought a 6950X for stupid cheap last week (~$250).
Insanely hot sucker though - was tough to keep cool even at 4.3GHz

 

[Carfax83]

Bought a 6950X for stupid cheap last week (~$250).
Insanely hot sucker though - was tough to keep cool even at 4.3GHz

How in the hell did you manage a 6950x for that price?! Whoever sold it to you, obviously didn't know its value.

 

[WhoBeDaPlaya]

How in the hell did you manage a 6950x for that price?! Whoever sold it to you, obviously didn't know its value.

Kind of. He thought it was defective and didn't want to bother troubleshooting it.
Brought it home, cleaned it up with IPA, and proceeded to Prime95 / IBT it at stock overnight, then OCed it

 

[Carfax83]

I'm wondering if I should stick with my 6800K or grab an 8700K. I'll likely stick with what I got, but I do like the sound of something new and shiny that's faster. I at least must wait for reviews...probably.

You have to ask yourself, whether you need the additional CPU power. If you already get good frame rates in games and aren't CPU limited, then I wouldn't bother. I upgraded from a 5930K to a 6900K, and that was more of a substantial upgrade as I got two extra cores and a bump in IPC. But I didn't do it because I was CPU limited in games. I did it because I'm doing a more video encoding than I usually do, plus Haswell-E's memory controller defect with the poor write performance ticked me off

 

[Carfax83]

Kind of. He thought it was defective and didn't want to bother troubleshooting it.
Brought it home, cleaned it up with IPA, and proceeded to Prime95 / IBT it at stock overnight, then OCed it

Well, he should have sent it to Intel as it should have been under warranty. Oh well, a sucker is born every minute. His loss, and your gain

 

[WhoBeDaPlaya]

Well, he should have sent it to Intel as it should have been under warranty. Oh well, a sucker is born every minute. His loss, and your gain

I honestly felt a little bad, but you do come across the odd deal like this on CL from time to time!

 

[moonbogg]

I honestly felt a little bad, but you do come across the odd deal like this on CL from time to time!

You should feel terrible. I wouldn't keep that thing. I'd just give it away to clear your conscience for sure. pm me I can help.

@Carfax83

I don't feel very CPU limited and its fast enough for me, but still though I want that new shiny, lol. I had a hard time knowing if the 6800K was better than my 3930K or not. I'm sure it is, but was hard to tell in gaming since I changed the GPU as well. Would probably be even harder to tell with 8700K vs 6800K. Being stuck at 100hz and also being a single GPU user now anyway, I doubt it would do anything for me. Rocketleague runs fine as it is, lol.

 

[tamz_msc]

Broadwell to Haswell IPC advantage is 5 percent, according to Intel themselves.

 

[Carfax83]

Broadwell to Haswell IPC advantage is 5 percent, according to Intel themselves.

It's more than 5% according to Intel themselves (unless > doesn't mean greater than), and that's in general code. In floating point and SIMD, the gains are higher, more like 10-15% and up depending. When I upgraded from a 5930K to a 6900K, general desktop performance was noticeably snappier.

Intel 6850K vs 5820K at same clocks and settings

 

[tamz_msc]

It's more than 5% according to Intel themselves (unless > doesn't mean greater than), and that's in general code. In floating point and SIMD, the gains are higher, more like 10-15% and up depending. When I upgraded from a 5930K to a 6900K, general desktop performance was noticeably snappier.

Intel 6850K vs 5820K at same clocks and settings

General desktop performance isn't a proxy for IPC, and why should improved SIMD performance result in faster desktop response in the first place?

Quad-channel DDR4 was improved in Broadwell-E, on top of that it had slightly higher default clocks. If we exclude TBM 3.0, then the difference between a 5930K(3.5GHz base, 3.7GHz boost) and the 6850K(3.6GHz base, 3.8GHz boost) is usually around 5%.

 

[Carfax83]

General desktop performance isn't a proxy for IPC, and why should improved SIMD performance result in faster desktop response in the first place?

I never said, or even implied anything of the above, so I don't know where you're getting that from. My comment about the general desktop performance was just anecdotal, you know, since I've actually owned these CPUs. And who the hell said that improved SIMD performance results in faster desktop response? Definitely wasn't me. You did that on your own. Too bad we can't down vote on these forums.

Quad-channel DDR4 was improved in Broadwell-E, on top of that it had slightly higher default clocks. If we exclude TBM 3.0, then the difference between a 5930K(3.5GHz base, 3.7GHz boost) and the 6850K(3.6GHz base, 3.8GHz boost) is usually around 5%.

I gave you the Intel slide which shows greater than 5% , and then I linked to a WCCFTech article which ripped some benchmarks from Overclock.net forums of a 4.2ghz 6850K and a 4.2 5820K, with the uncore and memory at the same speed. The 6850K was almost 15% faster than the 5820K in the Firestrike physics test!

So given this, why you're talking about TBM 3.0, memory controllers and stock clocks is a mystery. In fact, I don't even know why you bothered to respond at all given that you have no evidence to support your claims.

 

[tamz_msc]

I never said, or even implied anything of the above, so I don't know where you're getting that from. My comment about the general desktop performance was just anecdotal, you know, since I've actually owned these CPUs. And who the hell said that improved SIMD performance results in faster desktop response? Definitely wasn't me. You did that on your own. Too bad we can't down vote on these forums.



I gave you the Intel slide which shows greater than 5% , and then I linked to a WCCFTech article which ripped some benchmarks from Overclock.net forums of a 4.2ghz 6850K and a 4.2 5820K, with the uncore and memory at the same speed. The 6850K was almost 15% faster than the 5820K in the Firestrike physics test!

So given this, why you're talking about TBM 3.0, memory controllers and stock clocks is a mystery. In fact, I don't even know why you bothered to respond at all given that you have no evidence to support your claims.

I have all the evidence I need-

Anandtech's Professional performance on Windows.
Hardware.fr graphs.
IPC comparison at 4GHz across a range of applications.
5-7% on average, without accounting for clock speeds.

Anecdotes and a wccftech article vs hard numbers from reliable sources.

 

[Carfax83]

I have all the evidence I need-

Anandtech's Professional performance on Windows.
Hardware.fr graphs.
IPC comparison at 4GHz across a range of applications.
5-7% on average, without accounting for clock speeds.

Anecdotes and a wccftech article vs hard numbers from reliable sources.

From your own link, look at these juicy morsels hard numbers in heavy FP code. Intel 6850K is faster than a 5960x despite the latter having an extra two cores, and almost 30% faster than its predecessor the 5930K in the ray tracing benchmark and 25% faster in the NAMD molecular benchmark. Kind of supports by argument that Broadwell had some nice upgrades over Haswell in the FP and SIMD department, plus it all of a sudden makes WCCFTech Overclock.net forums look very credible

BTW, the IPC comparison you posted at the end from Hardware.fr is a joke, because the mainstream Broadwell parts have a L4 cache which makes doing an effective comparison all but useless.

Hey, but at least you said 5-7% this time instead of 5%, so at least we're making progress.

 

[tamz_msc]

BTW, the IPC comparison you posted at the end from Hardware.fr is a joke, because the mainstream Broadwell parts have a L4 cache which makes doing an effective comparison all but useless.

It's only a joke for those who don't understand that pseudo-L4 EDRAM in Broadwell only affects things like 7-zip and WinRAR, excluding games of course.

I specifically excluded Linux comparisons because of inconsistencies where 6850K is faster than 5960X but 6800K is slower than 5820K in the same test. Then there's things like these:

I guess people really look out for a particular kind of numbers in any given data that suits their beliefs and ignore everything else(everybody, including you, suffers from this disease to a varying extent).

EDIT: here, read this if you have the time, will save me some time in future discussions.

 

[Carfax83]

It's only a joke for those who don't understand that pseudo-L4 EDRAM in Broadwell only affects things like 7-zip and WinRAR, excluding games of course.

Yes of course, how could I have forgotten that it was actually OK in this scenario to introduce more variables into an IPC clock for clock comparison test

I specifically excluded Linux comparisons because of inconsistencies where 6850K is faster than 5960X but 6800K is slower than 5820K in the same test.

The ray trace benchmark showed inconsistences, but not the NAMD Molecular Dynamics test. Also, I'm curious to see what your excuse will be for why you didn't include these two:

I guess people really look out for a particular kind of numbers in any given data that suits their beliefs and ignore everything else(everybody, including you, suffers from this disease to a varying extent).

Well, you must definitely suffer from it big time!

 

[tamz_msc]

Yes of course, how could I have forgotten that it was actually OK in this scenario to introduce more variables into an IPC clock for clock comparison test

It seems averaging isn't your strong point - go down and see "Moyenne applicative", 120.6/116 = 1.04. So even with this one "L4" variable thrown into the mix, which according to you would make the comparison useless, the difference is still less than 5%. Now you'll complain about 4MB L3 vs 6MB L3.

The ray trace benchmark showed inconsistences, but not the NAMD Molecular Dynamics test. Also, I'm curious to see what your excuse will be for why you didn't include these two:

Now we've stooped to browser benchmarks. Great! Please tell me what the clocks were in those tests.

Well, you must definitely suffer from it big time!

Admit it, you're too much in love with the hardware you own to see past your biases. I'm still on Ivy Bridge, by the way.

 

[Carfax83]

It seems averaging isn't your strong point - go down and see "Moyenne applicative", 120.6/116 = 1.04. So even with this one "L4" variable thrown into the mix, which according to you would make the comparison useless, the difference is still less than 5%. Now you'll complain about 4MB L3 vs 6MB L3.

So you take gaming out of the equation because it suits your narrative? Much like how you ignored the NAMD Molecular dynamics test. You've already lost this argument, with that one test alone.

Now we've stooped to browser benchmarks. Great! Please tell me what the clocks were in those tests.

What does it matter what the clock speeds were? There is no way that clock speed could account for such a large margin when the CPUs are run at default speed. IPC and other architectural enhancements are the only explanation.

Intel 6850K is only 100mhz faster than the 5930K (for both base and boost clocks), yet it's 25% faster in Kraken 1.1 and 22% faster in the Google octane benchmark.

And what's wrong with browser benchmarks? Browsers are the most commonly used applications for consumers, and they are extremely sophisticated pieces of software with all kinds of technology implemented in them. You just don't want to use them because they go against your narrative.

Much like how you flat out ignored the NAMD Molecular dynamics test.

Admit it, you're too much in love with the hardware you own to see past your biases. I'm still on Ivy Bridge, by the way.

Whether I'm in love with the hardware I own or not is not the point. The point is everything I said is factual, and corroborated by hard evidence; much of it unwittingly provided by you!

I said that Broadwell has >5% of an increase in IPC over Haswell in regular code, and 10-15% and greater in FP/SIMD. This is a direct result of the architectural enhancements that Broadwell has, which are enumerated in the slide I posted above from Intel.

The fact that you stubbornly cling to your ridiculous narrative despite the evidence to the contrary, says a lot about you as a person, or you're just fundamentally dishonest. You just can't admit you're wrong, much like how you couldn't admit that CPU usage is tied to framerate, despite myself and several other posters telling you.

 

[tamz_msc]

So you take gaming out of the equation because it suits your narrative? Much like how you ignored the NAMD Molecular dynamics test. You've already lost this argument, with that one test alone.

Just like I expected - either you're fundamentally dense, or you deliberately ignore how one data point doesn't change the average picture significantly, especially when the total number of data points is relatively large.

What does it matter what the clock speeds were? There is no way that clock speed could account for such a large margin when the CPUs are run at default speed. IPC and other architectural enhancements are the only explanation.

Base/default speeds don't matter at all, how long do I have to repeat this same thing before it gets into people's heads? Based on what I know about Xeons, Broadwell-E can clock the same with more cores what Haswell-E can do with less cores. It is of critical importance to know about clock speeds in single-thread heavy web benchmarks.

Broadwell improved some things about AVX2 support and reduced some latencies with scalar and vector division instructions. It also improved DDR4 support. There's nothing significant apart from those changes in explaining the performance differences between Haswell-E and Broadwell-E, apart from clocks. On top of that, you also willfully ignore how TBM 3.0 can affect performance on Broadwell-E.

Here are the turbo bins for the 5960X. 3.3GHz. Now this is what The Stilt had to say about turbo clocks on the 6900K:

On Broadwell-E the advertized base clock seems to be during AVX2 load. In Anand review the 6900K appears to have worked continuously at 3.6GHz in non-AVX2 multithreaded workloads, despite the 3.2GHz base clock.

I wish the owners of these CPUs would dump the MSRs to find out the exact clocks they operate, depending on the utilized cores. Intel doesn't seem to provide this info for consumer CPUs newer than Haswell.

That's a 9% difference. And yet you still think I'm stupid to ask you to provide information about clocks. Geez.

 

[Carfax83]

Base/default speeds don't matter at all, how long do I have to repeat this same thing before it gets into people's heads? Based on what I know about Xeons, Broadwell-E can clock the same with more cores what Haswell-E can do with less cores. It is of critical importance to know about clock speeds in single-thread heavy web benchmarks.

I'm beginning to wonder whether you're dyslexic. I already told you the clock speed difference between the 5930K and the 6850K is only 100mhz. This is for BOTH base and boost clocks. The maximum boost clock for the 6850K is 3.8ghz, and for the 5930K it's 3.7ghz.

A 100mhz clock speed deficit is not going to account for a 25% and 22% increase in these benchmarks. Assuming the best case scenario which is perfect linear scaling, a 5930K would need to be clocked at around 4.6ghz to equal the stock 6850K in the Kraken benchmark.

Broadwell improved some things about AVX2 support and reduced some latencies with scalar and vector division instructions. It also improved DDR4 support. There's nothing significant apart from those changes in explaining the performance differences between Haswell-E and Broadwell-E, apart from clocks. On top of that, you also willfully ignore how TBM 3.0 can affect performance on Broadwell-E.

All of the changes that were outlined in the Intel slide are cumulative in their impact on performance. And don't kid yourself, you're not a CPU architect or a software engineer. You don't know the performance impact of these changes, especially as to how it relates to certain types of software.

Obviously the impact is substantial, otherwise the performance increase wouldn't be so large as to completely exceed the minor clock speed boost.

Here are the turbo bins for the 5960X. 3.3GHz. Now this is what The Stilt had to say about turbo clocks on the 6900K:

That's a 9% difference. And yet you still think I'm stupid to ask you to provide information about clocks. Geez.

You're literally grasping at straws. And while I don't think you are necessarily stupid, you are definitely slow when it comes to comprehending and accepting certain truths you don't like.

You can harp about the 5960x and the 6900K all you want, because even with the clock speed deficit and the assumption of perfect linear scaling, it still cannot account for such a large gap. That's why I preferred to use the 5930K vs the 6850K, because they are the most similar in terms of boosted clock speeds, with only a 100mhz deficit between them.

Now keep on believing that a mere 100mhz increase in clock speed is going to account for the 6850K beating the 5930K by 25%