top of page

Why I say no to Ivy, Haswell and Skylake CPU'S?

Many people think as they consume less energy 79 watts, 88 and 91 Watts they will run cooler they take it and find them self’s in a situation to find a new CPU cooler thinking it will solve their problem well it is not!!

 

Older Intel Chips had the CPU core directly attached to heat spreader with a fluxed solder as you can see here:
That solder had heat transfer conductivity 80 W/mK so the heat root to your cpu cooler was:
cores -> 80 W/mK solder -> Heat Spreader -> 5-6 W/mK Thermnal Paste -> CPU cooler.
All Ivy, haswell and skylake CPU's they have the cores burred even deeper into a box and they use plain thermal paste as you can see here:
​That thermal paste has thermal transfer condactivity 5 W/km, so the heat root to your cpu cooler is:
cores -> black box -> 5 W/km -> Heat Spreader -> 5-6 W/km 'better' Thermal Paste -> cpu cooler.

 

As you can imagine the road that leads the heat to your cpu cooler is now bigger but also 16 times thinner from the older cpu's!! This is WHY ivy, haswell and skylake cpu's running hotter and in many diffent web sides they found +20 C hotter in stock frequencies​.
​These cpu's are running hotter by the hour and is sure someone with Ivy, haswell and skylake can't run a 2 hours stability test of the machine like this:
Prime95 28.5 torture 4 threads 4096 Mbytes minimal fft (I use half the CPU cache in small tests always) and 6 Gbytes memory usage (-2GB always from total 8Gb to activate all the dimms) while at the same time Furmark 1.8 xtreme burning was running. Notice the 423 watts HEAT LOAD as the machine is overloaded with 6 satta ii in raid 0, Ocz drive, etc from the UPS the monitor is not connected, the test started in 13:17 am and ended 15:17:10 am  (Greek 1:17 μμ to 3:17:10 μμ) look the furmark.
The CPU (Q6600 105 watts) is running nonstop for 2 hours oveclocked at 3.3 Ghz and none of the cores are above 60 C, the bios shows 54 C to the CPU, 36 C to the System, the ambient temprature is 26,8 C. The cooling system is 3          fans for intake, 1 front, 2 by the side, 1 to the rear and the heat also goes out from 120x120 PSU. The cpu cooler is            and if you see more carefull only the 140x140 of            is connected to the onboard power fan connector, it spins at 1250 rpms. Everything else is directly connected to the PSU, always Noctua fans all over the box, there is no substitute.
Ivy, Haswell and skylake cpu's they can't have temprature termination above 105 c. With the above heat root they have a problem running even a simple 1 hour test of prime95 28,5 torture 4 threads, all CPU cache 8192Mbytes minimal FFT size and half the ram without to get 90 C to the cores (stock speed) and when Tcase max is 72c-74c only allowed (they will burned faster because they are smaller) ​and way above 105 c to the system!? If your motherboard has japanise solid capacitors are fully operational at 105 c, above that you risking malfunction, hardware failure, that lowers the power delivery all over the motheboard the volts are droping, making even the coolers that are connected onboard to run in lower rpms the situation get's even worst when the CPU starts to throttling, missing cycles, and with it all the components onboard start suffering.
With the cores at 90 c for long time you have the risk the glue that used to attach the heat spreader to start evaporating inside the chip! More problems can be created of this heat starts spreading in back of the board and though the copper all over the motherboard for those who doubt such a thing can happen I have these pictures taken by a thermal camera:

 

 

 
As you can see at the first image the heat spreads to a surface all over and covers everything, always running faster to the cooled area, you can also see the stable heat flow to the heat sinks. When the heat source stops, the other 2 images, the cooling effect seems to appears more random due to surface microstruction inperfections, some small edges that are not detectable by the eye, and the air movement that is not 100% predictable.  
Many crazy solutions can be found to the internet about cooling these chip removing the heat spreader and applying 'better' thermal paste directly to the cooler! You can't win nothing this way the road to the cpu cooler seems sorter:
cores -> black box -> 5 W/mk -> cpu cooler
but the attached sureface now is even smaller!! The old root to the cooler may seems bigger: ​​​
cores -> 80 W/mK soldier -> Heat Spreader -> 5-6 W/mK Thermnal Paste -> CPU cooler.​

 

But the 80 W/mK root looks so BIG as 17 ivy and haswel highways now!? Some people even try to melt down the black box that the cores  are burried deep inside, even indigo extream has 20 W/km heat condactivity can't help, is not effective as that 'small' solder did! See the heat road:​
cores -> black box -> 5 W/km -> Heat Spreader -> 20 W/mk indigo 2 xtream -> cpu cooler.

 

The plain thermal paste is the Bottleneck that makes even indigo extream useless!? An 'old' cpu is sure will drop the core to the ambient temprature at idle if indigo applied:
cores -> 80 W/mK soldier -> Heat Spreader -> 20W/mK Indigo 2 xtream  -> CPU cooler.​

 

Notice the attached surface is always BIGGER from any possible solution you can find by removing the heat sreader and  losing the warrante. The 'old' CPU runs -5 to -6 C cooler now. An ivy CPU at 4.2 ghz, 1.31v a​ consums arround 300  watts a haswell 4.4 ghz, 1.296v arround 155Watts and skylake (gets unstable above 80C no matter how much you push the voltage) with indigo 2 xtrem you get only -1.5 c to them it's wasted!?

 

Most websites they dont touch the heart of the issue. Why Intel didn’t use the same material of fluxed soldier attached to all of the cores driving the heat out faster and keeping the core temperatures to the lowest point? Is not hard Imagine the oc capabilities above 9 Ghz if they do such a move on air. They use a black box that offers no thermal advantage conductivity to cover the cores even deeper! More strange is they didn’t even bother to spread this thermal paste all over the back of the heat spreader!? It would be much more effective this way.

 

What is behind all these 'mistakes'? nobady can tell for sure, but i forced Suggesting to the users that they want to upgrade now the sandy bridge processors i7 3820 or i7 3930K for LGA​2011 motherboards, at list they will not change motherboard again to upgrade to the feature incoming intel chip's, smart or not, i feel beeing used as a mopet for the intel's stock that dosen't move! Not like i7 2500k and i7 2700k that are more popular to the public.
 
Bottom line is i have to turn down these cpu's as unusable to any type of enviroment i moved to Opteron AMD and there is no substitute, trust me.  
 
 
Better late than ever Intel forced to fix new processors fully soldered only Haswell-e i7-5960X, i7-5930k and i7-5820K and you will be able to have 4 Ghz and 5 Ghz in really low temratures using the right equipment. Enerdyne did relesed  the new                    with 40 W/km that can be aplied in the new chip (a new kit needed the old one can't be used in there...) so is safe to move if you like but void water cooling they are damaged from the over heating and the liquit can run into your motheboard. The 4 way architecture to the ram by plasing the dimms from each site of the motherboard can spread the heat from side to side (if you install 4 dimms) much faster and the grapchis card is also connected to the CPU!? Only NH-D14 extents under the cooler to cool the RAM and the componets with really strong air that is really close to the motherboard is a must any other choise will not be so effective.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Look these pictures taken from an Infrared camera to study the phenomenon of PWM 'cooling' the stoke speed is 3.3Ghz at 1.1850V (bios settings) and in idle consumes 17 watts and 75 watts at full load.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
The pc is just started, the motherborad is in idle mode, the temprature to the heat pipes is 30.80 C  and the chipset behind the cooler has some overheat spots that they are at 80 C and 70 C, if you are not experienced in infrader images you can't see it. The heat starts to spread all over the motherboard slowly but steady until you finally read 80 c to the chipset inside the box and as you see in this picture:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
The chipset from a different angle is stable finaly at 80 c and the components arround are at 60 C and 70 c!? The test starts the cpu is running stable at 40 c (reading at the bios) with fans of the cpu cooler finaly at the maximum speed the temprature to the heat pipes don't go above 34.42 C as you can see here:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
The ram is also cooled but this can't stop the heat stread all over the motherboard from behind that is now more rappid as you see spots with 75 C, 79 c, 80 C and 100 C, If you put your fingers in there you will burned!? So even when your CPU looks cool and is soldered the heat spreads and start to cover everything like it or not and what do you think all of the capacitors that are connected to your motherboard with the voltage regulators are all fully operational at 105C? You can see a 1500 Watts thermaltake PSU in here:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
From my experience and is not so difficult and for you to imagine they are really good you can read the specifications here: 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Thermaltake is saying is fully operational from 0 to 40 C and above that they can not guarranty that will work right!? If is used to drag the air from inside the box and the cooling you have is not so good you will hear it working afther you closed the computer to remove the heat out.
 
Some people don't get it even if you draw them a schematic:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
The ultra durable 4 Motherboards have voltage regulators on board that can't operate in more than 40 C as you see above and only the ultra durable 5 motheboards have regurators that can work at 60C yet some people claim some tempratures above 70C are safe!? They did put a double inner layer of copper to spread the heat faster away from the chips and they have better electromagnetic protection you can read the full article in Gigabyte site:
 
Was these enough to explain why some people get so many voltage anomalies in their readings as you see here:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
As you see better in the infrared the liquid cooling that I said so many times to avoid using it has nothing to blow cool air to the regulators and close to the motheboard!? The 4 ways architecture to the ram by placing the dimms by eatch side creates a wall from boath sides that making difficult even for NH-D14 to cool them if operates in low speed!? You can see the voltage regulatros and the motherboard burning as the temprature is above 105 C even when the CPU appears to be cool in 43.68 C!?
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
In order to  understand better why the FX-8370E appears more cool and stable from 5960X you need to see these motherboards:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Keep in mind the first one                            is Ultra durable 5 and the voltage regulators are operational in 60C is what you need to make it to 5 Ghz 'without' a problem but as can you see if 2 dimms installed they create a wall for the 2nd voltage regulators to cool down even from NH-D14!? This is the BIG problem but unlike other motheboards all the HEAT sources are connected with HEAT PIPES and this is helpful if you make top exhaust mounting for the CPU cooler (a smaller one is needed in order to do it) to cool down the CPU main voltage regulators that are between the wall of 2 dimms. These kind of problems are not exists in the second motherboard                         but it's Ultra durable 4 and the voltage  regulators are fully   operational only in 40C. Both have what I marked with red circle 3 x 12 volts the for PCI-E connectors by connecting them from the beggin you making sure the PCI-E card will not eat the fan out from the CPU if 4 sli cards are connected. I suggest to connect them anyway so with these 2 motherboards to be more possible to make it at 5 Ghz 'without' any problem and without to increase the voltage much keeping in mind what is forcing you to do it's the HEAT and the electromagnetic interference. In the test ASRock X99 WS was used as you can see it here:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
That has also 60 C operating voltage regulators but the HEAT beat them anyway in stock frequency!? The final conclusion is the FX-8370E can be overcloked in much higher frequencies always with better stability and temperatures from Intel. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
If the above wasn't enough to make you understand why the OC to the 'new' architecture can faild you must also see this image:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
The voltages to the haswell-e are coming from Vccin pin and VRin is the voltage that is getting and splits into Vcore for the cpu voltage, Cpu ring for the cache voltage, System Agent Voltage that is the pci-e and memory conroller, all these wasn't part of the CPU in the 'old' architecture and I never understood why they did this as the temperatures and the voltage needs inside the CPU are HIGHLY INCREASED without any REAL reason!? This is why I suggest VRin 2.0 Volts, vcore 1.3 volts and CPU ring 1.150 Volts for the 4.5 Ghz. All together are 1.3 + 1.150 = 2.45 Volts needs and the instability that may occures if the CPU dosen't have enough voltage for one reason or the other increase the VRin a bit more. The 5960x consumes 271 watts at 4.5 Ghz so you must have a really strong PSU in order to OC in higher frequencies. The pci-e card if it dosen't have enough power from the PSU connection may draw power from the motherboard and the extra connections in the Gigabyte motherboards are helping you if something like this  happens not to effect the CPU power needs. I forgot                   is a must as you apply it only once  and it works for a life time.
 
In the bottom line if your calculations are not right your OC may also failed as 1.89 volts VRin leads to a more stable situation for the stock frequencies in assrock motherboards (I'm joking  =;-)
 
For those that they fail to realize what happened this is the Ga-P35-DQ6:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
You see how LARGE heat sink had but also a 12 volts fan connector on board to install a small fan on top. From the back of the motherboard also had a LARGE heat sink known as crazy cool system as you see here:
 
 
 
 
 
 
 
 
 
 
That was cooling the CPU and the P35!? That was the memory and the pci-e card controller also known as south bridge and All this + one Voltage Regulator got inside the the CPU and nobody askes WHY this is happened? Just to BURN you much faster or to stop you to do 9 Ghz on air and for no other reason happend, we don't speack, we don't speack but I had enough. 
 
This is the new idea with the air acceletation channels:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Noctua saids "Boundary layer separation from the suction side of the fan blades leads to increased vortex noise and lower airflow efficiency. In order to suppress this phenomenon, the Noctua’s A-Series impellers feature suction side Flow Acceleration Channels. These channels alter the flow distribution on the suction side of the blades and speed up the airflow at the crucial outer blade regions. As flow separation is more likely to occur when the speed of the fluid relative to the fan blade is low, the increase in speed achieved through the Flow Acceleration Channels leads to significantly reduced flow separation, which permits lower vortex noise and higher airflow efficiency."
 
The big question is for how long? They will get dusty and if you smoke they will damaged much sooner and the result will be not to work in the end!? You need to remove them and to clean more frequently, this is why I reject even NH-D15 with this configurtion. The fans with the cuts in the wings you can have them running for 7 years and to clean them only once is more than enough. 
 
Let's cut the crap the GA-990FXA-UD7 I have above is out of stock from many places and some greedy people ended selling them in 1,259.98 and 2,228.97 dollars I have contact Gigabyte and I asked to stop them this is ridiculous!? 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
The truth is nobody with that type of motherboard will go to forum and ask why he has poor performance, lost frames, BSODS, as 2 x copper size inside the pvc can guaranty better thermal removal and higher quality of power delivery but is not for that price!? The most people choose AMD for the low price they get a poor quality motherboard and they get into trubble faster, is good to report these people to Gigabyte.  
 
                                                                                                                               
 
Some people don't get it even I draw them a schematic!? Following the tradition from 2007 with GA-P35-DQ6 the new GA-X99-UD7 is Ultra Durable 5 and as you see here:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
always had the better thermal removal protection to offer as ALL the HEAT sources are connected with HEAT PIPES and you don't need extra fans in the chipset or any other place to cool all the parts down, all is needed it to raise the the last heat sink in the rear above any other dimm (the 4way architecture that came and make all the simple problems extreamly complicated) to extent it and get out of the metal plate or to do a top mounting of the CPU cooler and 10nx to the HEAT PIPES to all the other parts will also cool down automaticly!? I strongly disagree with the wifi connection that offers as more electromagnetic infrearence  added without any real reason and that is also dangerus and to human body!? I don't even use the lan conntectors on board is far more better to use x1 Tp-Link card with the biggest receiver and transmitter buffers but also the biggest Jumbo Frame you can find linked with cat 6 SFTP ethernet cable that offers the best electromagnetic protection for send, recieve always with the maximum speed in long distances. The second 1 x M.2 Socket 3 connector is working in 10Gbytes unlike the GA-X99-SOC Force that offers 20 Gbytes connection on it but  you can call this is the motherboard I always like. In 2007 the GA-P35DQ6 had 12 power regulators to the CPU and why the got so low in the new builds I never understood!?  I always mark the 3 x 12 volts connection for the PCI-E cards on board that can help you make wonders with RED cycle. 
 
Finaly Gigabyte mobilized to the right direction GA-990FXA-UD7 3.0 is                    with all the capacitors fully operational in 105 C and the power regulators working in 40 C without any problem. They upgrade the bios of this greate motherboard to a new one that you can download from here:
 
 
 
Using the latest utilities that you can find here:
 
 
 
Don't hesitate to communicate with Gigabyte and report any anomalies you can find.
 
Personally and if you ask my opinion I do not wasting time with none Gigabyte products see again the GA-990FX-UD7:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
In top and rear mounting the most problems are already solved and you don't need to install the back plate of the motherboard for the air to be released and from there faster. NH-D14 is more than enough with Indigo Xtreme: 
                                              
 
to drop the tempratures of the Fx-8370E -5 C in 4.5 Ghz will run arround 40C on Prime95 28.5 torture CPU + FPU all cache your problems are solved for a life time. Just make sure if you use NF-P12 for intake all over the box to use double filters for the air the dust can create a blanket that covers everthing and you will be fine. The Only problem I have is how Gigabyte will control the market prices as I found GA-990FXA-UD7 to be sold in Italy for 793.94 euros!? The best solution is Gigabyte to buy a distribution company and allow us to order directly from them.
 
In the 'new' architecture for Z170 chipset they stopped using the ram from both sides that was blocking the coolers but as the last motherboard i had in here ended up to be soled for 1260 and 2229 dollars I have no reason to post something more in here Opteron 6300 series is a total new architecture used in high end servers
 
 
When i speak for Opteron 6300 architecture you end up with a system with 64 cores there is not Windows based operating system that can support so many cores and the only option is x128 bits unix you can see here:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
I was compiling firefox for a 128 bits version of unix tested in my vmware machine you can see the John Giatrakis in consol menu you can see the ping to youtube website is not more from 55 ms average:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
and after the compile was succeful you can see in here the visit to the mozzila website i was be able to view even the mobile version of their website:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Information of x128 bits unix you can find in here                                  and as you can imagine in a system like that you compile everything you download there is no other choice. In skylake  they searching ways to 'cool' then down by removing the IHS and loosing the warranty as we speack
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
  A friend of mine asked to run a test for him, he needed to see how the old computer is doing
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 Prime 28.5 torture, All the CPU cache, 8192k minimal fft and 6 Gbytes usage. The processor temprature is 41 C,  the system has temprature 32 C and none of the cores is above 45 C. The ambient temprature was 21,8 C. Finaly you can see the tempratures of the Hard Disks in RAID 0, 25 to 26 C each. 
 
 
 
John Giatrakis.
bottom of page