Is Skylake worth the investment?

[Kevin_Walter]

Okay, I have it on decent authority that I should be getting some extra pocket-cash from family for Christmas. This might not happen, but will probably happen. I should have around a $1000 budget. Hooray.

 

I like to start shopping and planning ahead though, and I plan on using the money to upgrade my PC. Basically, the only things I'm going to keep from my current build are my case, power supply, and storage.

 

So the parts that I'm keeping are as follows:

 

Case: NZXT S340

PSU: EVGA SuperNova 750 G2

Storage

 

I'm currently running an FX 6300 on a Gigabyte 990FXA-UD3, with 8GBs of RAM, and an R9 270X. I'm looking to replace most of this.

 

What I'm focusing on right now, is the processor and motherboard. I have a current theoretical build here...

 

CPU: Intel Core i7-4790K 4.0GHz Quad-Core Processor  ($309.99 @ SuperBiiz) 

Motherboard: MSI Z97-GAMING 5 ATX LGA1150 Motherboard  ($144.89 @ OutletPC) 

Memory: Corsair Vengeance Pro 16GB (2 x 8GB) DDR3-1866 Memory  ($71.99 @ Newegg) 

Video Card: MSI Radeon R9 390X 8GB Video Card  ($374.99 @ Micro Center) 

Total: $901.86

 

Now, I'm not TOO worried about the GPU. I know for a fact I'm going with AMD. My only question there is whether or not I plan on going ahead and buying a 390x, or if I want to just toss my 270x in the new Intel rig and wait for Fury prices to fall next year. But that's something I'll decide on when I'm closer to spending money.

 

The biggest question is whether I should spend the money on this sort of "end of the line" current/last-gen build, or if I should invest in the future and plan on investing in LGA 1151 and a Skylake i5. The only issue there is that Skylake seems to be priced quite high right now, though I'm sure the prices will come down in the next couple of months. I admittedly don't know much about Skylake.. I know it requires DDR4, which is obviously more expensive, unless you get a board that also supports DDR3, which is quite rare, really. 

 

Basically I'm just asking if Skylake is worth it... in many more words.

 

EDIT: I should probably add, I know i7s aren't an incredible improvement over i5 for gaming, which is what this build would mostly be used for, but I do enjoy other activities, such as web design and some video editing. I'd rather have the hyperthreading than not, and I know the i5s don't have them, but Skylake's i7s are priced re-fucking-diculously high.

[Predrag]

Welp. since broadwell-e is coming out haswell-e will be cheaper so do consider the 5820k.

[don_svetlio]

your theoretical build is far superior to a skylake 6600K

[Sam Z Man]

The price of the 6700k currently is not worth the investment.

[MageTank]

 

Okay, I have it on decent authority that I should be getting some extra pocket-cash from family for Christmas. This might not happen, but will probably happen. I should have around a $1000 budget. Hooray.

 

I like to start shopping and planning ahead though, and I plan on using the money to upgrade my PC. Basically, the only things I'm going to keep from my current build are my case, power supply, and storage.

 

So the parts that I'm keeping are as follows:

 

Case: NZXT S340

PSU: EVGA SuperNova 750 G2

Storage

 

I'm currently running an FX 6300 on a Gigabyte 990FXA-UD3, with 8GBs of RAM, and an R9 270X. I'm looking to replace most of this.

 

What I'm focusing on right now, is the processor and motherboard. I have a current theoretical build here...

 

CPU: Intel Core i7-4790K 4.0GHz Quad-Core Processor  ($309.99 @ SuperBiiz) 

Motherboard: MSI Z97-GAMING 5 ATX LGA1150 Motherboard  ($144.89 @ OutletPC) 

Memory: Corsair Vengeance Pro 16GB (2 x 8GB) DDR3-1866 Memory  ($71.99 @ Newegg) 

Video Card: MSI Radeon R9 390X 8GB Video Card  ($374.99 @ Micro Center) 

Total: $901.86

 

Now, I'm not TOO worried about the GPU. I know for a fact I'm going with AMD. My only question there is whether or not I plan on going ahead and buying a 390x, or if I want to just toss my 270x in the new Intel rig and wait for Fury prices to fall next year. But that's something I'll decide on when I'm closer to spending money.

 

The biggest question is whether I should spend the money on this sort of "end of the line" current/last-gen build, or if I should invest in the future and plan on investing in LGA 1151 and a Skylake i5. The only issue there is that Skylake seems to be priced quite high right now, though I'm sure the prices will come down in the next couple of months. I admittedly don't know much about Skylake.. I know it requires DDR4, which is obviously more expensive, unless you get a board that also supports DDR3, which is quite rare, really. 

 

Basically I'm just asking if Skylake is worth it... in many more words.

 

You live in Ohio, are you near any of the Microcenters? They have the 6700k for $360, or the 6600k for $220. You also save $20 when you pair any of these chips with a motherboard from them. DDR4 is not much more expensive than DDR3. 16GB on average, costs $8 more on DDR4 than it does on DDR3. http://www.newegg.com/Product/Product.aspx?Item=N82E16820211905&nm_mc=AFC-C8Junction&cm_mmc=AFC-C8Junction-PCPartPicker,%20LLC-_-na-_-na-_-na&cm_sp=&AID=10446076&PID=3938566&SID=

 

PCPartPicker part list / Price breakdown by merchant

 

CPU: Intel Core i7-6700K 4.0GHz Quad-Core Processor  ($359.99 @ Micro Center) 

Motherboard: ASRock Fatal1ty Z170 Gaming K4 ATX LGA1151 Motherboard  ($109.99 @ Newegg) 

Memory: A-Data XPG Z1 16GB (2 x 8GB) DDR4-2400 Memory  ($68.99 @ Newegg) 

Video Card: MSI Radeon R9 390X 8GB Video Card  ($374.99 @ Micro Center) 

Total: $913.96

Prices include shipping, taxes, and discounts when available

Generated by PCPartPicker 2015-12-11 16:45 EST-0500

 

As you can see, if you have access to a Microcenter, you will end up with pretty much the same cost, but a newer platform.

 

EDIT: i should also add that Microcenter has the 4790k for $250. If you want to stick with Z97, it is an amazing price.

[dizmo]

I'd stick with your Haswell build, it'll be more than good enough for what you're using it for.

The build @MageTank gave is also pretty awesome, provided you can get it.

 

Why the 390x? IIRC the 390 can OC to basically the same speeds, and they get very, very close results when looking at benchmarks.

Fury isn't worth it IMO.

 

Something to keep in mind is that you'll need a CPU cooler. It's not in your parts list.

 

It's been brought up by others, but why no go X99? If you sell your older hardware it's more than within your grasp and you can upgrade to Broadwell-E later on.

Also, if you look for CPU's on Reddit's hardware swap you can get a killer deal. All Intel processors have the same warranty, it just goes by the numbers on the CPU.

[manikyath]

i'm gonna pitch something that may sound a bit backwards, but if you're gonna be buying a new ram kit either way, you might as well upgrade to a DDR4 platform.

 

something to note tho, i've had a talk with a friend about skylake parts the other day. he's planning on picking one up at a local retailer, but they go out that quickly they're back to "on order" within sometimes less as a day, which makes prices a tad unstabile at times.

 

if you can pick them up for a decent price, maybe in a deal with mobo and ram it could be a thing to consider.

 

since you come from AMD, i'm guessing you have a decent cooler on hand, have you considered X99? if you can utilize the extra 2 cores its a significant boost in performance.

 

while some people have pitched the idea of waiting for the new releases around the corner... well... there'll always be new releases around the corner.

[Kevin_Walter]

 

You live in Ohio, are you near any of the Microcenters? They have the 6700k for $360, or the 6600k for $220. You also save $20 when you pair any of these chips with a motherboard from them. DDR4 is not much more expensive than DDR3. 16GB on average, costs $8 more on DDR4 than it does on DDR3. http://www.newegg.com/Product/Product.aspx?Item=N82E16820211905&nm_mc=AFC-C8Junction&cm_mmc=AFC-C8Junction-PCPartPicker,%20LLC-_-na-_-na-_-na&cm_sp=&AID=10446076&PID=3938566&SID=

 

As you can see, if you have access to a Microcenter, you will end up with pretty much the same cost, but a newer platform.

 

EDIT: i should also add that Microcenter has the 4790k for $250. If you want to stick with Z97, it is an amazing price.

 

 

Unfortunately, the nearest MicroCenter is over an hour away and my only vehicle is a gas-guzzling SUV.

 

Also, doesn't CAS latency factor more into memory than simple speed? The more common DDR4 has lower CAS than that particular set and also significantly higher price.

[MageTank]

Unfortunately, the nearest MicroCenter is over an hour away and my only vehicle is a gas-guzzling SUV.

 

Also, doesn't CAS latency factor more into memory than simple speed? The more common DDR4 has lower CAS than that particular set and also significantly higher price.

Yes and No. CAS on DDR4 is loose by nature, simply because its very low voltage (1.2V). I have a kit with CL16 2133mhz, i took it to 3000mhz AND lowered the cas to CL15. All by going up to 1.35V. My current kit however, is CL13 2133mhz, i took it to 3200mhz CL15. You have plenty of headroom to tighten the latency.

 

Also, games do not really care about latency, not according to my tests. They mostly care about raw bandwidth. Check my signature for more information regarding that.

[Imakuni]

Yes and No. CAS on DDR4 is loose by nature, simply because its very low voltage (1.2V). I have a kit with CL16 2133mhz, i took it to 3000mhz AND lowered the cas to CL15. All by going up to 1.35V. My current kit however, is CL13 2133mhz, i took it to 3200mhz CL15. You have plenty of headroom to tighten the latency.

 

Also, games do not really care about latency, not according to my tests. They mostly care about raw bandwidth. Check my signature for more information regarding that.

Not quite. DDR4 ram has higher CAS because it also has higher frequency. Remember, CAS is measured in cycles: even though the number is bigger and therefore "worse", it is offset by the higher clocks, in such manner that the ACTUAL latency is the same and sometimes even lower than DDR3's combinations!

 

And btw, games don't care about EITHER latency or speed (with a few cherry picked exceptions).

[Kevin_Walter]

I'd stick with your Haswell build, it'll be more than good enough for what you're using it for.

The build @MageTank gave is also pretty awesome, provided you can get it.

 

Why the 390x? IIRC the 390 can OC to basically the same speeds, and they get very, very close results when looking at benchmarks.

Fury isn't worth it IMO.

 

Something to keep in mind is that you'll need a CPU cooler. It's not in your parts list.

 

It's been brought up by others, but why no go X99? If you sell your older hardware it's more than within your grasp and you can upgrade to Broadwell-E later on.

Also, if you look for CPU's on Reddit's hardware swap you can get a killer deal. All Intel processors have the same warranty, it just goes by the numbers on the CPU.

 

I generally don't OC my GPUs. It's never been worth the hassle for me. Also, the 390x isn't terribly more expensive than the standard 390s.

 

I also have a cooler. A Hyper 212 Evo. Forgot to mention it.

 

As for selling my current rig, I'm actually going to move it downstairs to replace an old shitty Dell we've hardly been using. I'd like to dump some extra storage in there over time and use it as a sort of HTPC, and something for the kids to play games on.

 

 

i'm gonna pitch something that may sound a bit backwards, but if you're gonna be buying a new ram kit either way, you might as well upgrade to a DDR4 platform.

 

something to note tho, i've had a talk with a friend about skylake parts the other day. he's planning on picking one up at a local retailer, but they go out that quickly they're back to "on order" within sometimes less as a day, which makes prices a tad unstabile at times.

 

if you can pick them up for a decent price, maybe in a deal with mobo and ram it could be a thing to consider.

 

since you come from AMD, i'm guessing you have a decent cooler on hand, have you considered X99? if you can utilize the extra 2 cores its a significant boost in performance.

 

while some people have pitched the idea of waiting for the new releases around the corner... well... there'll always be new releases around the corner.

 

Yeah, that's the thing. It's always with the waiting.

 

Also, I've never really looked into X99... ever. Are there any real obvious benefits or is it just the usual "higher numbers = better" deal?

[manikyath]

I generally don't OC my GPUs. It's never been worth the hassle for me. Also, the 390x isn't terribly more expensive than the standard 390s.

 

I also have a cooler. A Hyper 212 Evo. Forgot to mention it.

 

As for selling my current rig, I'm actually going to move it downstairs to replace an old shitty Dell we've hardly been using. I'd like to dump some extra storage in there over time and use it as a sort of HTPC, and something for the kids to play games on.

 

 

 

Yeah, that's the thing. It's always with the waiting.

 

Also, I've never really looked into X99... ever. Are there any real obvious benefits or is it just the usual "higher numbers = better" deal?

well.. if you can make use of those higher numbers, they're considerably better (higher numbers being more cores in this case)

[Kevin_Walter]

well.. if you can make use of those higher numbers, they're considerably better (higher numbers being more cores in this case)

 

Yeah, probably not. I'll probably never use the extra features of X99, such as the extra memory and multi-GPU "optimization". If it were a lot cheaper than it is, it would be a "why not?" question, but I can't really justify it currently.

 

Seems I was already on the right track, really. I think I'll pop for the 4790k when I get the money.

 

Thanks, guys.

[MageTank]

Not quite. DDR4 ram has higher CAS because it also has higher frequency. Remember, CAS is measured in cycles: even though the number is bigger and therefore "worse", it is offset by the higher clocks, in such manner that the ACTUAL latency is the same and sometimes even lower than DDR3's combinations!

 

And btw, games don't care about EITHER latency or speed (with a few cherry picked exceptions).

You can still scale the voltage up to 1.65 and reach the same cas latency you would see on DDR3 (CL11, 12, etc). People like Alex or Bullant on HWbot have some very aggressive overclocks doing it. Though, you are correct on that end

 

However, that last part is simply false. For the past few years, people have been studying the effects of memory speed's impact on gaming performance in situations with high CPU overhead. If you would actually read my signature before completely dismissing the data, you will see the evidence for yourself. Screenshots from my personal testing, Videos from Digital Foundry, and here is a link from OCN where others have been testing it themselves. http://www.overclock.net/t/1487162/an-independent-study-does-the-speed-of-ram-directly-affect-fps-during-high-cpu-overhead-scenarios

 

Trust me, the performance difference is real. If you do not believe me, test it yourself.

[Nena Trinity]

I would wait if I where you but you can get the 390X now and get rest later?

[dizmo]

I generally don't OC my GPUs. It's never been worth the hassle for me. Also, the 390x isn't terribly more expensive than the standard 390s.

 

I also have a cooler. A Hyper 212 Evo. Forgot to mention it.

 

As for selling my current rig, I'm actually going to move it downstairs to replace an old shitty Dell we've hardly been using. I'd like to dump some extra storage in there over time and use it as a sort of HTPC, and something for the kids to play games on.

 

 

 

Yeah, that's the thing. It's always with the waiting.

 

Also, I've never really looked into X99... ever. Are there any real obvious benefits or is it just the usual "higher numbers = better" deal?

 

 

Yeah, probably not. I'll probably never use the extra features of X99, such as the extra memory and multi-GPU "optimization". If it were a lot cheaper than it is, it would be a "why not?" question, but I can't really justify it currently.

 

Seems I was already on the right track, really. I think I'll pop for the 4790k when I get the money.

 

Thanks, guys.

It depends entirely on your use case. I'm going X99 with my next build because with Broadwell-E coming, I figure the 10c/20t CPU will easily last me 5 years. Since it's only $125 more than a comparable Z170 build, and extends the life, it made sense to me.

If this is strictly a gaming rig though the 4790k does make sense.

 

Hassle? It's a few sliders, you just move your mouse back and forth. It's pretty easy.

[Imakuni]

You can still scale the voltage up to 1.65 and reach the same cas latency you would see on DDR3 (CL11, 12, etc). People like Alex or Bullant on HWbot have some very aggressive overclocks doing it. Though, you are correct on that end

 

However, that last part is simply false. For the past few years, people have been studying the effects of memory speed's impact on gaming performance in situations with high CPU overhead. If you would actually read my signature before completely dismissing the data, you will see the evidence for yourself. Screenshots from my personal testing, Videos from Digital Foundry, and here is a link from OCN where others have been testing it themselves. http://www.overclock.net/t/1487162/an-independent-study-does-the-speed-of-ram-directly-affect-fps-during-high-cpu-overhead-scenarios

 

Trust me, the performance difference is real. If you do not believe me, test it yourself.

Congrats on the cherry pick (and from a forum post). The majority of the tests indicate otherwise, though.

[NumLock21]

Core i7 6700K, price gouging level: over 9000!

PCPartPicker part list / Price breakdown by merchant

CPU: Intel Core i7-6700K 4.0GHz Quad-Core Processor  ($698.89 @ OutletPC)
Total: $698.89
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2015-12-11 18:15 EST-0500

 

For the price of this 6700K, you get almost 2 5820K!

[Shakaza]

Congrats on the cherry pick (and from a forum post). The majority of the tests indicate otherwise, though.

And you have yet to provide a counterexample. Congratulations on making such an amazing argument.

[MageTank]

Congrats on the cherry pick (and from a forum post). The majority of the tests indicate otherwise, though.

Would a video suffice? I can give you a list of my games. I'll even let you pick which memory speeds i run at, along with whichever cas latency you want me to run them. This PANRAM kit is pretty great, so i am certain i can hit anything you specify as long as its within proper limitations of my VCCIO/SA voltages. It's one thing to not believe me (or the others in my post, or Digital Foundry, or several pages of people on OCN) but to not test it yourself while still calling it false? That's just silly, don't you think? 

 

Unless you have evidence showing otherwise (Don't try to use that LTT video, they completely botched their own test results), then i have to believe what i (and others) have seen with our own eyes. But hey, you could accept my offer, and let me prove it to you. I really don't mind the challenge.

[Imakuni]

And you have yet to provide a counterexample. Congratulations on making such an amazing argument.

DDR3: http://www.anandtech.com/show/7364/memory-scaling-on-haswell/7

DDR4: http://www.anandtech.com/show/8959/ddr4-haswell-e-scaling-review-2133-to-3200-with-gskill-corsair-adata-and-crucial/6

http://www.tomshardware.com/reviews/memory-bandwidth-latency-gaming,3409-6.html

 

Would a video suffice? I can give you a list of my games. I'll even let you pick which memory speeds i run at, along with whichever cas latency you want me to run them. This PANRAM kit is pretty great, so i am certain i can hit anything you specify as long as its within proper limitations of my VCCIO/SA voltages. It's one thing to not believe me (or the others in my post, or Digital Foundry, or several pages of people on OCN) but to not test it yourself while still calling it false? That's just silly, don't you think?

Sigh.... if I said that I DID tested it myself and reached the conclusion of ram being irrelevant, would that change anything?

 

Also, I'm not saying the tests are wrong. I'm saying that you chose one of the very few instances where RAM matters (though that tomb raider test is a bit weird, as anandtech's 4 different tests all showed the same numbers).

[MageTank]

DDR3: http://www.anandtech.com/show/7364/memory-scaling-on-haswell/7

DDR4: http://www.anandtech.com/show/8959/ddr4-haswell-e-scaling-review-2133-to-3200-with-gskill-corsair-adata-and-crucial/6

http://www.tomshardware.com/reviews/memory-bandwidth-latency-gaming,3409-6.html

 

Sigh.... if I said that I DID tested it myself and reached the conclusion of ram being irrelevant, would that change anything?

 

Also, I'm not saying the tests are wrong. I'm saying that you chose one of the very few instances where RAM matters (though that tomb raider test is a bit weird, as anandtech's 4 different tests all showed the same numbers).

It would matter to me. It would show me your test methodology, and i would be able to discern whether or not you were in a scenario with high CPU overhead. It would also add more data to a subject that is currently lacking in data. More data is always helpful to everyone. 

 

Also, i did not cherry pick my tests. I actually ran the rests 3 times, and the results were very consistent. I still have them saved, along with their time stamps if that will help ease your mind. The margin of error within each test was on average, 1fps difference, with the highest difference being 2fps in minimum FPS between tests of the same memory speed. 

 

If you look at my thread, you also see that i noted absolutely no difference in performance on Metro 2033, and even negative performance on Shadows of Mordor. If i was going to lie about something, i certainly would not have included that data. However, they were the only AAA titles with built in benchmarking tools that i had, so i used them. The data that came from them was going to be used whether it proved my point or not. If memory speed had no impact, i would be saying it didn't. I do not get paid by memory companies, and i certainly do not benefit from telling people one way or the other. I do it because I personally noticed a difference, and because I want to help people get more performance for their dollar. 

 

I also want to point out that in my results, my ram is not running XMP. I've overclocked my kit thoroughly. From Primary timings all the way down to Tertiary timings. One thing you have to consider when overclocking ram, is that when you go to a higher speed (Say, going from 2133 to 2400mhz) your RTL's and IO-L's have to change with it. When these become looser, you actually LOSE memory bandwidth. For me, 3200mhz CL15-15-15-30-CR1 is faster than 3400mhz CL16-16-16-32-CR1, because with the 3400mhz setup, i cannot lower my RTL's enough to make up for the difference in clock speeds, and my overall bandwidth actually suffers from it. By leaving RTL timings on auto, and switching between XMP profiles in these tests, you are actually running the risk of not improving your bandwidth enough per clock change, and it will skew the results. In my tests, i kept the Tertiary timings as tight as i possibly could at every single speed stepping. Only leaving the Primary and Secondary timings the same. 

 

Another reason why I would have loved for you to test it, so i could evaluate your testing methodology. All we know so far on this subject, is not every game or hardware configuration will benefit from faster memory. HOWEVER. We know that certain games and hardware scenario's do, and the price difference between 2133mhz DDR4, and 3200mhz DDR4 (8gb kits) are only $10. My personal tests showed an 8% difference in minimum frame rates on average. $10 more for 8% more minimum FPS? That is a worthy investment for me. If it's just a placebo, hey, I am out $10. It'll come in handy if i ever decide to get into heavy compression though, so it's still win-win. 

[Prysin]

 

You live in Ohio, are you near any of the Microcenters? They have the 6700k for $360, or the 6600k for $220. You also save $20 when you pair any of these chips with a motherboard from them. DDR4 is not much more expensive than DDR3. 16GB on average, costs $8 more on DDR4 than it does on DDR3. http://www.newegg.com/Product/Product.aspx?Item=N82E16820211905&nm_mc=AFC-C8Junction&cm_mmc=AFC-C8Junction-PCPartPicker,%20LLC-_-na-_-na-_-na&cm_sp=&AID=10446076&PID=3938566&SID=

 

PCPartPicker part list / Price breakdown by merchant

 

CPU: Intel Core i7-6700K 4.0GHz Quad-Core Processor  ($359.99 @ Micro Center) 

Motherboard: ASRock Fatal1ty Z170 Gaming K4 ATX LGA1151 Motherboard  ($109.99 @ Newegg) 

Memory: A-Data XPG Z1 16GB (2 x 8GB) DDR4-2400 Memory  ($68.99 @ Newegg) 

Video Card: MSI Radeon R9 390X 8GB Video Card  ($374.99 @ Micro Center) 

Total: $913.96

Prices include shipping, taxes, and discounts when available

Generated by PCPartPicker 2015-12-11 16:45 EST-0500

 

As you can see, if you have access to a Microcenter, you will end up with pretty much the same cost, but a newer platform.

 

EDIT: i should also add that Microcenter has the 4790k for $250. If you want to stick with Z97, it is an amazing price.

 

good suggestion

 

BUT

 

need 32-64GB of memory.

 

but why?

 

FOR THE GLORY OF RAMDISK OFC!!!!!!

[MageTank]

good suggestion

 

BUT

 

need 32-64GB of memory.

 

but why?

 

FOR THE GLORY OF RAMDISK OFC!!!!!!

Primocache all the things!

[Imakuni]

It would matter to me. It would show me your test methodology, and i would be able to discern whether or not you were in a scenario with high CPU overhead. It would also add more data to a subject that is currently lacking in data. More data is always helpful to everyone. 

 

Also, i did not cherry pick my tests. I actually ran the rests 3 times, and the results were very consistent. I still have them saved, along with their time stamps if that will help ease your mind. The margin of error within each test was on average, 1fps difference, with the highest difference being 2fps in minimum FPS between tests of the same memory speed. 

 

If you look at my thread, you also see that i noted absolutely no difference in performance on Metro 2033, and even negative performance on Shadows of Mordor. If i was going to lie about something, i certainly would not have included that data. However, they were the only AAA titles with built in benchmarking tools that i had, so i used them. The data that came from them was going to be used whether it proved my point or not. If memory speed had no impact, i would be saying it didn't. I do not get paid by memory companies, and i certainly do not benefit from telling people one way or the other. I do it because I personally noticed a difference, and because I want to help people get more performance for their dollar. 

I was referring to the link you posted, not the one in your sig.

 

I also want to point out that in my results, my ram is not running XMP. I've overclocked my kit thoroughly. From Primary timings all the way down to Tertiary timings. One thing you have to consider when overclocking ram, is that when you go to a higher speed (Say, going from 2133 to 2400mhz) your RTL's and IO-L's have to change with it. When these become looser, you actually LOSE memory bandwidth. For me, 3200mhz CL15-15-15-30-CR1 is faster than 3400mhz CL16-16-16-32-CR1, because with the 3400mhz setup, i cannot lower my RTL's enough to make up for the difference in clock speeds, and my overall bandwidth actually suffers from it. By leaving RTL timings on auto, and switching between XMP profiles in these tests, you are actually running the risk of not improving your bandwidth enough per clock change, and it will skew the results. In my tests, i kept the Tertiary timings as tight as i possibly could at every single speed stepping. Only leaving the Primary and Secondary timings the same

Oh, ffs, 2nd and 3rd timings? Really??? Sure, if you are going to go THAT deep on tweaking, go ahead and get the most ballz to the wallz hardware you can.

 

But for the other 99% of the people, that is NOT a representative test. Just like benchmarks aren't accurate reproductions of real world games, playing with anything over primary timings is NOT a good representation of the auto settings.

 

 

Another reason why I would have loved for you to test it, so i could evaluate your testing methodology. All we know so far on this subject, is not every game or hardware configuration will benefit from faster memory. HOWEVER. We know that certain games and hardware scenario's do, and the price difference between 2133mhz DDR4, and 3200mhz DDR4 (8gb kits) are only $10. My personal tests showed an 8% difference in minimum frame rates on average. $10 more for 8% more minimum FPS? That is a worthy investment for me. If it's just a placebo, hey, I am out $10. It'll come in handy if i ever decide to get into heavy compression though, so it's still win-win. 

Do you want a consistent methodology or a real world one (with stuff like antivirus, music, skype, and other things that would be what would actually happen for everyone else, but completely screw up with results)?

 

Personally, I vote for the second, but I'm pretty sure I'm a minority here.