What to upgrade for 144hz? 4670k GTX 1060

[jerubedo]

@GoldenLag, never mind, it wasn't Tech Powerup, it was PC Gamer:

 

Similar story to a lesser degree:

 

[GoldenLag]

2 minutes ago, jerubedo said:

The Anand is October 2018 and Tom's was November 2018. Not terribly long ago, and not far off of Steve's December results. There's also data from Techpowerup that agrees with Anand and Tom's, 1 sec. That would be 3 against 1. 

Both of older datasets.

 

Great to see more benchmark data. 

 

But then there is the discrepancy of having similar performance characteristics but at a lesser degree. Something GN showed between his older vid and his newer. 

 

But for the purpose of this thread neither the 8700k or the 9700k is the right choice with Zen 2 right around the corner

[jerubedo]

Just now, GoldenLag said:

But for the purpose of this thread neither the 8700k or the 9700k is the right choice with Zen 2 right around the corner

Well yeah, that I can agree with. Everyone will come out a winner if people wait. Ryzen might be the best new thing. And if not Intel prices will drop or a response will come. Win-win for everyone.

[GoldenLag]

1 minute ago, jerubedo said:

@GoldenLag, never mind, it wasn't Tech Powerup, it was PC Gamer:

 

Similar story to a lesser degree:

 

These are all 97th percentile results. Which shows dips but not stutter like 0,1%

[GoldenLag]

1 minute ago, jerubedo said:

Well yeah, that I can agree with. Everyone will come out a winner if people wait. Ryzen might be the best new thing. And if not Intel prices will drop or a response will come. Win-win for everyone.

More or less.

 

We will have a fundamental disagreement on HT vs non-HT where i will be rather concerned due to historical data with core i5 parts stuttering. 

[jerubedo]

1 minute ago, GoldenLag said:

These are all 97th percentile results. Which shows dips but not stutter like 0,1%

This one is, yeah. Anand and Tom's were 1% lows, though. We still only have one data point for the 0.1% lows but as shown he has very different results from the 1% lows, so I'm not sure how much stock we can put in it. And again, my own personal data from TODAY conflicts with his results as well.

[GoldenLag]

2 minutes ago, jerubedo said:

This one is, yeah. Anand and Tom's were 1% lows, though. We still only have one data point for the 0.1% lows but as shown he has very different results from the 1% lows, so I'm not sure how much stock we can put in it. And again, my own personal data from TODAY conflicts with his results as well.

Yup. 

 

Which then mostly leaves 4c/4t that have matured enough to be deemed consistant. 

 

Which is honestly the bigger convern rather than games right now.

 

But at least we can both agree "wait for Zen 2" > "250$++ CPU"

[jerubedo]

I just got a bit of extra relevant information from my friend over at Intel pertaining to CPU usage percent with hyperthreading that I figured I'd share here since it's relevant to the conversation on the whole.

 

So apparently CPU usage monitoring applications and Task Manager alike aren't great at estimating real CPU usage percentages with chips that have Hyperthreading enabled because there's no real way to accurately measure the load that's actually being put on the core. So from what he said, when you see a hyperthreaded chip that has 2 threads (that share the same core) at ~65% load, that core is more like 95% - 100% loaded and won't be able to crunch much more than it is. Likewise when you see all threads at 65% you can safely assume that all cores are 95% - 100% loaded. At this level of load, the threads generally aren't fighting over resources to any extreme extent.

 

When you see all the threads at 100%, they are now in a situation where there is fighting over resources happening to an extreme extent. In these cases, each thread is only operating at ~65% efficiency. If each thread were actually working at 100% with 100% efficiency it would mean gains of 100% for HT vs. non-HT, which is obviously not the case.

 

Hence, when a chip is loaded at 65% on all threads, it won't perform very differently from a chip loaded at 100% on all threads.

 

The 65% is a generalization for optimal performance gains of ~30% (and the math does indeed work out. All threads at 65% being the equivalent to 100% load on the cores is the equivalent to 30% gains in performance on the whole).

 

He did note that there are some VERY rare workloads (all productivity related) in which ~70% usage would represent 100% on the physical cores (for a mathematical gain of 40% gains in performance). I've personally yet to see a situation like this but /shrug.

 

So basically when you see a HT chip being reported at less load % than a non-HT chip being reported at a higher load %, the difference is far less than it seems.

 

This was sort of a revelation for me, because while I did of course realize that when all threads are at 100% load, it's at a lower efficiency, I never thought about a reported 65% load being equivalent to a reported 100% load.

 

@GoldenLag I figure you'd probably find this interesting as well, if just from an academic standpoint.

[GoldenLag]

24 minutes ago, jerubedo said:

@GoldenLag I figure you'd probably find this interesting as well, if just from an academic standpoint

that is fairly interesting. not that i trusted the readings of taskmanager much. why taskmanager would report like that is very much reflected in how HT and SMT would operate. 

[jerubedo]

3 minutes ago, GoldenLag said:

that is fairly interesting. not that i trusted the readings of taskmanager much. why taskmanager would report like that is very much reflected in how HT and SMT would operate. 

Agreed. He had said it was all reporting apps, though. Not just task manager. He said there was no real reliable way to get a reading on the core's actual load % since the OS is only aware of the threads and it doesn't know what efficiency they'd be running at in any given moment.

[Beowulff83]

so you want this

21 hours ago, Glarity said:

8700k or a 9700k

to pair with this

21 hours ago, Glarity said:

GPU: MSI 1060 6GB Armor

i stand by my first statement

21 hours ago, Beowulff83 said:

upgrade to a used i7 4790k?

if you are looking at upgrading your graphics card soon, by all means go with 8700k or 9700k.  but if you are planning on keeping that 1060 for awhile i would just upgrade the cpu on your current rig