Intel Skylake Core i7-6700K IPC & Overclocking Review @ [H]

Dan_D

Extremely [H]
Joined
Feb 9, 2002
Messages
60,420
Murder it at what? Games? No

Agreed. Haswell-E doesn't clock as well as Skylake, uses more power, and is based off an older microarchitecture with less IPC. Lets not forget, memory clocking on Haswell-E is a fickle thing and you rarely can get 3600MHz or anything like that. On Z170 its often as easy as setting XMP to enabled and being done with it.
 

DF-1

2[H]4U
Joined
Jun 17, 2011
Messages
2,645
I continue to question where people are getting these price figures. Haswell-E is going to break the bank in comparison.

I just ordered:
170 Chipset MB: $139.
16GB DDR4 3000: $139.
All that's left is the i7-6700k CPU that isn't available in North America yet.
Let's assume it's full price $350.

That's $628.

You are going to be hard pressed to go with anything socket 2011-v3 with DDR4 for that kind of money.

5820k $380
http://www.tigerdirect.com/applicat...ffiliateID=rGMTN56tf_w-3ykX_PvS8mi04mjXdK2M3g

MB $200 (or $170 after rebate)
http://www.newegg.com/Product/Product.aspx?Item=N82E16813128752&cm_re=5820k-_-13-128-752-_-Product

16gb ddr4 $125
http://www.newegg.com/Product/Produ...233752&cm_re=16gb_ddr4-_-20-233-752-_-Product

total is $675

675-628 = $47. I don't see how thats significantly different.
its available now for ordering, you have lots of reviews on boards, ram, etc so you know what you're getting, and you dont have to wait.

plus an extra two cores.
 

Burticus

Supreme [H]ardness
Joined
Nov 7, 2005
Messages
4,651
And my Phenom II 955 will keep on limping on until something exciting happens. Shouldn't we all be running 8 core i7's by now or something?
 

rinaldo00

[H]ard|Gawd
Joined
Mar 9, 2005
Messages
1,862
I am not a CPU expert so can someone tell me why [H] used an i7-4770K rather than an i7-5820K or i7-5930K? Is it a better overclocker or is it price?
 

WildMonkey

Gawd
Joined
Nov 1, 2007
Messages
887
I am not a CPU expert so can someone tell me why [H] used an i7-4770K rather than an i7-5820K or i7-5930K? Is it a better overclocker or is it price?

Well obviously we understand which processor is the fastest, but these debates factor in price, o/c potential, technologies available with the chipset, and a more subjective VALUE analysis is constructed.

Your question is along the lines of getting a fast car and you asking why doesn't everybody just buy a Ferrari?
 

jwcalla

2[H]4U
Joined
Jan 19, 2011
Messages
3,628
I guess there's no harm in waiting for the refresh and / or whatever AMD brings to the table, if anything. Maybe if Zen is half-decent Intel will actually get off its rocking chair.
 

rinaldo00

[H]ard|Gawd
Joined
Mar 9, 2005
Messages
1,862
Well obviously we understand which processor is the fastest, but these debates factor in price, o/c potential, technologies available with the chipset, and a more subjective VALUE analysis is constructed.

Your question is along the lines of getting a fast car and you asking why doesn't everybody just buy a Ferrari?

well, this IS [H]ardOCP and the i7-6700K is the top end Skylake CPU which might go $350 and the i7-5820K is $300 at microcenter ($380 elsewhere) with the i7-5930K at $500 ($570 elsewhere). Notice I didn't mention the i7-5960K, $900
 
Last edited:

undertaker2k8

[H]ard|Gawd
Joined
Jul 25, 2012
Messages
1,718
in my case 5820 k is cheaper than 6700k even with the board factored in

if u pm mc to staples, hard to argue there.
 

Wag

[H]ard|Gawd
Joined
Aug 29, 2006
Messages
1,133
So will Tri/Quad SLI setups on PLX chip on Skylake have a significant performance hit?

I'm wondering if it's even worth it, because motherboards like The EVGA Confidential look to be 8x/8x/8x/8x using PLX which isn't bad.
 

misterbobby

2[H]4U
Joined
Mar 18, 2014
Messages
3,814
And my Phenom II 955 will keep on limping on until something exciting happens. Shouldn't we all be running 8 core i7's by now or something?
I am guessing you are either in no financial position to upgrade and making excuses or either just joking around.
 

Lightworker

Limp Gawd
Joined
Jan 12, 2009
Messages
457
Anybody want to take a guess at how well a 6700K/Z170 would stack up to an old X5680/X58 combo?
 

drescherjm

[H]F Junkie
Joined
Nov 19, 2008
Messages
14,936
And my Phenom II 955 will keep on limping on until something exciting happens. Shouldn't we all be running 8 core i7's by now or something?

You can get that with the extreme edition. For the mainstream a 4 core / 8 threaded processor is enough for most. I believe Intel would have produced a 6 core / 12 threaded mainstream processor if AMD had a chip to that was better than a 4 core i7 meaning one with 5 or 6 modules. 4 modules will not cut it unless AMD could have clocked it over 6GHz stock (with headroom to overclock) to make up for having smaller less powerful cores.
 

Quartz-1

Supreme [H]ardness
Joined
May 20, 2011
Messages
4,257
The Core i7 6700K can pull about 150 watts. That's basically with an unrealistic load on it and represents a maximum worst case scenario according to my information.

Thanks. Given that different CPU coolers have different TDP limits, I wonder if it would be possible to do a follow-up article that explores overclocking vs TDP?
 

Zarathustra[H]

Extremely [H]
Joined
Oct 29, 2000
Messages
34,071
And my Phenom II 955 will keep on limping on until something exciting happens. Shouldn't we all be running 8 core i7's by now or something?
I am guessing you are either in no financial position to upgrade and making excuses or either just joking around.

Yeah, those Phenom II's were great and fun in their day, and back in 2009/2010 they could actually still keep up with most things.

Today, they are hopelessly behind though, on IPC alone, even before we start talking about how you can almost hit 5Ghz with some of the newer chips.
 

Dan_D

Extremely [H]
Joined
Feb 9, 2002
Messages
60,420
And my Phenom II 955 will keep on limping on until something exciting happens. Shouldn't we all be running 8 core i7's by now or something?
Some of us already are.

So will Tri/Quad SLI setups on PLX chip on Skylake have a significant performance hit?

I'm wondering if it's even worth it, because motherboards like The EVGA Confidential look to be 8x/8x/8x/8x using PLX which isn't bad.

The short answer is yes, there is a performance hit to 4-Way SLi on Z170 and Skylake compared to X99. Is it worth it? In my opinion the answer is a simple "no".

Here is the long answer: Skylake / Z170 will be no different than Devils Canyon / Z97 in this regard. A PLX chip adds more lanes but still has to go through the same 16 lanes supplied by the CPUs PCIe controller. It must be this way as the PCH and DMI bus are even more constrained in terms of bandwidth. This increases latency as well.

X99 combined with a similarly clocked Haswell-E CPU will certainly be faster. Most of the time you don't see a PLX on X99, but your still getting x8 lanes for each card. With Z170 your effectively only providing x4 lanes to each GPU in 4-Way SLi. Now that's over simplifying things as bandwidth allocation isn't the only consideration with such a setup. GPU scaling comes into play as cards three and four aren't taxed to the degree the first two are, the bandwidth allocation or lack of it isn't as apparent as one might imagine but the hit is there.

Still 3 and 4 way configurations are of limited value due to the often poor performance scaling. The price / performance ratio flat out sucks. If you want to spend the money is think you would want to give that configuration the best possible chance at performing well by having the resources necessary for it to do so in those few instances where 3 or 4 way configurations are actually beneficial. Keep in mind that Nvidia doesn't certify 3-Way configurations with 4x slots. Tests in previous generations showed a fairly significant hit in performance to the third GPU when using an x4 slot on AMD graphics cards compared to 8x8x8 configurations. (It wasn't something one could test with Nvidia cards.) It was still an improvement to performance but not the same degree as 8x8x8 implementations were.

Granted this is based off data I recall from an article and benchmarks I can't find right now. Nvidia doesn't allow asymmetrical lane allocations for graphics cards outside of x16 + x8 type situations.

This is probably because 8x PCIe isn't overly saturated by modern graphics cards but x4 almost certainly is. If I was going to make the investment beyond 2-Way SLi, I would only do so on the HEDT/LGA2011 platform. In short that's the only way to actually provide the real, unbottlenecked bandwidth such a configuration requires. Having done 3-Way and 4-Way configurations in the past, I have made this decision several times.
 

Zarathustra[H]

Extremely [H]
Joined
Oct 29, 2000
Messages
34,071
Anybody want to take a guess at how well a 6700K/Z170 would stack up to an old X5680/X58 combo?

Should be fairly easy. Just go to Anandtech.com/bench. Do a comparison between the 6700K and any x58 chip (the i7-920 will do)

Since we don't want to muddle in core count, lets specifically pick only the single threaded benchmarks. For these two chips the only single threaded benchmarks they have in common are 3D Particle and Cinebench R10.

Linearly interpolate by dividing all those scores by max turbo of chip in stock form, and multiply by the expected overclock.

Find the percentage for each score.

Based on this,, guestimating max x58 OC at 4ghz and max Skylake OC at 4.7Ghz my estimation is as follows (and keep in mind, linear interpolation is only an estimation)


At Stock Speeds:
3D Particle: 6700k is 54% faster
Cinebench R10: 6700K is 134% faster

Per clock (~IPC):
3D Particle: 6700K is 8% faster
Cinebench R10: 6700K is 63% faster

Max OC:
3D Particle: 6700K is 27% faster
Cinebench R10: 6700K is 91% faster

Now, this doesn't take into account how many cores each has, as it muddies the water, as turbo clocks are less predictable when loading multiple cores for the base benchmarks, so use best judgment there.

Linearly interpolated comparisons like these are imperfect, but they are the best we will get without comparing the exact setups side by side.
 

yourgrandma

[H]ard|Gawd
Joined
Apr 15, 2011
Messages
1,389
thats great,
let me tell you my experience.
upgraded from a 2600k to a 3770k to a 4970k all at 4.7-4.8 and the improvement was outstanding for every upgrade
I for one plan on upgrading to a Asus Maximus VIII with a skylake i7 at 4.7 with DDR 4 at 3000+mhz

Also when the IPC is stronger it makes emulators run alot faster. Hardocp should put up some dolphin wii /epsxe ps2 benchmarks at 2560x1440
retro games look incredible

I can tell you with 100 hours of playing/testing with the 2600k at 4.5 already plays ps2 and dolphin maxed out with no bottleneck.
 

Wag

[H]ard|Gawd
Joined
Aug 29, 2006
Messages
1,133
Some of us already are.



The short answer is yes, there is a performance hit to 4-Way SLi on Z170 and Skylake compared to X99. Is it worth it? In my opinion the answer is a simple "no".

Here is the long answer: Skylake / Z170 will be no different than Devils Canyon / Z97 in this regard. A PLX chip adds more lanes but still has to go through the same 16 lanes supplied by the CPUs PCIe controller. It must be this way as the PCH and DMI bus are even more constrained in terms of bandwidth. This increases latency as well.

X99 combined with a similarly clocked Haswell-E CPU will certainly be faster. Most of the time you don't see a PLX on X99, but your still getting x8 lanes for each card. With Z170 your effectively only providing x4 lanes to each GPU in 4-Way SLi. Now that's over simplifying things as bandwidth allocation isn't the only consideration with such a setup. GPU scaling comes into play as cards three and four aren't taxed to the degree the first two are, the bandwidth allocation or lack of it isn't as apparent as one might imagine but the hit is there.

Still 3 and 4 way configurations are of limited value due to the often poor performance scaling. The price / performance ratio flat out sucks. If you want to spend the money is think you would want to give that configuration the best possible chance at performing well by having the resources necessary for it to do so in those few instances where 3 or 4 way configurations are actually beneficial. Keep in mind that Nvidia doesn't certify 3-Way configurations with 4x slots. Tests in previous generations showed a fairly significant hit in performance to the third GPU when using an x4 slot on AMD graphics cards compared to 8x8x8 configurations. (It wasn't something one could test with Nvidia cards.) It was still an improvement to performance but not the same degree as 8x8x8 implementations were.

Granted this is based off data I recall from an article and benchmarks I can't find right now. Nvidia doesn't allow asymmetrical lane allocations for graphics cards outside of x16 + x8 type situations.

This is probably because 8x PCIe isn't overly saturated by modern graphics cards but x4 almost certainly is. If I was going to make the investment beyond 2-Way SLi, I would only do so on the HEDT/LGA2011 platform. In short that's the only way to actually provide the real, unbottlenecked bandwidth such a configuration requires. Having done 3-Way and 4-Way configurations in the past, I have made this decision several times.

Thanks for that long reply. That makes my decision even more complicated.:D

Since my gaming setup is on a 980Ti SLI setup in 4k, I have to decide whether I want to go Haswell-E for a later upgrade to a possible 3rd 980Ti or just stick to a Skylake SLI setup. They are equivalently priced.

http://www.newegg.com/Product/Product.aspx?Item=N82E16813130841

Here is a nice 3-Way SLI Haswell motherboard (at least I think it's 3-way) with all the features of a Skylake motherboard for $230. I just might do this- unless someone can convince me to go with Skylake at this point.
 
Last edited:

Zarathustra[H]

Extremely [H]
Joined
Oct 29, 2000
Messages
34,071
I can tell you with 100 hours of playing/testing with the 2600k at 4.5 already plays ps2 and dolphin maxed out with no bottleneck.

To add to that, what percentage of the [H] community even runs emulators regularly?

I'd argue maybe 0.5%?

I had a desire to replay Katamari Damacy a while back, so I loaded up the PS2 emulator, and found it to be a rather disappointing experience, not for lack of performance, but rather for how buggy it was.

Nah, going forward I'll keep my emulators for what they do best. NES games.

The reviews need to cover what we do most. Emulators are not it.
 

TDSlam720

Limp Gawd
Joined
Mar 30, 2014
Messages
366
Agreed. Haswell-E doesn't clock as well as Skylake, uses more power, and is based off an older microarchitecture with less IPC. Lets not forget, memory clocking on Haswell-E is a fickle thing and you rarely can get 3600MHz or anything like that. On Z170 its often as easy as setting XMP to enabled and being done with it.

Some games like Ryse, and a few of the racing games will take advantage of the extra COREs. Also, must note the obligatory DX12 games may make better use of multi-core CPUs. That said.. Murder -- probably not overall right now.. But.. for a few extra bucks you get the extra COREs which may be beneficial for productivity/multitasking, and some gaming. I've seen a few instances of Skylake winning in productivity benches, but for the most part this is all Haswell-Es area.

You get extra PCIe lanes which provide comfort of expandability and "future proofing". There's benches ( Guru3d PCIE scaling article ) showing 3-6% differences from PCIE 2.0 to 3.0. And yes I know 5820k can only do x16 x8, but it's still better than x8 x8.


I agree with you on the memory though. Skylake seems way easier to achieve those high frequency numbers. This is probably Haswell-Es biggest issue right now. As far as OCing goes.. I think it's a bit early to tell. I know H-es binning is inconsistant, but over all most people seem to be able to do 4.2-4.4 on air. Go AIO/Water and it seems pretty standard to go higher than that. I'm interested to see how Skylake does on Air, I think most reviews I've seen have been AIOs. We'll also see how consistent binning is when there's more of them out there in regular peoples hands.


Skylake to me just feels like a more refined mainstream Haswell-E that's a bit less trouble, but is more limited overall. I think it's a toss up between them overall, H-e/X99 just seems a bit more high end forward thinking than Skylake. I like the idea of doing a system upgrade to a hexacore and hoping future programs can take advantage of it instead of upgarding to another quad-core after a decade of having quad-cores. I'm not really losing out performance wise by making that gamble either. It's not like Haswell-Es gaming performance suffers vs Skylake on games that don't use many cores.

Plus it also helps that I have a microcenter close by to score a 5820k cheaper than a 6700k. Was able to pick up a gentle used Asus X99 Deluxe for $240. Thing that hurt was the DDR4 3200 I grabbed was 190.
 

Dan_D

Extremely [H]
Joined
Feb 9, 2002
Messages
60,420
Some games like Ryse, and a few of the racing games will take advantage of the extra COREs. Also, must note the obligatory DX12 games may make better use of multi-core CPUs. That said.. Murder -- probably not overall right now.. But.. for a few extra bucks you get the extra COREs which may be beneficial for productivity/multitasking, and some gaming. I've seen a few instances of Skylake winning in productivity benches, but for the most part this is all Haswell-Es area. As for talking about PCIe lanes, 16x8 isn't really much if any better than 8x8. This has been proven several times over the years. Where Haswell-E shines is in how many lanes it has on the on-die PCIe controller compared to Skylake.

You get extra PCIe lanes which provide comfort of expandability and "future proofing". There's benches ( Guru3d PCIE scaling article ) showing 3-6% differences from PCIE 2.0 to 3.0. And yes I know 5820k can only do x16 x8, but it's still better than x8 x8.


I agree with you on the memory though. Skylake seems way easier to achieve those high frequency numbers. This is probably Haswell-Es biggest issue right now. As far as OCing goes.. I think it's a bit early to tell. I know H-es binning is inconsistant, but over all most people seem to be able to do 4.2-4.4 on air. Go AIO/Water and it seems pretty standard to go higher than that. I'm interested to see how Skylake does on Air, I think most reviews I've seen have been AIOs. We'll also see how consistent binning is when there's more of them out there in regular peoples hands.


Skylake to me just feels like a more refined mainstream Haswell-E that's a bit less trouble, but is more limited overall. I think it's a toss up between them overall, H-e/X99 just seems a bit more high end forward thinking than Skylake. I like the idea of doing a system upgrade to a hexacore and hoping future programs can take advantage of it instead of upgarding to another quad-core after a decade of having quad-cores. I'm not really losing out performance wise by making that gamble either. It's not like Haswell-Es gaming performance suffers vs Skylake on games that don't use many cores.

Plus it also helps that I have a microcenter close by to score a 5820k cheaper than a 6700k. Was able to pick up a gentle used Asus X99 Deluxe for $240. Thing that hurt was the DDR4 3200 I grabbed was 190.

Well there are a few games that do leverage the extra cores. I never disputed or even forgot that. And while it is increasingly more common, the sad truth is that these games are still few and far between when you use GPU limited resolutions.

I'm simply stating that "murder" wasn't the right word to describe Haswell-E's advantages compared to Skylake.

On the matter of overclocking I can't speak to the 5820K, but for the 5960X 4.5GHz is usually the best you'll see on air or water. Many chips do far worse. If you get a dog you'd be better off with Skylake so long as you are only talking about a single GPU. And PCIE 16x8 isn't really much higher if any better than 8x8. This has been proven time and time again over each generation. Where Haswell-E shines, even on the 5820K is in the number of PCIe lanes it has via the CPU compared to Skylake. That was your point but your only partially correct. Two GPU do not benefit from a 16x8 configuration. Skylake has 28 compared to 16 which is why it makes more sense for multiple GPUs. With only two you won't see a difference. But you can use three or more with x99 where as with Skylake you have to use a PLX chip or AMD cards which support a 4x slot through the PCH. Neither choice is ideal.
 
Last edited:

bezant

Gawd
Joined
Oct 7, 2009
Messages
780
Great review. However, I can't help feeling a bit let down by Intel. Power efficiency, more USB ports, or even faster storage - these are all marginal improvements, esp when the GPU is the bottleneck. Where is the USP (unique selling proposition)? Where's the 6 core/12 threaded consumer CPU overclockable to 5 Ghz? After all these years, I regret not going for a 2600k, and cheapening out on a 2500k given how long it has lasted...
 

Dan_D

Extremely [H]
Joined
Feb 9, 2002
Messages
60,420
Some games like Ryse, and a few of the racing games will take advantage of the extra COREs. Also, must note the obligatory DX12 games may make better use of multi-core CPUs. That said.. Murder -- probably not overall right now.. But.. for a few extra bucks you get the extra COREs which may be beneficial for productivity/multitasking, and some gaming. I've seen a few instances of Skylake winning in productivity benches, but for the most part this is all Haswell-Es area.

You get extra PCIe lanes which provide comfort of expandability and "future proofing". There's benches ( Guru3d PCIE scaling article ) showing 3-6% differences from PCIE 2.0 to 3.0. And yes I know 5820k can only do x16 x8, but it's still better than x8 x8.


I agree with you on the memory though. Skylake seems way easier to achieve those high frequency numbers. This is probably Haswell-Es biggest issue right now. As far as OCing goes.. I think it's a bit early to tell. I know H-es binning is inconsistant, but over all most people seem to be able to do 4.2-4.4 on air. Go AIO/Water and it seems pretty standard to go higher than that. I'm interested to see how Skylake does on Air, I think most reviews I've seen have been AIOs. We'll also see how consistent binning is when there's more of them out there in regular peoples hands.


Skylake to me just feels like a more refined mainstream Haswell-E that's a bit less trouble, but is more limited overall. I think it's a toss up between them overall, H-e/X99 just seems a bit more high end forward thinking than Skylake. I like the idea of doing a system upgrade to a hexacore and hoping future programs can take advantage of it instead of upgarding to another quad-core after a decade of having quad-cores. I'm not really losing out performance wise by making that gamble either. It's not like Haswell-Es gaming performance suffers vs Skylake on games that don't use many cores.

Plus it also helps that I have a microcenter close by to score a 5820k cheaper than a 6700k. Was able to pick up a gentle used Asus X99 Deluxe for $240. Thing that hurt was the DDR4 3200 I grabbed was 190.

Well there are a few games that do leverage the extra cores. I never disputed or even forgot that. And while it is increasingly more common, the sad truth is that these games are still few and far between when you use GPU limited resolutions.

I'm simply stating that "murder" wasn't the right word to describe Haswell-E's advantages compared to Skylake.

On the matter of overclocking I can't speak to the 5820K, but for the 5960X 4.5GHz is usually the best you'll see on air or water. Many chips do far worse. If you get a dog you'd be better off with Skylake so long as you are only talking about a single GPU.
 

doug_7506

2[H]4U
Joined
Oct 17, 2004
Messages
3,247
I guess when Kyle hinted at "low" availability, he really meant no availability... :(

It doesn't look like a 6700k will be readily available until sept.
 

Rikki

2[H]4U
Joined
Oct 8, 2000
Messages
2,302
Great review Kyle. Ive got a 3 year old Ivy bridge 3770k at 4.5 and will stick with that for the time being though.

God I miss the days of massive speed jumps each year and the likes of the Celeron 300a
 

rinaldo00

[H]ard|Gawd
Joined
Mar 9, 2005
Messages
1,862
Some of us already are.



The short answer is yes, there is a performance hit to 4-Way SLi on Z170 and Skylake compared to X99. Is it worth it? In my opinion the answer is a simple "no".

Here is the long answer: Skylake / Z170 will be no different than Devils Canyon / Z97 in this regard. A PLX chip adds more lanes but still has to go through the same 16 lanes supplied by the CPUs PCIe controller. It must be this way as the PCH and DMI bus are even more constrained in terms of bandwidth. This increases latency as well.

X99 combined with a similarly clocked Haswell-E CPU will certainly be faster. Most of the time you don't see a PLX on X99, but your still getting x8 lanes for each card. With Z170 your effectively only providing x4 lanes to each GPU in 4-Way SLi. Now that's over simplifying things as bandwidth allocation isn't the only consideration with such a setup. GPU scaling comes into play as cards three and four aren't taxed to the degree the first two are, the bandwidth allocation or lack of it isn't as apparent as one might imagine but the hit is there.

Still 3 and 4 way configurations are of limited value due to the often poor performance scaling. The price / performance ratio flat out sucks. If you want to spend the money is think you would want to give that configuration the best possible chance at performing well by having the resources necessary for it to do so in those few instances where 3 or 4 way configurations are actually beneficial. Keep in mind that Nvidia doesn't certify 3-Way configurations with 4x slots. Tests in previous generations showed a fairly significant hit in performance to the third GPU when using an x4 slot on AMD graphics cards compared to 8x8x8 configurations. (It wasn't something one could test with Nvidia cards.) It was still an improvement to performance but not the same degree as 8x8x8 implementations were.

Granted this is based off data I recall from an article and benchmarks I can't find right now. Nvidia doesn't allow asymmetrical lane allocations for graphics cards outside of x16 + x8 type situations.

This is probably because 8x PCIe isn't overly saturated by modern graphics cards but x4 almost certainly is. If I was going to make the investment beyond 2-Way SLi, I would only do so on the HEDT/LGA2011 platform. In short that's the only way to actually provide the real, unbottlenecked bandwidth such a configuration requires. Having done 3-Way and 4-Way configurations in the past, I have made this decision several times.
Is x4 a bottleneck only in 3 or 4 way SLI or are modern GPUs that much faster because according to this x4/x4 was not a problem a few years ago, 2010, and that was with PCIe 2.0 not 3.0
http://www.hardocp.com/article/2010...ie_bandwidth_perf_x16x16_vs_x4x4#.VcUGXq2Ds5I

The Bottom Line
The results are actually a bit shocking to us to be honest. We weren’t so surprised that in the previous evaluation x8/x8 did not cause any differences at 2560x1600 but did at 5760x1200. However, we thought certainly at x4/x4 PCIe 2.0 mode there would be some kind of a bottleneck at 2560x1600, but the results have proven otherwise. Even with all the data that GTX 480 SLI is pushing across the PCIe bus, x4/x4 is NOT a bottleneck in a single display setup at 2560x1600 with AA enabled. The only game to show us any difference was AvP, but it did not affect the gameplay experience. Therefore, if you are on an aging PCIe 1.X system at x8/x8 mode (equivalent to PCIe 2.0 x4/x4) on a single display fear not, you are not holding back the performance of GTX 480 SLI or we guess with any CrossFireX or SLI configuration.
 

Dan_D

Extremely [H]
Joined
Feb 9, 2002
Messages
60,420
I guess when Kyle hinted at "low" availability, he really meant no availability... :(

It doesn't look like a 6700k will be readily available until sept.

According to a friend of mine It looks like you can buy Skylake equipped PCs in the US, but that's about the only way right now. If that's true (I haven't verified) then the OEMs seem to have purchased the bulk of the chips available in the US or at least a significant portion of them.

Is x4 a bottleneck only in 3 or 4 way SLI or are modern GPUs that much faster because according to this x4/x4 was not a problem a few years ago, 2010, and that was with PCIe 2.0 not 3.0
http://www.hardocp.com/article/2010...ie_bandwidth_perf_x16x16_vs_x4x4#.VcUGXq2Ds5I

Actually, I did some further checking on this. It seems that the most recent data I've found indicates that there is a small drop in FPS using PCIe 3.0 x4 slots and at least a single GPU. I couldn't find anything where SLI or Crossfire was used in such a configuration with modern hardware. I did see though that some specific games do lose a lot of performance though when only x4 or fewer lanes are used. This problem could increase with SLI and Titan X's and 4K resolutions in some or all games as the only article I found on this used a single GTX 980. The article you linked from the [H] showed a difference at higher resolutions so again, it's possible that things may change in 4K or Eyefinity/NVSurround configurations.

In any case I stand by the statement that PCIe x16/x16 vs. x8/x8 is a moot point. I also stand by the statement that a PLX chip adds latency which reduces performance slightly compared to X99. This has been tested several times and has always proven to be the case. The difference isn't Earth shattering but it's bothered some people enough that several motherboard models with PLX chips have been built with a dedicated PCIe slot for single card usage that bypasses the PLX.

I would love to see 4-Way Titan X's at 4K on Z170 vs. X99. But I'll leave such a thing to Brent unless someone wants to send me two more Titan X's.
 

Wag

[H]ard|Gawd
Joined
Aug 29, 2006
Messages
1,133
I need to upgrade pretty much now- my PC is crashing left and right so I guess I have no choice but to buy Haswell-E.
 

Dan_D

Extremely [H]
Joined
Feb 9, 2002
Messages
60,420
I need to upgrade pretty much now- my PC is crashing left and right so I guess I have no choice but to buy Haswell-E.

You could actually try diagnosing the problem. If the issue isn't motherboard related, or CPU related then you really don't need to go with Haswell-E unless you want to. If you are going to use your current situation as an excuse to upgrade then by all means. Go for it. :)
 

psyshack

Limp Gawd
Joined
Feb 15, 2004
Messages
311
Can one delid the cpu and go to the slug as it should be with a cooling solution? If not I will keep my 5930K @ 4.8 ghz ( will delid it when the enthusiast warranty runs out ) and my delided 4790K at 5 ghz and continue on with my happy geek dance.

For those that care there are some x99 comparisons out there. And the X99 based systems still are slut slapping Skylake. And with ddr4 dropping in price soon there will be no reason to not have 64/128 gig in your x99. With that ram on board for folks like me that never turn there rigs off a complete boot drive put into ram becomes a real option. Once booted off say a pcie drive and all that thrown into ram and stable. OS op's become insanely fast after a boot and data transfer to ram.

Ain't it great to be a geek now days if you have money to pay for product over priced by 80%. :)
 

nilepez

[H]F Junkie
Joined
Jan 21, 2005
Messages
11,827
According to a friend of mine It looks like you can buy Skylake equipped PCs in the US, but that's about the only way right now. If that's true (I haven't verified) then the OEMs seem to have purchased the bulk of the chips available in the US or at least a significant portion of them.

I've seen machines on Fry's website and one other, but I can't remember if it was Microcenter, Amazon, Newegg or some other random site. Whether they're shipping or not, I don't know, but it didn't say they were back ordered.
 

Wag

[H]ard|Gawd
Joined
Aug 29, 2006
Messages
1,133
You could actually try diagnosing the problem. If the issue isn't motherboard related, or CPU related then you really don't need to go with Haswell-E unless you want to. If you are going to use your current situation as an excuse to upgrade then by all means. Go for it. :)

Did that already. I already replaced my PSU- still shutting down with 980Ti SLI setup in every slot. PCI-e slot 4 doesn't even work. It's either the motherboard or one of my 980Tis and since slot 4 doesn't work with either card I'm willing to bet it's the motherboard. Ironically this didn't happen for 2+yrs with my OG 6GB Titans. Not until I installed my 2nd 980Ti, then it all went fubar. Gigabyte POS motherboard. 2 weeks past warranty.:rolleyes:

Anyways, MA tax free weekend next week. I can pick up a 5820 in store @ Micro Center for $299. The MSI X99S SLI Plus is on sale @ Newegg this weekend for $150 after rebates and discounts. All I need is DDR and a cooler and I'm all set.
 
Joined
Dec 13, 2002
Messages
1,003
Did that already. I already replaced my PSU- still shutting down with 980Ti SLI setup in every slot. PCI-e slot 4 doesn't even work. It's either the motherboard or one of my 980Tis and since slot 4 doesn't work with either card I'm willing to bet it's the motherboard. Ironically this didn't happen for 2+yrs with my OG 6GB Titans. Not until I installed my 2nd 980Ti, then it all went fubar. Gigabyte POS motherboard. 2 weeks past warranty.:rolleyes:

Anyways, MA tax free weekend next week. I can pick up a 5820 in store @ Micro Center for $299. The MSI X99S SLI Plus is on sale @ Newegg this weekend for $150 after rebates and discounts. All I need is DDR and a cooler and I'm all set.

You'll have to let us know how it overclocks.
 
Top