*Official* NBR Desktop Overclocker's Lounge [laptop owners welcome, too]

http://www.overclock.net/t/1636529/watercool-heatkiller-for-amds-tr4
More on the heatkiller block.

 

http://hwbot.org/submission/3705000_

 

If you haven't already tried, try using drivers older than 378.66 on Skydiver, for my system I lost nearly 10,000 Graphics Points from 378.66 onwards! Praps the earlier drivers are better optimised for extremely high framerates that you see in Skydiver with a strong GPU. 378.57 and 376.19 were good Sky Diver drivers.

 

Simple answer: yes, complex answer: depends and maybe not.

Simple math: 500W CPU + 300W graphics card + 40W memory + hard drives + fans and pumps + lighting = 840-900+W.

Now, it depends how heavy you are overclocking, what all has to run off the power supply, where on the efficiency curve the psu considered falls, etc. So, really need more for a good answer to this question. If you aren't going to shoot for a high OC on the graphics card or CPU, then no problem. But there is a lot of little things that can add up.

 

I agree with @ajc9988 on this. I would recommend going to at least 1200W and getting one of the EVGA PSUs with the 10 year warranty. PSU is not something to cut corners on or settle for "adequate" IMHO.

I'm barely hitting 500W on my system with 160W on the CPU and about 320W with the GPU. The 7960X is going to pull a lot more than 160W when you start spurring it in the flanks like an angry cowboy.

 

Definitely agree with that! I've got an EVGA 1600W supernova T2 with the 10 year, although some tests show smoother power delivery with the gold and platinum. The reason for additional ripple is the extra stuff needed to reach the higher efficiency at load. But I've got two pumps, 17 fans and two memory fans, couple hard drives, couple rgb lights. My CPU alone pulls over 400W. I forgot to check draw on my GPU, but it will be up there (980 ti classy). When I later upgrade, I'll have multiple graphics cards, unless no one does mGPU by the time I upgrade (likely waiting for 7nm, as Volta seems like milking at this point, as seen with the BS 1070 Ti). No reason to pay their premium when they'll price the next xx80 GPU above the current Ti, then repeat. Bad business model for a consumer to feel good at paying the asking price when they'll drop the price drastically in a couple months or launch a new product more powerful at the same price in a couple months. At 7nm, you have much more to look forward to because of release time frame.

 

Nice system @Mr. Fox, I saw the video, definitely cool. You're looking good brother. I was going to ask you why you went with a 1000 watt PSU. So even when you run SLI the 1000 will still be adequate. I didn't know the 8700K would pull 160watts. Do you think they'll have a hard time cooling that in the p870?

 

Thank you. I am having fun with it. And, thanks for the compliment. I am about 13 pounds away from a 100-pound weight loss, LOL.

I was shocked it could pull that many watts, too. But, I am glad it can because that makes it even more awesome. It would suck if it were TDP limited. As far as cooling in the new P870, I have no idea. I hope it works out well, but it remains to be seen at this point.

 

http://hwbot.org/submission/3706193_

 

That is EXACTLY the PSU I wanted and the one @Johnksss@iBUYPOWER recommended. The only reason I do not have it is that would have taken me about $200 over my spend cap, so I got the best PSU available for the money. It's a good one, but not as good as that EVGA 1600W Supernova Titanium. Being limited to a budget sucks, but the good thing is that my system was paid for in cash and I can add to or change it however I want to when I can afford it. So glad to be done with the laptop ball-and-chain that controls what we can do with upgrades based on what the laptop OEM, in their infinite lack of wisdom and greed-driven guiding principles, decides in advance for us. And, the more times I change parts, the more hardware points I will rack up at HWBOT.org LOL.

 

I definitely hear you there! I paid cash for mine, but it was spread out between last November and now, so a little easier to swallow (moving from a 6700K to a 1950X watercooled system), although a lot was recent. But, I'm still not sure if I'm waiting for Zen 2 or 3 for the next chip (might do Zen 2 to get a better resell on this chip, so it is only the intermediate cost of stepping up instead of half the value going two gens). And that freedom to change it how I want it is exactly why I went back to a desktop. Once I got burned on not being able to upgrade my ZM, I decided to hold off on DTR laptops until they pull their heads out of their backsides (or each other's executive's backsides) related to GPUs upgrading, at least, until I come back. Figure a couple years (plus, my ZM still has plenty of life now that the desktop is running). I'm still waiting for the laptops to move to socket GPUs, similar to a CPU. We've all seen the mixed AMD/Intel chip for low power. Cut the Intel part off, generally have the power components beefy enough on the board to supply the chip, resolve the HBM2 packaging issues, and you would have a nice laptop configuration where you trade out GPUs without needing the MXM3.0B socket which is being abandoned slowly. Unfortunately, Ngreedia seems to be waiting for HBM3 before it moves to it from GDDR6, which was chosen over HBM2. But that gets into speculation where things are going. Also, how to do a chip the size of TR for a GPU and what sort of cooling would be needed? (Nvidia is going to a multi-die setup after Volta most likely)

That is my Skylake ram timings to beat! Just to give you something to shoot for! I don't know about hitting my latency because of the new mesh, but the rest isn't too far off from yours.

 

Oh, if you don't have it:
https://hexus.net/gaming/news/pc/111887-watch-dogs-free-pc-gamers-13th-nov/

 

Its a good idea to get the 1000w PSU minimum. I mean, I wont use my current setup with single GPU with anything less than 1000w. Dual GPU needs 1600w IMO.

Hell, you definitely know your sh*t overclocking PC wise. I think you will do well in hwbot. I may not agree with your philosophy on BGA laptops and etc, but you are always a very nice person willing to learn and help so I am sure you will do fine!!

On that note, I may post some numbers about my new desktop soon.... Not going to be very high clock, but lots of cores

DO NOT RECOMMEND! As someone who have used GP102 and Intel HCC Die, highly do not recommend at all.

The 7960x can pull a lot of power and accounting for VRM inefficiencies, I would not put the second 1080 TI on there.

 

Here is my HWBot profile and Mr. Fox's:
http://hwbot.org/user/ajc9988/
http://hwbot.org/user/mr._fox/

Still jealous he has worked with more hardware than me and crushes me on everything graphics! CPU and ram, I can hold my own, mostly. But that should give you a better context on us as overclockers.

 

Although I reckon you're only gonna get to those kind of RAM speeds with Samsung B-dies on your sticks, don't know if Mr Fox has those. I don't have Samsung B and I couldn't really overclock at all, I could tighten timings though. Typhoon Burner can tell you the type of dies on your sticks for those that are unsure.

 

I've got B-dies! I made sure of that. Also, you don't need the speed to reach the same level of performance, mostly, if the timings can be tightened so much further at the lower speed. This is why, even though that was 4000@CL16, I ran my skylake ram at 3733@CL14 and much tighter timings. Also, having the spread of 4 ticks between the setting for CL and the timings on RCD and RP. How many people do you see do that? But the ram wants what the ram wants. I have to work within what the sticks are capable of, not make them conform to my ideal of what their timings should look like. It is like a woman, treat her well and let her be who she is, she will show you so much love; but try telling her how she should be and constantly put her down and one morning when you are drinking the coffee she made for you.... You get the point!

Edit: @Mr. Fox - My MaxxMem score from the Skylake build

 

Yep, I was pretty sure you'd had to have had B-dies!

 

Interesting that mr.fox actually took my advice, I'm suprised to say the least.

@ajc9988
How much does RAM frequency impact your threadripper build?

 

Tighter timings can impact performance more than the ram frequency. Because the frequency effects the infinity fabric, people think that you absolutely must hit the higher speeds to get the real performance from it, and that is true to a degree. You see large jumps in performance up to 3200MHz on the ram (1600MHz on IF). Beyond that, you start seeing the effects taper. But, properly tight 3200MHz ram has beaten looser timings on 3333, 3466, and 3600. So the key is to actually balance the speed and the tightness of the timings, including trying multiple speeds with as tight of timings as you can get at each speed to then see which, for sure, provide the best performance. AIDA64 isn't a good indicator of this, neither is MaxxMem. Instead, you need to do a slue of benchmarks to find what works best, specifically SuperPi, WPrime, CB15, and a couple others. Basically, if you see the score of benches fluctuate with the timing on your ram, you should be running that as you tighten and as you compare different speeds. AIDA makes you feel good, but it doesn't give the full story on how it impacts the cache performance (and timings DEFINITELY effect cache on this platform, especially L3 performance or going to the other die to pull the info, which is infinity fabric domain, or even to the other CCX).

This can show roughly equivalent performance between the two speeds on AIDA, but performance in other benches is practically identical as well. So it isn't the worry some have made it out to be. As with any system, if you are not willing to optimize for your hardware, you won't get performance like this at all. So you try to get as fast and as tight as possible, but sometimes slow and tight is better than fast and loose, as with so many things in life.

 

I've posted it before... But have you seen it?
Intel i7-8700K Coffee Lake Memory Benchmark Analysis

 

That was one of the points when I was on less, but makes perfect sense.

Also, from testing by PCPer, if mesh is used, it also sees reduced latency with speed to a degree. But...

 

I am struggling with the RAM overclocking. I'm sure it's a simple noob mistake I am making somewhere.

I used to get awesome MaxxMem scores with DDR3, but my DDR4-3000 and higher are lower than DDR3-2133 and 2400 scores. Is there something I need to do to MaxxMem to make it work correctly with DDR4? It seems to not read things correctly.

Edit: Another observation... if I look at the leader-board for Corona benchmark, it appears having more memory capacity lowers the benchmark results. Everyone with 16GB, with only a few rare exceptions, has higher Corona benchmark scores than those with 32GB or more. Not sure what to think of that.

 

Micro$oft CEO Nadella Mocks iPad Users Telling Them To ‘Get A Real Computer’

"Aaccording to Micro$oft CEO Satya Nadella, the iPad is not a "real computer." What with the thin/flimsy JokeBook's his own company have designed/developed and now throw out everywhere to the masses? Micro$oft Surface Pro series? A real computer? Take a brake. What a Moron. Stop talking BS!! All his company push out nowadays, is pure poison and should be totally avoided. And he can't see it, as it's seems And have he forgot that he's company in fact design their newest OS for tablet/phones now and not for "real" computers?

 

Which RAM kit do you have ? As in what are the default timings on that GSkill 3200Mhz kit? Is it a TridentZ or Ripjawz?

The higher capacity observation might be because, the more the modules, the less you are able to push them to higher frequencies.

 

So, you have to first remember that 2133/2400 DDR3 gives performance like 2800-3000 (or it was 2666-2800, something like that) DDR4. Changes in the ram, as well as architecture changes in the CPUs, created a situation where you had to have a fair amount of speed more on DDR4 to get about the same, even though DDR4 has some advantages. So, part of that can be discounted off the bat.

The next part comes in the changes for mesh, cache, etc. For example:


Literally, according to MaxxMem, I'm getting less performance from TR than I got out of my 2740qm CPU! Think about that! In other words, the benchmark needs retired as it no longer accurately reflects performance with the changing tech that has since came out. In fact, half the time it read my Skylake as having 150ns latency, which is ludicrous with these sticks and that CPU. So, you aren't crazy, it doesn't work properly anymore (and I have my doubts on it's usefulness for awhile now).

Now, with all of that out of the way, we can get to some of the things I did. First, I ran it in debug windows mode, and I believe I also had Asus ramdisk hosting all of the smaller benchmarks on a 4GB ram drive. Second, priority set to realtime in windows AND set as high as the program allows (there is a priority setting in the program itself). Third is timings. You are running 16-18-18-38-450 T2 at 3200. In comparison, I was running 4000 CL16 20-20-40-310 T2 for my run. So, almost identical timings (slightly looser by cycle, way tighter in ns), but with 800MHz faster. Literally, we are talking 7.5ns on the cas latency. It is pretty insane compared to 9+ns CL. So we need to work on speeding those up or tightening them down.

So, I found a utility today meant to help calculate ram timings for Ryzen. Not all settings translate, but this tool shouldn't be taken as the be all, end all. Instead, it is to be thought of as a guideline. The main thing is getting boot stable and no errors before you start tightening. Sometimes finding a boot solution is hard. So this does some of the math conversion for you and helps give a starting point.
http://www.overclock.net/t/1640919/ryzen-dram-calculator-overclocking-dram

You use Thaiphoon Burner to get the XMP settings, but go to the bottom to switch to ns. You punch in those to the calc, plus target speed, and it converts it for you. Then, use the safe or fast timings to try to boot without error. From there comes a long process of tightening timings, checking with TM5, etc. You'll also need to have AIDA64 to watch the memory and cache reactions, SuperPi to see reactions to the timings, WPrime also reacts to the changes in timings, and even CB reacts. Now, for some of those, I see more reaction on Ryzen than Intel, but SuperPi, especially 32M, is reactive to memory timings. What you are looking for is overall best performance. Sometimes going too tight, even if it isn't giving errors, can hurt performance (but it will be close to as tight as you can go and not kick errors, so you could actually tighten it all down, then loosen certain ones to see how performance reacts). Remember, some of those timings are not a min or max time, they are windows in which to perform certain operations. Too tight, you end the window too soon and the operation has to start over, thereby sucking performance out from your grasp. Too big of a window and it is still sitting there when the next item arrives, usually kicking out errors, or, if nothing else, robbing cycles.

So, hopefully that helps some (and I'll be here for more assistance if you need it on this).

Edit:

 

Let's set the record straight here. I have no idea what you are talking about, but one thing is for certain... I have not taken your advice on anything related to computers. So, I hate to burst your bubble on whatever you think it is I took your advice about, but it didn't happen. Maybe you had a lucky guess about what my next steps were going to be. I can assure you that whatever it was, you had no influence on anything I did.

16GBx2 Ripjaws V CL16.

Thanks. I will try to absorb that and ask questions if something doesn't make sense. Part of the problem may be that I haven't figured out exactly what some of the obscure RAM related BIOS settings are for. The depth of the settings eclipses anything I am used to using. Changing from XMP to Manual and reproducing all of the settings that were populated automatically from XMP as a starting point is unbootable, so I think I am setting something wrong somewhere.

I have been trying to follow this, with minimal success (thus far): [GUIDE] Skylake Memory Timings on Asus Motherboards ! - HWBOT forum

 

http://hwbot.org/submission/3707458_

https://www.3dmark.com/3dm05/5545650 | http://hwbot.org/submission/3707452_

https://www.3dmark.com/3dm06/18005647 | http://hwbot.org/submission/3707455_

https://www.3dmark.com/3dmv/5655486 | http://hwbot.org/submission/3707474_

 

I would start with increasing tRFC and push clock first on memory.

 

I just love how 3DMark Vantage pushes CPU performance... Compare to Cinbench, LOL.

 

Sure buddy, keep telling yourself that

I know you absolutely hate it when I'm right and won't ever want to admit that u were wrong.

Has anyone here pushed their 8700K to 5.4k ghz yet?
I've seen a couple of people who managed to do so on air, but they were reviewers so I think Intel might have given them silicon lottery ticket.

 

So, I ditched the pair of ugly blue 140mm fans that didn't move any air to speak of, replaced them with three better 120mm fans. Moved the 2TB HDD and 2 1TB SSDs to the back of the mobo tray. I ordered four Phanteks Halos RGB frames to go on the front panel side of those three fans and the GPU radiator fan. Those RGB frames will plug directly into the mobo RGB headers and sync with my ASUS Aura mobo lighting control scheme.

 

Some of the better fans that i love are the Noctia NF-F12's and the NF-F14's. Both 3000rpm and PWM. They do get loud, but can be easily controlled, specially in a push pull config.

 

What you are getting at is the problem of Auto-tuning! I've had it also. To deal with it, I usually try to pick a CL that should work at a given speed and put the rest on auto. If it doesn't work, then I add a couple more of the primary timings and see if it boots. That is why I put so much emphasis on a boot solution. You cannot control the algorithm(s) used for the ram. You just can't. So, when you copy over ALL of the timings from what the algo said manually, it is unseen timings and timings you didn't change that make it not boot. In other words, the algo cannot find it's own solution if you fill in too many variables.

Next, you mentioned using mode 2 for the ram. That's good! On Asus boards, they have two algos: 1) I'm going to rip off your head and **** down your throat, and 2) aggressive, but a normal type of aggressive. If mode 1 doesn't work on an Asus board, always try mode 2.

So, generalized memory timing advice:
1) don't be afraid to leave a larger gap than 2 ticks between the CL and RCDWR and RCDRD timings (RCD). I regularly need 3-4 ticks between the two (CL14 18; CL14 17). In fact, you can go even looser if you need to in finding a boot solution.
2) populate the CWL! - Sometimes, the board will use the old, tighter Cas write latency instead of matching it to the CL! That can prevent it from booting A LOT OF THE TIME. Just populating that with the same CL you are trying can save on many no boot situations.
3) The RP DOES NOT have to match the RCD timings. It can be a couple ticks looser (I usually keep it within 2, like 18 20 when trying timings, but the point is, it doesn't have to be the same).
4) RAS is normally calculated as 2*RCD or tCL + tRCD + tRP (so a range is possible). You can do 2*RCD-1 to 2*RCD+10 or more. So sometimes having that looser can help find a boot solution. Try 2*RCD+(5 to 7). That can sometimes help.
5) Keep the RFC Loose. Loosen it up a lot when starting. Like 600 or more. You will tighten it later in the process.

So, those are the first ones to populate to try to find a boot solution. Start with just the CL, then CL and CWL. Then add in the RCD, then the RP. If that still doesn't work, manually enter the RFC. Finally, add in the RAS. If all of that is still not working, then we move onto the RC, the FAW, and potentially the RRDS and RRDL timings with the FAW.

Quick Note - Setting the tREFI and tREFIx9 - When setting these two timings, it is best to be as loose as possible. But, I found with my Asus board that the tREFI fully maxed on being loose kicked errors (specifically bit flips on the memory). That means, if you are having issues, you may want to take a moderate approach to setting those while looking for a boot. Too tight and you will have issues and lower performance, or errors, but too loose can cause errors also. We will loosen them back up later in the process, so just go with something decently high, but not maxed out.

Here AMD put out a little sheet on the meaning of the timings back when Ryzen came out and people were having issues. It is a quick discussion of what the different timings are and do. https://community.amd.com/community/gaming/blog/2017/05/25/community-update-4-lets-talk-dram
They later tested different ram settings (not all applicable on Intel boards) to give a bit of guidance to people trying to OC their ram. It is similar in some ways to the post above by papusan, where the different speeds and timings were tested. https://community.amd.com/community...emory-oc-showdown-frequency-vs-memory-timings
Basically, AMD was trying to educate its users as to how to OC the ram and what to focus on to get better performance. That is how you create enthusiasts, teach them that it isn't that hard and build their confidence in playing with settings they otherwise never would touch in a million years. I'm not trying to overly sing their praises, as these skills are transferable to a large degree between AMD and Intel, but anything to help get people interested in higher performance is AWESOME!

6) Now the tRC. Here is what AMD had to say about it: "Row cycle time, or the number of clock cycles required for a memory row to complete a full operational cycle. Lower values can notably improve performance, but should not be set lower than tRP+tRAS for stability reasons." Generally, they are right. But I still try to tighten it a lot and sometimes tighten it more than this recommendation but not by much. Overall, try doing tRP+tRAS+(5 to 10) to get it booting. This, just like tRAS, acts as a window. tRAS is "the minimum number of cycles that a row has to be active ... to ensure ... enough time to access the information." So a lot of the general guidance is adding up other settings to figure out the amount of time to allow for certain actions (hence, timings). But, back to the point, the tRC effects performance and you will want to get this one low as well.
7) FAW. AMD says "Four activation window, or the time that must elapse before new memory banks can be activated after four ACTIVATE commands have been issued. Configured to a minumum 4x tRRD_S, but values >8x tRRD_S are often used for stability." Now this can get you some speed, but is often overlooked. Also, as you tighten this timing, you may need to, once errors form, try to tighten the tRRDS. When tightening that timing, you may also want to tighten the tRRDL at the same time (I found with my sticks, keeping tRRDL = tRRDS+3 worked well). This can sometimes allow you to get the tFAW tighter than it otherwise would be.
8) Now that the other timings are tightened, it is time to tighten the RFC and loosen the tREFI and tREFIx9.

After all of that work, you should be ready for a drink and to watch what comes from nice, tight timings. @tgipier - to help with your endeavors as well.

 

So this is with 16-18-18-38-1t at 3600mhz, I am running 4 x16gb so naturally the speed isnt that great compared with the 8x8gb kits.



Edit: Added Cinebench

 

Looks good, but always show the speed the ram is running at. It helps inform people. But, nice score there!

I'm noticing the L3 on those are slower than TR (by 3.5-4x slower), the L2 similar but slightly faster, but the L1 is double TR. You also have about 33% over my read and just under 20% lower latency on the ram. Not to mention 2 cores and 4 threads on me. But nice!

 

After comparing the Noctua iPPC3000 and the EK Vardar F4-120ER, the Vardars move more with the resistance of my rads! Literally, they are insanely good! Also, it works at a different pitch than the iPPC3000 and at 2200 RPM is easily comparable. I'm thinking of picking up 5 more. If you don't mind display samples, you can get them on EKs website at half the cost plus shipping (that is what they list as bulk on their website). They are the Vardar Predators. Those ****ing move air!

As an example of timings that the Utility came up with (remember, just a guide)

 

I am probably going to remove the three Vardar Furious fans from my U3 cooler and put them in the desktop once the Phanteks RGB rings get here. These wicked fans move a butt-load of air and have extremely high static pressure. I was going to use them use them on the 360mm TT Riing radiator. The problem is, I cannot use the Riing fans on the chassis because of the proprietary end connectors and the Vardar fans will not work with the Riing controller. I suppose I could just eliminate the Riing controller and manage the pump and fans using the BIOS.

https://www.ekwb.com/shop/ek-furious-vardar-ff5-120-3000rpm

Edit: Actually, the Riing RGB fan controller doesn't know if the fans are attached to the radiator or not, LOL. The cables are long enough to reach the front panel, so I will move them there and control the Vardar Furious fans mounted on the radiator using the PMW headers on the mobo. Yeah, that's the ticket.

The Vardar Furious fans are double the CFM of the Arctic F12 fans in the front of the chassis now.

 

I had the TT Water 3.0 Ultimate (also 360mm). Pretty much same but without the RGB stuff. Just threw on 6 fans and used a powered fan hub and worked like a wonder (hence the 5GHz benches on my skylake with an AIO). Didn't know they went proprietary with the Riing (glad I didn't get it). But, the AIO will be used with the skylake router once I figure out a slimmer case that can hold an eATX board. That is literally what is holding back my router build (and maybe a new psu for it, although I'm not max OCing it, just trying to have fast but stable router/firewall/squid for the LAN).

 

The controller connectors are proprietary because they have a PWM and RGB header connection on the same socket (which makes good sense). But, as noted in the edit in my post above, that doesn't matter because the cables are long enough for me to use them wherever I want to.

 

Weirdly enough when i was using the Vardars with my Predator 280 last year, they were moving about 25% less air at the same rpm than the iPPC3000 NF-F14's. But the fin density on the Predator 280's radiator is about average.
But one thing i seriously love about the Vadar's is that that they are definitely a touch quiter than the Noctua's when running at about 1900rpm.

 

See, I'm using HWlabs Nemesis Black Ice 480GTX, which have a fairly high fin density and are 55mm thick. Hell, my loop holds around 2L if you include the res. Each of the rads (3 of them) holds like .35L. Then you have tubing and blocks. But I hear ya there. I'm just looking at, with gasket, the vardar puts out about the same in pull as the ippc in push pull. But I haven't controlled every variable.

I'm thinking of, on two rads, having the Vardars pushing with the iPPC3000 pulling. I think that may increase the air flow the most.

 

This gave me goosebumps.

 

Might give your computer goosebumps from being cool, too.

On one hand he is right, because... iPad is not a "real computer" LOL. But, he is talking from both sides of his mouth. I agree about Windows 10 being ideally suited for disposable trash consumer-level hardware--like an iPad--rather than "real computers" and it is a hypocritical double-standard there. We have to remember they (Micro$lop) are pretty desperate because everything they touch now turns to crap, and there is no bottom to how low they will stoop to create an impression of success. I think it is pretty clear that their approach to business is that nothing they say needs to be true or accurate as long as they can bluff their way to the other side and stay at the top by telling lies and using deception.

 

Well, the real problem is MBAs getting involved. They are looking for too many ways to monetize their product, which includes adding "features" to spy and sell data, upsell features, etc. It really is getting bad. They could easily go back to the pinnacle of their work, Win 7, but don't. Further, they are trying to head to a subscription OS. So many will jump ship, especially with recent marketshare increase in Linux.

On Linux, look at every product sold to teach kids to code. What OS is used? Think of every mobile device that kids come into contact with and most IoT is on. Literally, the younger generation is going to go open source and there is little that can be done about it! Started with my generation with those just younger than myself and the creation of Ubuntu.

But back on M$. They ****ed up when they tried doing too much international market segmentation, charging S. America and parts of Asia HIGHER prices than even retail in the US. And you wonder why they got pirated! Just so much crap from them!

 

If by "on air" they don't mean "industrial-strength fans blowing -200c air at a heatsink the size of nevada" then that's about as believe-able as the fact that my girlfriends are Jessica Nigri, a clone of Jessica Nigri, and a second clone of Jessica Nigri

 

I would rep you a 100 times over if i could. Every single time i pointed this out in the real world, people just shoot it down. No one realizes that these short term profits are killing the future and the freedom of the consumers.
Done even get me started on the strategies they implement to maximize profits in Asia and their reasoning behind it, its just plain infuriating.

 

When I was in law school and we were talking about distributor licensing and how to secure market segmentation by region, I saw Office home and office in S.Am. for $300+ USD. WTF!

Edit: and they claim the price is justified due to piracy! If you want to make it less pirated, try making it a reasonable ****ing price for consumers in that market!

 

Some air coolers are just as good as closed loop liquid coolers with 240mm radiators, and if you delid it makes a massive difference too, so air cooling is not rubbish for CPUs, it's actually can be superior in a number of ways: cheaper, less maintenance, less to break or fail, some perform better than some closed loop liquid coolers, can be quieter. I can imagine an 8700K at 5.2Ghz on air cooling, especially if delidded, and a good chip that doesn't need massive amounts of voltage. EDIT: oh re-read, was 5.4GHz that was talked about, less sure about that, don't know how much voltage would be required, but I still think water cooling is mostly only really worth it if you go the more extreme custom loop route, otherwise I think you may as well get a good air cooler.

 

And then there is **** like this -

http://www.news.com.au/finance/busi...a/news-story/158aa7824a22e1e66611c2afb5e7a61a

XD

Their CEO literally dodged the same question multiple time when asked about it umteen times directly to the face.

 

Good 240mm+ AIOs blow air out the water, so to speak. I know top notch air coolers work fine but they're not going to encroach on something like a Kraken X62 Rev 2 or a H115i with replaced fans, probs corsair maglev ones.

I bought into the AIO is terribru for a good while until some people started running Noctua NH-D15s versus H115i with the replaced fans on Broadwell-E etc and there was a clear benefit.

But... reviewers don't delid, and 5.4 stock chip on air is about as impossible as ron stoppable taking over kim possible's show

I am punny today.

Sent from my OnePlus 1 using a coconut