Anybody got a Skylake HQ/HK CPU yet? I want guinea pig-- I mean testers.

It should beat it in cinebench as well, though your 3720QM can only hit 3.8GHz on all 4 cores at its max OC bins. Your 4GHz bin is only for single-thread.

 

OP is updated with a link to Throttlestop Beta 6.

 

3720QM at 4ghz (so 3.8 on 4 cores)
3dmark physics: 8641
Cinebench: 644 (after some time it throttles down i believe due to running out of the turbo power boost)
Passmark cpu: 9659

3720QM stock
3dmark physics: 8269
Cinebench: 595
Passmark cpu: 8519
ts bench: 277.629

6700HQ
3dmark physics: 7856
Cinebench: 679
Passmark cpu: 8539
ts bench: 277.49

 

Yeah, your cinebench should be better for the 3720QM. I believe my cinebench at 3.5GHz here was 692 or so on my haswell? 3.8GHz on ivy bridge should be different.

There's also the possibility of RAM having a big factor in it, and less of a factor in 3D Mark physics (which would make more sense by far). Even the passmark CPU makes more sense.

 

taken from: http://www.anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation

please be aware that those are IPC changes, i.e. cpu performance based at identical clocks on all five architectures!

Sent from my Nexus 5 using Tapatalk

 

Thanks to everyone who provided some response to my post#202. But that is not what I want (and I have seen those anandtech articles which make no sense to me because I won't buy a new platform and limit it (or overclock it) to get a 'clock for clock' comparison.

I'm asking for any performance/productivity improvements in your actual day to day workflows on a mobile Skylake platform w/DDR4 (@ 2400MHz or better) vs. any older mobile platform.

Benches mean nothing in the real world and even the ones above show discrepancies within themselves as others have noted.


Ideally, I would like something like:

'I use my system to do 'X' and with my SB/IB/Haswell/Broadwell system it took me 45 minutes to complete a loop of my daily workload and the fans would be maxed out and the system noisy and throttled. With the Skylake platform (and ideally, everything else being equal w/regards to RAM, O/S, Programs and Workflow), the system finishes 'x' minutes faster, with no/little throttling and much quieter too.

Benchmarks are fine for the online rags - but I don't buy anything based on synthetic benchmarks with artificially imposed limitations and/or other handicaps...

I hope to be able to do my own internal testing for mobile Skylake at the end of this quarter or early Q2, but any comments today in the form above would be great to know beforehand.

 

Spoiler: Book

If what you want is a stock v stock comparison, that's going to be ridiculously hard to find. Though I'd be glad to run a bunch of tests if I had a skylake notebook compared to my P370SM3 for you.

The point of clock for clock comparisons is that it can generally be calculated how much better or worse things are, in general, without being handed a system to manually check. But you want DDR4, which means it's an uneven comparison which isn't in those articles I suppose.

I'll give you some generalizations here.

The i7-6700HQ is a straight sidegrade or downgrade from any haswell system running decent DDR3 (RAM with decent timings, unlike the 1600MHz 11-11-11-27 RAM I have) that has a CPU speed of 3.4GHz or above.

If you can get the 6820HK or 6920HQ at high speeds (3.9GHz or higher) without overheating or power throttling in your notebook of choice, coupled with some decent RAM (DDR4 2400MHz cl14 for example) it'll likely be an upgrade in performance from any previous CPU that is NOT a: i7-29*0XM, i7-39*0XM, i7-49*0MQ or i7-49*0MX

The 29*0XM chips likely will be near the limit of their cooling for OCing to combat a 4GHz skylake chip, needing to be about 4.6GHz or higher.
The 39*0XM chips will hold their clout a bit longer, but you'd need at least 4.5GHz to beat a 4GHz or higher skylake chip on average.
The 49*0MQ chips, being capable of hitting 4.2 or 4.3GHz on 4 cores, will hold their own against a 6920HQ in compute capability, and their RAM potential is better (2133MHz cl11 RAM is much better than 2400MHz cl14 RAM), however the cooling system required for holding those clocks may be unreachable without a modified Clevo or an Alienware 17 R1/AW18, both of which have issues with current GPUs, and the AW18 has problems with 780M SLI as well using drivers beyond ~345.20. So those are not systems I would require heavily.
The 49*0MX chips can pass the 4.2-4.3GHz limit that the 49*0MQ chips have, however again, cooling them is nearly impossible without heavy modding, and the limit of cooling for 24/7 usage with heavy workloads (even with CLU) is achieved near the same 4.3GHz mark or so, which makes a potential 4.2GHz 6820HK (maybe in a GT72 to cool it?) a superior performing CPU.

On the other hand, the i7-4790K in the P7xxZM machines would more than beat any of the mobile skylake chips, due to the cooler being designed for the heat and overclocking. But if considering that, then the 6700K in the P7xxDM and P870DM machines are also very worthy considerations, and with their potential to get even faster RAM at the cl15 range (like DDR4 2666MHz or even DDR4 3000MHz at cl16) combined with the cooler skylake CPUs, I'd say they're even better equipped for your purpose.

If considering mobile CPUs ONLY, then you'd need something with a beefy cooling system like a GT72 to do the heavier workload performance at high clockspeeds.
If considering STOCK mobile CPUs only, then an i7-4810MQ/4900MQ or better will beat any stock skylake mobile CPU on the market. The i7-3840QM (3.6GHz 4-core boost) should match or beat all but the 6920HQ CPU (3.4GHz 4-core boost).

I hope the above book gives you enough insight into what to expect. Please note that while I generalized a lot and gave stock and overclocked comparisons to great detail, some programs will like improved RAM a lot, and some programs might even like Skylake's cache system better, etc. There's a few variables that don't show up in clockspeed and IPC considerations alone, but those are usually rare and it's hard to tell which programs love them. For example, 3DMark firestrike's physics test *LOVES* Broadwell's L4 DRAM cache and gives it a gigantic boost over haswell and skylake in those tests. But the DRAM cache doesn't really do much for most other programs.

 

Thanks for the effort; but I don't want to guess on what I might experience in the real world.

It is easier to see if a user's reported workflow matches mine or not (or at least; how closely it matches) instead.

I have stopped guessing and estimating performance gains for years. But yeah; in general - the newest the better (overall; heat/noise/efficiency/battery life/etc.).

No problem; I'll do my own testing as usual when the time comes. But if anyone with any real world experience can contribute here, it would be most welcome.

While synthetic benchmarks and their 'analysis' creates great 10K+ word articles for click bait, they offer little relevance to a productivity/ROI focused buying analysis for anyone except (maybe) others that benchmark for a living too.

 

Well that's the thing. There's too many variables. If an end user says they had an i7-2920XM and it was at stock with DDR3 1333MHz cl10 RAM and then tried an i7-6700HQ with DDR4 2400MHz cl14 RAM, the skylake *WILL* perform better. There's no doubt.

On the other hand, a user with a 3920XM at 4.3GHz with DDR3 2133MHz cl10 RAM will demolish almost anything you can glean from a skylake system, better cache etc or no.

What you need to give is your current machine, and the planned machine (and whether or not you wish to OC) for a good response. Either someone can copy the scenarios for you and test, or I can tell you off the bat which is better or worse based on said benchmark information you don't bother with. And of course HDD or SSD is a factor. Otherwise, you NEED your own testing and nobody else can test for you.

 

By adding synthetic benchmarks; the variables only increase - they don't clarify anything at all.

While I can almost agree 100% with your first two examples; that is the thing... I want to know by 'how much'. And for the OC'd 3920XM setup... I would simply discount it as OC'ing is not what I do with any of my systems. Not that it wouldn't potentially 'win' from performance perspective - but because OC'ing and it's hassles (heat/power/efficiency) is a known downside to my workloads, productivity and reliability requirements.

There is no benchmark that will give me what I want (except to do the testing myself; i.e. 'my' benchmark(s)...) and that is that with this new platform my workload will be 'x'% faster or, alternately, my workload will be completed 'x' minutes faster.

Thanks for the offer, but I too can (and have) read all the standard reviews done by all the standard online rags.

 

then theres just going out there,buying the machine urself and comparing with ur current setup and workloads. no one else can do that for you, unless u pay someone to do it, otherwise its just a waste of time for them unless they just happen to have the exact same workloads as u do and the exact same setup u have now and upgraded to a newer system

thats exactly what benchmark tools are for, they do not increase the amount of variables but reduce them. clock for clock comparisons between cpu archs gives u info on the IPC differences.based on that its just simple arithmetics to calculate actual stock vs. stock differences between cpus of different gens and with different clockspeeds. that, combined with ddr3 vs ddr4 tests will give u a pretty good average of expected performance gains for any given hardware.

Sent from my Nexus 5 using Tapatalk

 

but the improvements are marginal. Maybe 10-20% max difference, nothing like the improvements made in the 1st, 2nd and 3rd gen i7's. So it wont even be necesarry to know its exact preformance, higher CPU preformance is not worth upgrading for. The only real improvements are made in the iGPU and the low TDP processors.

 

Ran another test in case the over-TDP time limit was just increased for Skylake, since the longest I'd tested was like 4 minutes.
Looks like it can hold 75+W for well over 20 minutes! There's a couple brief instances of thermal throttling because one of the cores fancies itself a free thinker and wants to run 10C over the rest, but other than that, no throttling whatsoever.

Spoiler

 

Thanks very much @Katiecat! I'm wholly satisfied that the 6920HQ can handle its TDP, at least for your machine.

If I can find one more 6920HQ, even if only in a GT72, that can hold its TDP, even if Clevos and Alienwares don't allow it, I'll know the chip itself is capable by design, since I'm still convinced by the extremely rare Haswell chips that could hold 70W constantly that some chips seem to "slip under the radar" or something. But I'm glad to see what you have. One more 6920HQ and I'm set, especially if it's from another machine and can be overclocked.

Seems all the 6700HQs are indeed locked. I wish I knew whether or not the 6820HQs were limited-OC or not, but there's no reason to purchase that chip with the 6820HK in existence (there's actually no reason for the 6700HQ to exist, since the 6820HQ is the SAME PRICE, 100MHz faster and has 2MB more L3 cache).

 

I see people are still missing the point. I can adjust my expectations based on what others report - except for benchmarks that correlate to nothing but other benchmarks (and if then...).

Even if someone posted their experience with my exact workload - I'd still be testing for myself before I committed to actually handing over any funds...

An example of how insufficient benchmarks are; if we were to believe them (and of course, which aspect to believe in too...) we should be seeing order of magnitude improvements over HDD's when using our SSD's.

But that is not the case by a long shot. At least not in the workloads I'm immersed in (and yes; they are heavily storage subsystem dependent).

You may think that benchmarks make it easier to compare different systems because they are done in a controlled and consistent manner. But that is not how the world works. If my competition can use/find/invent/steal a better workflow than I to get ahead; they will.

That is the difference; what I am after is how will a new component/platform/etc. will increase my productivity in the real world (i.e. decrease my time for a given workload) - not whether I have bragging rights to the fastest (theoretical/on paper) hardware today.

I sincerely appreciate everyone's input. But would appreciate concrete examples even more.

Now, back to your regular programming.

 

Spoiler: Boooook

Well this is the thing. Benchmarks are designed to test either a specific workload (like 3DMark, or AIDA64, or Cinebench) and that alone. For example, a "blender benchmark" would be perfect for checking various systems' performance in Blender. And you can check different PCs (in their entirety) during that. But the reason benchmarks like the ones linked are "normalized" is so that they narrow down variables to as few as they can; ideally one variable. When this variable is down, known and documented, we can run other variable tests. With enough single-variable benchmarks, we can get down to what makes a difference where. When you know what makes a difference where, you can, without the help of testing a system manually, know within a small margin of error what the best thing for a job is. This is the purpose of these benchmarks, though as-is, I've not seen enough of them, but that's because my level of attention to detail is far beyond what most people (and thus those websites' testers) care about.

Now, you refuse to list what your workloads are (with some good reason, I will bet), and without that I can't help you. Maybe your workload is related to certain RAM speeds. Maybe your workload is related to overall RAM performance; similar to most programs in the world (I.E. faster speeds but comparably slower timings will equate to an equal or negative performance). Maybe, like x264 live encoding, the speed of the RAM (not necessarily the latency) is important, and similar performance RAM (1600MHz CL8 vs 2133MHz CL11, for example) will give the nudge to the 2133MHz RAM purely because 2133MHz and above helps x264 encoding. Maybe you need a large pagefile on a SSD or even multiple pagefiles to get things really working. Maybe Broadwell's 128MB L4 DRAM cache that does not exist for Sandy/Ivy/Haswell/Skylake will be the best benefit for you, and even a comparatively "slower" broadwell CPU will actually still improve on performance purely due to that cache.

Now I'm calling out a bunch of random numbers here, and most likely none of it means anything to you, or even applies to your workload much. I wouldn't know. But this is the point of the benchmarks. When you can single things out (say, Physics-based rendering, or ray tracing, etc is beneficial on a CPU with the L4 DRAM cache, but nothing else benefits) you can look at your workload and judge whether or not things would be beneficial. Then you have a smaller subsection of things to test when you are indeed ready to test. I want to be clear though: under no circumstances am I trying to state that you *should* do this kind of information filtering using available benchmarks. Maybe it's too much time and effort for too little benefit. But this is what benchmarks that are "normalized" are for. A benchmark in itself is a static workload capable of being run on various machines to test their performance. If you had say.. a "test project" you could transfer between machines to run through your productivity workload, and you ran that same project on every machine to see which one would benefit you the most and whether or not you should upgrade, you're doing a benchmark. When you do your testing, you'll likely do benchmarks of your own =D.

I just keep getting the feeling that you've been misunderstanding the purpose of a benchmark with your desire for purely real-world testing, when real-world testing is best done using benchmarks of some sort. There is a "general feel" that can't be shaken (like say... "oh, my new PC feels slower than my last") but that doesn't really quantify anything. But yes, good luck with your testing. For your sake, I hope you find something that gives you a really nice benefit.

 

You and I are saying the same things; but we still are not agreeing on how beneficial BM's are (at least to me).

The main issue is that a real world workflow is not a single, narrowly defined 'workload', rather, it is a combination of many tiny 'workloads' that gives us the final outcome desired. Trying to guess how 1,000 narrowly defined workloads will interact with my overall workflow is an exercise in frustration. As is attempting to breakdown my overall workflow so that I can map it to the tiny ones offered all over the web (why? Because my workflow is not set in stone; it is fluid and constantly changing as I learn new processes/tricks and/or software updates/upgrades make them available to me). In a general sense, it can seem the same overall (from the cpu+ram viewpoint), but I have been around enough that I know it sometimes doesn't matter what the theory may predict; it could be the exact opposite (with regards to my overall productivity).

D2 Ultima, thank you for continuing this discussion with me. I do know about and understand all your points you make. They are just not enough for me to even begin to base business decisions on.

If Intel were to hang their reputation on just these narrow workload/BM's (like Samsung does with a lot of it's products, including SSD's), they would have lost my business decades ago.

Instead, with each new platform I have tested properly in house from them, they have proven that regardless of what the benchmarks may say, productivity increases are still obtainable.

As I may have mentioned before, a $400 MB with appropriate RAM, CPU and PS+Case is a small investment to make if I even achieve a 'mere' 5% productivity improvement, and, sometimes even much less (i.e. my competitors are already onto the new platform...).

It is easier to wait until I:

1. Have the time to properly setup a test system (i.e. find the best combination of components that will make it sing).
2. Have the time to properly optimize it (i.e. find the best combination of settings for the O/S, Program Suites and any tweaks available).
3. Have the time/energy to do the testing fairly between the new test setup and my current platforms.

At this point, the 'better' system is given to me in minutes saved. I can now use simple math to multiply by the number of workstations I will commit to upgrading/replacing today and know when this technology investment will be paid off (based on the work I have scheduled, of course).

What I hoped for in asking my question here is help with #1 (best combo of components).

Again, thanks for trying to help and for letting me explain myself a little.

I will continue to follow your thread (silently) and see what I can gleam from it.

Ty.

 

Hello everyone. Don't think I've been here for 5-7 yrs lol. Finally upgrading Gateway P-7811fx , T9900 (which was a later cpu upgrade) lol. Anyway, I think 6820hk is for me. But I have one little problem...I can't seem to easily see what laptops you all are using ! I know there is some AW, prob some MSI. But I'm doubting that the Sager 8678 I have in mind even has any BIOS OC settings, and is meant to rely on software only. Anyone know? Any of you that could kindly tell what brand and model you are using, what's in the BIOS, and just approx what stable OC you're getting would be greatly appreciated! I would be ok with 3.9 even, but very satisfied with 4.0 everyday, no overvolt or anything.

 

Mine is a Clevo P650RG. I literally haven't OCed at all because of current limit throttle

 

hmm thanks, yours should be really similar to the Clevo P670-RG I'm considering. Umm pls pardon my ignorance..where is this throttle you speak of? I assume you haven't flashed a custom bios, and I'm thinking these models might be a bit new for one to be out yet? So if you try to run intel extreme tuning utility XTU, it won't do it?
BTW I think I'll go for the PCIe SSD : ). (prob 950pro) Hey you know, a guy at Sager told me the 2400MHz mem was unstable, so they took it off the compatibility specs, you know anything about that? Thanks for ur rig tag..if the other posts here had that it would help me.
One thing I'm wondering...is about hardware/MB support for OC...don't know if that's a factor here too, and if any brands are better suited for that. MSI said they have an OC page in the BIOS, so that's something, though I havn't found any screenshot of that yet. If anyone has those please link Thanks!

 

So.. why am I leaning to Sager 867x instead of MSI GT72 ?..strictly because of about 1.4 pounds or so... since I want the 6820HK. The lighter MSI's like the GS, GE apparently don't support the chip...only the GT. Which leads me to wonder ....does the lighter Clevo 'fully' support it, or as much as the MSI GT? Re" GPU, I don't think any of these are MXM, and MXM as a future proofing thing seems like non-starter anyway..though I havn't a complete picture yet. I kinda..like the idea of an external GPU amp, but I would want that to be DIY option since I bet they suck, like the AW I've read sucks, and I bet the new Razor one is no better. Would be fun if there was a hack to put at GPU outside like that! Or better if there was a port meant for that purpose of course.

 

Clevos will support the 6820HK. Even if stock sBIOS and EC don't, @Prema is a magical phoenix that will make it support. Love him, he is magical. And phoenix-like. And somehow takes hikes for fun despite my certainty of his ability to fly.

That being said, @Spikes042 have you attempted asking for a new BIOS/EC from your reseller, and/or checking of Clevo's website has an update? The P870DM had random current throttle poop for quite some time from what I remember and a BIOS/EC update fixed that.

Alternately, have you tried increasing your current limits via XTU or throttlestop? XTU lists it as ICCMax and throttlestop should just say "PP0 Current Limit" under the TPL button's popup window.

 

Most of the CLEVOs are quite similar in specs.
Yes, I haven't flashed a custom bios. Prema currently doesn't even have it so I might have to wait a while.
XTU works but the system throttles because the it has passed its set current limit. I need a custom bios to remove/up that limit so the system can use as much current as is available.

I didn't need the PCIe SSD as it wouldn't be of any use for me. I don't do any I/O intensive stuff that would benefit from such speed. I never knew about the ram problem.

 

I haven't asked my reseller, would do that now. There's no BIOS listing on the clevo download page.
XTU doesn't work, still does the same thing. Will try with throttlestop

 

@Spikes042 Probably can change that without a BIOS mod. If your interested post / PM a link to your BIOS backup and I'll have a look.

 

I can't find one, so I guess it probably doesn't have an update yet... but officially, it should be fixed sooner or later. I know that rampant current throttle isn't by design.

I am certain it's a problem with the EC rather than the BIOS, going by the P870DM's problem and subsequent fix.

 

Can you suggest any BIOS backup utility?

Let me keep my hopes high then. Thanks for all the help

 

@D2 Ultima

if you still need help, i got an asus gx700vo with an intel 6820hk that i can throw around for you if you'd like ;-)

 

4ghz..... on all 4 cores, can the Asus do it..??? lol

 

Of course I do! I want from all vendors =D

 

haha yah actually ^^ no clue yet for how long though ;-)
out of the box those ****ers blasted the cpu with 1.355v (!!!) to make sure those 4.1ghz are stable...unbelievable!


Gesendet von iPhone mit Tapatalk

 

cool, so tell me what you want to see ^^


Gesendet von iPhone mit Tapatalk

 

Everything is all on the first page.

 

ah damn, alright i will go check. that's the prob when you get into a thread through google and don't even look at the startpage ;-)


Gesendet von iPhone mit Tapatalk

 

lets see how this Asus performs under pressure......

1.355v is one hell of a vcore ....on a 14nm chip..?? no..?? mr ultima will know!!

 

totally crazy..temps also skyrocket with that much vcore. i run it at 4ghz using 1.18v now, temp difference is huge. also degrading of the cpu with 1.355v must be quite something.


Gesendet von iPhone mit Tapatalk

 

so handbrake i got 4ghz constant. max tdp was 50.9W..max temps around 70degrees c

prime95 temp is around 77deg. c max, 56W max tdp, cores trottle down to 3.4 and mostly hang around that speed =P --> powerlimit trottling according to xtu..limit reasons somehow is greyed out in throttlestop and starting it manually from the folder brings it up but it won't display anything.

ts bench 1024 max tdp 40.9W, 65 deg. c max temp, 4ghz stable.

xtu benchmark temp max 75 deg. c., 56tdp max, cores mostly at 4ghz, sometimes shortly dive down to 3.8 and come right back up again. xtu says power limit throttling

those are the settings i use in intel xtu:

http://www2.pic-upload.de/img/29494368/Unbenannt.jpg

 

You haven't bumped your power limits via XTU. Try setting turbo boost power max and short power max to something like 100W and then testing with XTU benchmark and P95 again. If you managed a lengthy test holding 56W constantly already, hopefully it will be straightforward. It's good to see the 6820HK actually holding its stuff in multiple laptops, at the least. I hate BGA but if they work my complaints have taken a steep dive.

Also, what is the max OC you have on the 6820HK? It seems like the 40x mark is quite near to the max multiplier for that machine.

 

That was re-written to use MSR's and the same library as TS. Maybe those MSR's have changed in Skylake, idk. Perhaps you could run the attached and zip back the "dump.txt" file so I can check, preferably run after you have experienced throttling.

Thanks.

 

alright, thx for your feedback guys, will try all of your suggestions later on when i get home (as i'm back to work now ). will report back


Gesendet von iPhone mit Tapatalk

 

Just remember to quote me so I get an alert. And pictures of your testing mid-test I'll also need, if you want a screenshot-taking program you can use lightscreen and then upload them to imgur.

 

@D2 Ultima @Dufus

so i figured i'd quickly have a look inside the notebook. just e glimpse. shortly later i put it back together and had repasted gpu and cpu with liquid ultra... it's always the same story xD
so i upped the turbo boost power max (65W) and short power max (75W). those are the highest settings, that's the end of the slider.
ran xtu @4GHZ and it never dropped down, went right through with 4ghz. max temp: 54c!! ma tdp: 53W

screenshot xtu

now on to prime 95! will report back.

edit: so i rn p95, had to up the voltage in xtu from 1.180v to 1.190v cause i had a bsod. after i upped it it was ok so far. 4ghz straight through (well..3.999 at some points xD). max tdp: 62W. max temp: 72

screenshot p95

 

Sounds good to me. Though a 65W max for that notebook seems... so low. Ugh.

 

yeah it does. hmm, but so far it's performing quite nicely, so i will not ***** too much ;-)

 

It's probably Asus's throttling crap in this notebook due to the dock.. Do you have the dock, deadbydawn?

 

yah, i do ;-) same limits with or w/o dock attached.

 

you lot are really starting to annoy!!

hearing you all go on about 'BGA JUNK'..... do you think the guy who just entered this thread to submit his Asus machine for testing really wants to hear how crap his machine is b/c it has soldered parts..... it shouts of jealousy tbh!! (you cant afford to jump on the BGA bandwagon, whether you want to or not..?? at least thats how you're coming across)

if you dont have anything constructive to add here.... go haunt some other thread!!

you had a pop at me over on the AW forums cos i said i could overclock and you told me it was bull **** and i had a BGA pile of junk which couldn't hold past TDP or OC on more than 1 core..... we've provided endless evidence that it DOES overclock and hold passed TDP

why not start a thread dedicated to ppl who hate BGA ..... you can all polish each others 'e.peen' over there..

unbelievable!!!


to the guy with the Asus.... nice results, yet another success for BGA and overclocking!!

 

Okay, enough fighting everyone. Please and thank you.

I quite hate the lack of choice and the forcing of BGA as much as you do @Papusan and @TomJGX or maybe even more, but I hate them a lot less if they're all working better than Haswell and Broadwell and high performance computing, even if not as good as what the Clevos with a Prema mod and a 4790K/6700K can get. It's something that was lacking before, and now I'm glad people have working equipment.

@skunk1 I understand why you got ticked off, but I wanna be clear. Nobody... NOBODY. On this forum. That considers themselves a hardware or performance enthusiast. Is "sad" that they "can't afford BGA". You were not around when BGA was first forced on us. The first and second BGA-only waves of laptops (the second half of Haswell's lifespan and Broadwell's full lifespan) could barely hold stock clocks most of the time, had wildly varying voltages (1.3v and 1.4v at stock for 3.3GHz quadcores were "uncommon" in rarity) and were very power limited. It's like the saying "blood runs deep"; just because there are some changes on the top doesn't mean people forget so easily. And I think people forget TOO easily in the gaming and tech industry... not enough fire is applied to companies when they screw up; the consumer generation has gotten rather complacent. Granted, I do agree they went too far in bashing what appears to be a perfectly working system (albeit overpriced), hence why I first addressed them. But it by no stretch of the imagination means the BGA chips something to be lauded.

The point of this thread was to make sure the information in my mobile CPU guide is up to date, and accurate. I'm happy that the BGA chips can hold their clocks and overclock about as well as the MQ chips before them could, but that's just it. They're simply showing the ability to perform like what we had before, this whole time, except they're STILL soldered to the board, to save the cost of a socket that isn't transferred to the end-user. In other words, it saves NOBODY but Intel/ODMs/OEMs money (it trickles down, they all get some savings). Sockets should always be used if possible, but if we're stuck using soldered hardware, it should at least work properly. I'm glad it does, since there's no choice for users considering mobile chips. But even if socketed chips come back, I hope the soldered ones that inevitably remain will continue to work as well.

 

haha @Papusan and @TomJGX are at it again lets not drive away the bga test subjects guys

greetings from the gym on my new highend bga hardware@huawei mate 8, now thats some bga id recommend to anyone!

Sent from my Huawei Mate 8 using Tapatalk

 

thanks i was just about thinking that may be why not a lot of people contribute here haha j/k ;-)

well i am actually quite happy with the machine. i know it's hella expensive, but i like that it holds an oc of 4ghz (haven't really tried 4.1 yet as i'm happy with the resulty plus temps according to it) which i just wasn't able to accomplish with my aw17-r2 before, no matter what i did.. (don't know if the r3 is different). i do not like soldered cpus and gpus either just because i liked to play with stuff and also upgrade it. but since i have a quite hefty desktop (it really really really is hefty performance wise.. ;-) ) my main goal was to get a powerfull notebook which is not as thick as all other "bricks" out there..since i bring it to work sometimes and i don't want a 5 cm high brick in front of me at a meeting with everyone staring at me like "dafuq is this pre-world-war-2 machine haha i needed portability and performance and that's what i got :-D