With the launch of X99 Motherboard lineup, Asus has also introduced something new i.e. the OC Socket. This patent-pending(as of now) OC Socket will feature some extra pins compared to the reference LGA2011-v3 grid patten. As per Asus it offers some good Overclocking benefits and Compatibility wise, its compatible with the newer generation Intel LGA2011-v3 socket based CPUs. The following X99 motherboards from Asus will feature the OC Socket :

• ROG Rampage V Extreme
• ASUS X99-Deluxe
• ASUS X99-Pro
• ASUS X99-A

What is OC Socket?

Haswell-E (Intel Core i7-5960X, 5930X and 5280K) processors actually have more pads than the reference LGA2011-3 socket design, so ASUS added extra socket pins to tap this resource. OC Socket is an enhancement based on the reference LGA2011-3 design, so for users not overclocking or tweaking voltages in the BIOS, OC Socket does not activate. However for those that do overclock and tune performance, there’s significant benefits.

Here are some comparisons between Reference Socket, OC Socket and the Haswell :

As per Asus other benefits of the OC Socket as follows –

Very Tiny Voltage drop :

Per Core Finetuning :

Even more DD4 performance :

Source : Asus

The post Asus X99 Motherboards to feature OC Socket appeared first on OCFreaks!.

Intel likes to update their processor architectures every couple of years. This however comes in two forms, which Intel likes to call ‘Tick-Tock’. Tick basically gives the present architecture a die shrink, adds a couple more features and optionally new instruction sets, while the more evolutionary Tock represents a new micro architecture.

Let’s look back at the last couple of years before we delve into today’s review.
With Sandy Bridge, Intel introduced a major architectural change, probably their best barring Conroe, and shook the market quite a bit. Ivy Bridge, the die shrink to Sandy Bridge, didn’t meet with as much enthusiasm as its predecessor due to the heat issues and a new chipset almost forced upon consumers besides providing near identical performance as comparable Sandy Bridge parts. All this time, Intel really did not have any true competition on hand. With AMD focusing on mobile parts and APU solutions and receding themselves into value oriented markets, they probably didn’t need to. Enter Haswell – Intel’s newest Tock, featuring a 22nm process, and a plethora of other options including the voltage regulators now being on die and not on motherboard itself. With it, Intel has also introduced a chipset in the form of Z87 (as is obvious).
With Haswell though, Intel’s focus is clear – providing the same/better performance with much better thermals and power consumption. It might not look important for the average Joe, but when you consider mobile space, and Ultrabooks or tablet hybrids like the Surface Pro, the advantage becomes really apparent.
Z87 as a chipset is going through same iterations from motherboard manufacturers – most of them have today a consumer range, a gaming range, and an enthusiast range.
Asus is no different. With Republic of Gamers, Asus already has probably the best branding in their hands, and it is divided into the following categories –

1. Impact – Mini ITX ROG board introduced in Z87, best for SFF gaming systems.
2. Gene – Micro ATX suited for best bang for the buck systems.
3. Hero – Entry level ATX board introduced in Z87.
4. Formula – Prosumer gaming board with best on board audio.
5. Extreme – Geared mostly for LN2 overclockers and enthusiasts.

These boards range from $195 to $380 depending upon your choice/preferences.

We are looking today at the Maximus VI Formula, priced at a hefty $300. Asus puts the Formula as their crème de la crème Gaming board, and equips the board with a set of overclocking and gaming features which they say are significantly better than others. The main one among them is what they call SupremeFX Audio. ROG shield also debuts with the Formula, with a plastic/metal sheath covering the board also seen in their Sabertooth series. So is this the best gaming board that you have been looking for? Let’s take a look.

The post Asus Maximus VI Formula Z87 motherboard review : the gaming board with an overclocking gene. appeared first on OCFreaks!.

So if you are sporting an Sandy bridge or Ivy bridge rig or rocking with a Piledriver or Bulldozer based PC then come and compete for the prizes that are up for grabs.

Rules

• Only Sandy/ivybridge or Sandy Bridge-E /Ivy Bridge-E or Piledriver/Bulldozer based processors.
• Super Pi 8M benchmark

For a more detailed set of rules click here .

Hoping to see you there !
–

Sumon

The post OCFreaks presents SuperPi 8M competetion appeared first on OCFreaks!.

Well…i was kinda riding high with my last review and was kinda sad that the P8Z77 V Deluxe had to go. But the guys over at Asus had some other plans! So one fine day…I found this at my doorstep!

My latest toy..the Maximus V Formula!

Now..before we dive into the product showcase and review lets have a look into the history of ROG line of Hardware and this board in particular Rog was established way back in 2006 with the Crosshair(yes its an AMD board and back in the FX-64 days AMD literally thrashed Intel until the 45nm C2D and C2Q came out),soon they launched another Board Commando and soon other boards followed.

In short ROG offers the best from the stable of Asus and will have everything you need in a motherboard from gaming specific features to overclocking.
Now,enough of history lessons.. Lets have a look at the board and what it offers.

The post Asus Maximus V formula review appeared first on OCFreaks!.

This RAM doesn’t have flashy heat spreader but is indeed a decent looking one.

Overclocking & Results

The Rig that we used for Benching was :

The stock or defaults values for the RAM are as follows :

Following are the different overclock Settings / Profiles used for Benchmarking:

Graphs for Benchmarking Results

Results in Details

Conclusion

As seen from the results Overclocking this RAM is very easy. In my case I reached 1920-1940 Mhz @ 11-11-11-28-38 @2T on 1.6 Volts. From this its quite obvious that the RAM can go past 2000Mhz with more relaxed timings i.e at CL12 and on 1.7Volts if required. I tried loosen the timings as much as I could and set memory Voltage to 1.7Volts on Gigabyte 880GM-USB3L; but the motherboard won’t accept it and would reset the CPU clock and RAM multiplier to default. This doesn’t mean that the RAM is not capable of reaching 2000Mhz+ since I am pretty sure that the motherboard was limiting the MAX Overclock and not the RAM. On a side not : Specs for Gigabyte 880GM-USB3L says it supports “1666+ Mhz” Ram so it was pretty good getting 1920-1940Mhz stable on this motherboard.

I would recommend this RAM for low to mid-end PC builds while there are better alternatives from Gskill and Corsair for Beefy RIG Builds. The price to performance of this RAM is rock solid!

The post GSKILL RipjawsX 4GB DDR3 1600Mhz CL9 Review appeared first on OCFreaks!.

Introduction

The motherboard which we are going to review today is Gigabyte 880GM USB3L. This motherboard falls in low to mid-end range and provides good set of features for the price. It was around $90-$100 at the time of this review. Like many motherboards out there in the market this one too has a smaller and a bigger brother. 880GM-D2H is a little bit cheaper than this and lacks USB3 support , except for USB3 both the motherboards are almost Identical. 880-GM USB3 being its bigger brother has some extra features and is priced higher as compared to the one being reviewed. When it comes to overclocking this board doesn’t fall behind and is indeed a serious Overclockable motherboard.

First , lets have a look at the box:

The post Gigabyte 880GM-USB3L Review appeared first on OCFreaks!.

LGA1156 Platform was released way back in 2009 just 1 year after the release of LGA1366 Platform.

The main differences between lga1156 and lga1366 are as follows:

1. Elimination of northbridge in lga 1156.
2. Only dual channel memory support.
3. Better turbo mode on lga1156.
4. Lower tdp of i7’s on lga1156(95watt opposed to 130 watt in 1366).

This the list of cpu’s that belong to this platform

Core i3(Clarkdale)Dual cores with ht but no turbo:

• -i3 530
• -i3 540

Core i5(Clarkdale) Dual cores with HT and turbo:

• -i5 650
• -i5 660
• -i5 661
• -i5 670

Core i5(Lynfield) Quad cores with turbo but no HT:

• -i5 750
• -i5 750

Core i7(Lynfield) Quad cores with ht and turbo:

• -i7 860
• -i7 870
• -i7 875k
• -i7 880

The chipsets supported are:

• P55–Mainly for core i7 and i5 lynfield versions,although you could run i3 and other i5’s too but you won’t be able to utilize the on cpu graphics unit.
• H55–Mainly for non lynfield i5’s and i3’s as it supports on cpu graphics unit.
• Two More chipsets are available but they are not much different,H57 and Q57.

This concludes the intro.

Setting Up Your System

1. Goto bios(Press F10/Del/F2 at startup) and set everything at default then restart your pc.
2. Again goto bios and disable all power saving features,c-states etc ,.. then set all voltages from auto to normal in the bios except the PCH(discussed later), save settings and exit bios.
3. Disable or remove any unnecessary usb devices PNP devices that you don’t need.
4. Make sure you have adequate cooling and a good cpu cooler(advisable for system safety not mandatory).
5. Make sure you have a good cpu cooler installed before going for heavy overclocking(Eg. cooler master hyper 212+ evo)

Softwares/Utilities Required

1) Core Temp–Used to measure CPU temps.

http://www.alcpu.com/CoreTemp/

2) HWMonitor–This gives temp and voltage info for the whole system including motherboard PCH temp.

http://www.cpuid.com/softwares/hwmonitor.html

3) CPU-Z–Used for getting memory and cpu specific detailed info.

http://www.cpuid.com/softwares/cpu-z.html

4) Prime95–A time tested and very good stability tester for cpu’s of any make and brand.

http://www.overclock.net/t/137251/prime95

5) IBT(Intel Burn Test)–Includes linpack test which is apparently used by intel too for testing the processors.

This is better than Prime 95 as it takes much less time to find stability issues in the system.

http://downloads.guru3d.com/IntelBur…load-2047.html

6) MEMTEST86+–Used to test the memory stability.

http://www.memtest.org/

Well established safe limits–Its better to leave this at auto.(This setting can easily damage your board and components and do very less for ocing so its better left on at auto,sometimes when ocing beyond 4ghz+ and if your cpu is not stable this might help,no guarantees though).

1) Don’t set a target in mind,ocing is a recursive process,there is no guarantee you will achieve your desired overclock.

2) If you know your mother board is weak then don’t take the bclk too high,rather increase CPU multiplier as much as you can to achieve the desired overclock.

3) If you know your mother board can take the heat then its always better for stability reasons to decrease your CPU multiplier as low as you can and increase the bclk to achieve ocing.

4) Before increasing the bclk set the CPU multiplier as low as possible and set the PCI frequency as auto.

5) Decrease the memory multiplier before hand too because when you start ocing through bclk every other clocks/frequencies in the system will also rise.

Finally Overclocking

1) After reading the previously mentioned rules and setting up the bios multipliers and frequencies, goto Bios and start increasing bclk in steps of 5 or 10(if you are confident).Save settings then again enter bios.

2) Now after reentering bios check 3 frequencies every time you increase the bclk:

1. CPU Frequency.
2. Memory Frequency.
3. PCI Frequency(although it is set to auto,doesn’t hurt to check for yourself to be extra sure).

If everything is fine then move on to next step.

3) Save bios settings and exit then boot up windows,start cpuz and core temp and hwmonitor,let them run in the background.

4) Start Prime95 or IBT whichever you trust or like(imo ibt/linx are better):

1. Run it for at least 6 hours and check for errors or bsod’s or any decripency in system behaviour.
2. If you get any error/bsod then you might need to review your bios setting and try increasing vcore/qpi voltage a bit.
3. If you did not get any error then congratulations you have done your first successfull overclock.

1. Set the run times to 20(even 10 will do,but to be extra sure),and run it and then see for any errors bsod’s and any decripency in system behaviour.
2. If you get any error/bsod then you might need to review your bios setting and try increasing vcore/qpi voltage a bit.
3. If you did not get any error then congratulations you have done your first successfull overclock.

5) Do steps 1-4 again and again till you stabilize the system(keeping voltage under safe range as stated by me in voltage section) and achieve your desired overclock.

6) In case your pc stops booting reset cmos by removing the battery in the motherboard and keep it out for 5 minutes and then put it back in and reboot.If your system still does not start it means something has gone kaput.

7) If you encounter step 6,its highly likely that either your power supply or motherboard were not upto the task or you have been overenthusiastic and pushed your system too hard.

This concludes ocing section.

End Notes

This concludes the guide.

The post Overclocking Guide for Intel LGA1156 Platform CPUs appeared first on OCFreaks!.

Legacy Intel LGA 775 Overclocking

Basically Front Side Bus (FSB) is set of wires i.e. a Bus that connects CPU to the “Northbridge(Chipset)”.By the way – Intel also calls the Northbridge as “Memory Controller Hub” or MCH for short.

Wait … what is the Northbridge in the first place?

Now back to where I was …

This FSB is the address and data bus which the CPU uses. The CPU interacts with all the other components of the computer through the FSB via the Northbridge. Since the Northbridge is so central to all the connections in the computer – increasing the FSB will also increase the speed at which other components interact with the Northbridge which in turn will increase the system performance. Overclocking the FSB may be a little tricky for someone new because the base clock rate differs from the effective(actual) clock rate which is termed as ‘Quad-Pumping’ by Intel.

On systems that support Intel processors the FSB is Quad Pumped which can be stated as :

Hence for eg. a system running at 1333 MHz FSB has a base FSB frequency of 333 MHz.

One of the key factors that makes FSB the first choice to be overclocked is that the CPU’s speed is determined by the Base FSB which is then multiplied by the CPU Multiplier which sets the operating frequency of the CPU.This relation of proportionality between the CPU and FSB frequency gives an instant CPU speed boost when the FSB is increased given that the multiplier remains the same.

The formula is given as :

The CPU multiplier is also referred to as CPU to FSB ratio in some BIOSes – its logical because if you divide CPU frequency with FSB freq what we get is the CPU Multiplier.

Lets take an example of Q9300 which has native FSB of 1333 MHz (Quad Pumped) and Multiplier ‘Top-Locked’ to 7.5

Native Processor FSB is the default FSB which is used to generate the CPU frequency. Each processor has its own predefined default FSB. When every thing is set to default – i.e. no Overclocking – the CPU dictates the Northbrigde that it wants to work at the default CPU FSB.

When is comes to FSB we are mainly concerned with max FSB supported by the motherboard and the native FSB of the processor. The Q9xx0 series Quads from Intel may have the edge of being 45nm based and slightly faster than the 65nm based Q6×00 series Quads. But the 45nm Quads have higher native FSB of 1333 MHz which makes it difficult to overclock beyond a limit and it even gets worse for quads like Q9300 which has the multiplier at 7.5 max. Consider today’s p35 , x38 , p45 , x48 chipset based motherboards which can be OCed to 1600MHz easily but not quite beyond 2000 MHz i.e. a Base FSB of 500 MHz is limit for the current generation chipsets. Now with Q9300 the ‘max possible (can be pushed more)’ overclock will be = 500 x 7.5 = 3.75 GHz which does not make it suitable for Overclocking because such a high FSB will tax both the chipset and also the RAM to maintain good FSB : DRAM ratios [ill explain RAM OCing in next Article].

On the other hand lets consider Q6600 which has 1066MHz with multiplier of 9 max. With Q6600 the ‘max possible’ OC will be = 500 x 9 = 4.5 GHz which some have reached. I remember that the Tom’s Hardware team from France managed to push Q6600 up to 5 GHz on liquid nitrogen. Due to this relation between the Base FSB and the CPU Multiplier the CPU having native FSB less then the max FSB supported by the motherboard has a more Overclocking head room than a CPU have native FSB same as the max FSB of the motherboard.

In our consideration of Q9300 and Q6600 I would go with Q6600 because its more Overclockable and will not tax the chipset with extreme high FSB as with Q9300. But on the other side the main advantage of Q9300 is the higher FSB itself. Since the native FSB of the processor dictates the default FSB at which the system will run having higher default FSB will make system run fast due to higher FSB. But the down fall is that further OC is possible from the default FSB , hence more the head room more is the Overclockability.

AMD With HTT

On AMD systems the bus used to interface the CPU with Northbridge uses HyperTransport Technology (HTT) which was earlier called Lighting Data Transport (LDT). This is similar to FSB on Intel systems which uses Assisted/Advanced Gunning Transreceiver Logic (AGTL+). On AMD systems the bus speed is the base HT speed from which we get the HT Link speed.

Here’s the formula :

Usually the HT multiplier swing is limited between 1 and 5(max).
So if we have bus speed of 200 MHz and HT multiplier set to 5x then our HT link speed will be : 200 x 5 = 1000 MHz HT Link

The HT link speed is used to get actual data rate of the system bus. Since HTT can transfer data twice per clock pulse (Double Data Rate) the effective can be calculated as :

In our case we have: 1000 MHz x 2 = 2000 MHz Effective Bus Data Rate

As on Intel systems the CPU speed here is also obtained by multiplying the CPU Multiplier with the bus speed (Base HT) which is as follows :

Lets say we have an AMD X2 5000+ which has its multiplier set to 13 . Hence in our case the CPU would be : 200 x 13 = 2600 MHz = 2.6 GHz

The post Legacy Intel LGA 775 / AMD AMx CPU Overclocking Guide appeared first on OCFreaks!.

Introduction

The fermi architecture was reavealed by Nvidia way back in 2009 and the first Nvidia GPU series to be based on that architecture was the GTX 4XX series.The first card to be released was the GTX 480,and later on various other cards were released.Then the GTX 5XX series was released not too long ago with the first card being the GTX 580.

Major differences Between previous gen GPU’s like GTX 2XX and the new fermi architecture are as follows :

1. Fermi based cards have GDDR5 memory modules with ECC(Error Correction) capabilities while the older architectures had GDDR3 non ECC modules.
2. Fermi based cards had core clock and shader clock speeds interlocked in the ratio 1:2 while the older architectures need not be so.
3. Fermi based cards were the first ones to support OpenGL4 and DirectX11 API’S(Application Programming Interface).
4. Fermi based cards had tessellation units specifically for its hardware based implementation while the older architectures lacked it.

Well there are many more differences but for the sake of simplicity these are the most important ones to be concerned about.

List of GPU’s based on the Fermi architecture:

1. Dual GPU : GTX 590
2. High End : GTX 580,GTX 480,GTX 570,GTX 470
3. Mid End : GTX 560 TI,GTX 560,GTX 460,GTX 465,GTX 550 TI,GTS 450
4. Low End : GT 545,GT 440,GT 530,GT 430,GT 520,GT 510

Setting Up Your System :

1. Firstly no BIOS adjustments required for the GPU OC unless you have changed your PCI bus frequency,if yes then change the PCI frequency to default.
2. Now login to windows and install some essential Softwares/Utilities(Discussed later on in the next section).
3. Now set all your card Frequencies to default/stock.

Softwares/Utilities Required:

1) Geforce GPU Drivers–Do we need an explanation here .

Download–http://www.geforce.com/Drivers

2) MSI Afterburner–“The” best and “The” most user friendly OC’ing app available on earth.

Download–http://downloads.guru3d.com/MSI-Afte…load-2850.html

3)Unigine Heaven–A very good benchmark for DX11 Tessellation enabled GPU’s.

Download–http://unigine.com/products/heaven/download/

4)OCCT–A gem among stress testing apps,a very versatile GPU testing/stressing Tool.

Download–http://www.ocbase.com/perestroika_en/index.php?Download

5) 3DMark11–A very Good DX11 GPU/CPU stresser.

Download–http://www.3dmark.com/3dmark11/download/

6) GPU-Z–GPU monitoring Tool.

Download–http://www.techpowerup.com/downloads/SysInfo/GPU-Z/

Understanding Some Basic Stuff :

Important Voltages

GPU VCORE : Its the amount of volts fed into the GPU’s core,its the most Important voltage for OC’ing GPU’s.

Safe Limit : Stock – 1.1Volts

Mem Volt-Important for OC’ing memory.

PLL Volt-Its just a gimmick by the manufacturers ,till now there has been no evidence of the card being able to OC higher and be stable by increasing the PLL Voltage.

Some Basic Rules Before Moving On

1. Don’t set a target in mind,OC’ing is a recursive process,there is no guarantee you will achieve your desired overclock.
2. Make sure your GPU temps are within 85C before any overclock @ stock,because if your cards already above 85C at stock,there won’t be much overclock headroom available to you,it does not mean your card can’t overclock,but it means that its not too safe to run your GPU at 90-100C 24X7.
3. Make sure your GPU is dust free as dust will add in to the temps of your GPU.

Finally Overclocking

—–Without Messing With The Voltages

• 1) After booting into windows start GPU-Z and Afterburner,close any unnecessary applications,verify your cards stock clocks and voltage.Also update your GPU drivers.

• 2) Now to start OCing bring afterburner into view and you will see 4 sliders by defaults namely

Here our aim will be to stabilize all the clocks one by one.We start with core clock(core and shader clocks are linked so essentially you will be increasing both automatically if you increase any one of those).

• 3) Start raising the core clock speed 10 MHz at a time(you can also increase more if you are feeling adventurous but increasing in lower steps ensure that you will find more stable and safer highest frequency).After each increment run 2 loops of unigine heaven and 3dmark vantage/11 benchmarks,as these will immediately show the stability of your GPU OC by either crashing/slowing down/texture corruption or any graphical anomalies.

• 4) Repeat the above step till you GPU OC passes all the tests without any problem,after you start getting problem reduce the core clock in increments of 5 MHz and again run 2 loops of unigine and 3dmark benchmarks each,and keep decrementing till your card becomes stable and stops artifacting or giving any problem.

• 5) Through the last step we have now successfully found out our stable and highest core clock ceiling.So now in a way we have isolated the core clock from memory clock as memory clock have been untouched till now.So now start incrementing memory clock in increments of 10 MHz(Please don’t be over adventurous with memory like with core clock,memory is the component which can be easily destroyed by the slightest of mistake).Now repeat the same benchmark runs and strategy followed for core clock in earlier steps and you would have found out your memory clock ceiling too.

• 6) Congratulations! you have achieved a stable OC on your GPU without messing with voltages.Now to keep these settings click “Apply Overclocking At Startup” in afterburner,and save the OC settings in a profile represented in afterburner as numbers at the bottom.

—–With Messing With The Voltages

• 1) Here we follow the above 6 steps word to word and follow up from there.Now that we have got stable OC without increasing voltages,and if we need even higher OC then we need to up the voltage in small increments.

• 2) Now for this we need to make some changes in the options of afterburner,click options in afterburner and then under the generals tab check the “unlock voltage and enable voltage monitoring options”.

• 3) After setting up afterburner increase the core clock by 10 MHz again and again benchmark it and see if it crashes,if it does increase the voltage in increments of 5 mV and then again benchmark and see if the card has stabilized or not.

• 4) Repeat the procedure given above till either you reach your desired OC or you reach 1.1 V.

• 5) Once you reach the max possible OC apart from running the benchmarks also run OCCT GPU test for at least 30 minutes and check the stability of the card,if it survives well you have achieved a successful Overclock,if it does not decrement the core clocks by 5 MHz and keep doing this till your GPU is stable.

• 6) Final step enjoy gaming with increased performance .

The post Nvidia Geforce Fermi GPU Overclocking Guide appeared first on OCFreaks!.

1. RAM Clock Frequencies ,
2. RAM Timings/Latencies ,
3. FSB : DRAM ratio ,
4. SPD Chip ,
5. and Voltage.

DDR(1/2/3) RAM modules have 3 types of ‘clocks’ associated with them:

1. The first is DRAM Core clock or memory Clock.
2. The second comes I/O Bus clock.
3. Finally the 3rd is the Effective Data Rate.

The prime difference between DDR1 and DDR2 is that DDR2 can run its I/O bus clock twice the memory clock but with higher latencies and DDR2 has prefetch size of 4 bits as opposed to DDR1’s 2 bits. And the the main difference between DDR2 and DDR3 is that DDR3 can run its I/O bus clock at four times the memory clock but with higher latencies and DDR3’s prefetch buffer is 8 bits deep.

For DDR2 :

• I/O Bus Clock = DRAM Core Clock x 2
• Data Rate = I/O Bus Clock x 2 (i.e ‘DDR’)
• Data Rate = 4(bits per clock) x I/O Bus Rate [4n prefetch]

For DDR3 & DDR4 :

• I/O Bus Clock = DRAM Core Clock x 4
• Data Rate = I/O Bus Clock x 2 (i.e ‘DDR’)
• Data Rate = 8(bits per clock) x I/O Bus Rate [8n prefetch]

Here Onwards, Whenever I Refer ‘Base Memory Clock’ or ‘I/O Bus Clock’ Or ‘DRAM Frequency’ ALL MEAN THE SAME.

For DDR(1\2\3\4)-SD-RAM the data rate is twice the base clock rate.

For e.g. RAM running at DRAM Frequency (Base Memory Clock) of 400 Mhz has an effective (data rate) Frequency of 800 Mhz i.e. effective signalling rate of 800 MT/s [MegaTransfers per Second].

Before going further into details I’d like you to have a look on the screen shot that I have taken on my PC for explanation so that you can understand better; that’s what I hope!

Screenshot 1: Showing the SPD chip readings and memory details.

Screenshot 2: Showing the ‘current’ RAM timings , frequency and FSB : RAM ratio.

SECTION 2 : Ram Timings

When it comes to Overclocking RAM we are just concerned with 4(out of many) primary timings as seen in screenshot 2. The chips used on the DDR ram modules have different types of timings(Primary & Secondary) which give us a sense of the speed of ram along with its stability and rated frequency of-course. When it comes to overclocking loosening(increasing) the timings is very effective for increasing the RAM frequency making the modules stable at high frequencies and on the contrary reducing the RAM frequency the timings can be tightened (lowered). But ironically lowering the timings decreases access time but at the cost of the bandwidth.

Now lets get a feel of some of the Signals(strobes) for RAM:

1. /CAS[Active low] : Column access strobe(signal) When this signal goes low , the column in the selected row is ready to be accessed in burst mode of 2 , 4 or 8.
2. /RAS[Active low] : Row access strobe(signal)
3. Precharge : Used to activate/deactivate a row in the selected bank before it can be used for read/write operation.

These timings are specified are in the following order : tCL-tRCD-tRP-tRAS and as you have guessed ‘t’ stands for time.

The 4 important timings with respect to RAM Overclocking are :

1. t-CL [CAS Latency] : It is time elapsed between the memory controller sending the address of the column and the data that “first” arrives in response. Since data is sequentially placed in memory and a row contains sequential data so its quite simple to catch the fact that columns will be switched more frequently than rows so CAS will have a big impact on performance. Though some say that CL is not that important, but generally it is and its only in the case of bizarre memory access patterns that CL may become less significant.
   Typical t-CL Values in clock cycles :

   • DDR1 – 2,3
   • DDR2 – 4 to 6
   • DDR3 – 6 to 10
   • DDR4 – 10 to 18

    

2. t-RCD [RAS to CAS Delay] : It is the amount of delay between a RAS and a CAS.=or= Simply speaking it is the time taken to select a particular row first and then selecting the particular column for data access. It doesn’t have a huge impact on performance.
   Typical t-RCD Values in clock cycles :

   • DDR1 – 2 to 4
   • DDR2 – 3 to 5
   • DDR3 – 6 to 10
   • DDR4 – 10 to 18

    

3. t-RP [RAS Precharge Time] : It is the time required to deactivate the current row and activate next row.
   =or=
   Simply delay caused by switching between rows.
   Typical t-RP Values in clock cycles :

   • DDR1 – 2 to 4
   • DDR2 – 3 to 5
   • DDR3 – 6 to 10
   • DDR4 – 10 to 18

    

4. t-RAS [Active to Precharge Delay / Row Active Time] : Time required between an active and a precharge command.
   =or=
   The time taken between 2 memory access / Data requests.
   =or=
   The time taken to activate a memory bank(row) and then deactivating it. It affects stability more than performance. It is approximately equal to tCL + rRCD + tRP [=>tRAS] while in some cases it may not be so. And here’s a quote from a wiki article :

   “in practice for DDR RAM Modules , it should be set to at least tRCD + tCAS + 2 to allow enough time for data to get streamed out”

   Typical t-RAS values in cycles :

   • DDR1 – 5 to 12
   • DDR2 – 10 to 19
   • DDR3 – 15 to 30
   • DDR4 – 20 to 36

    

SECTION 3A : FSB:DRAM Ratio on Legacy Systems

In case of legacy systems where the memory controller was present on the northbridge, while talking about FSB : DRAM ratio we are concerned with the base FSB frequency and the DRAM frequency (base clock rate).This ratio tells us “who is running faster than whom” and when DRAM frequency is more than base FSB or both are the same then we won’t have any issues with system performance. Depending on both the frequencies, FSB : DRAM ratio can yield any one of the 2 operating modes which are Synchronous and Asynchronous.

Synchronous Mode: (sync)

• In sync mode both the frequencies are equal which means that both the RAM and FSB are running synchronously and yields max performance.

Asynchronous Mode: (async)

• Note that Async mode can give either max performance or average performance as follows:
  • If the FSB frequency is more than DRAM frequency then u’ll get average or poor performance because FSB(hence CPU) is running faster than RAM and so RAM cannot cope with data hungry CPU’s requests and eventually CPU has wait i.e be in a idle state for a while till the data arrives from RAM.
  • If the DRAM frequency is more or equal to FSB clock speed then u’ll get max performance since here CPU doesn’t have be idle in between and also DRAM being faster than FSB is not a concern since CPU will read/write data at FSB’s rate.

Common FSB ratios :

• 3:4 -> For each 4 DRAM ticks , FSB ticks at rate of 3
• 2:3 -> FSB is at 266 Mhz and DRAM is at 400 Mhz
• 1:1 -> Both are equal for e.g both running …at say 266 Mhz
• 5:4 -> You can guess this!

SECTION 3B : FSB:DRAM Ratio on current Systems

On Current systems be it Intel or AMD where the Memory controller is integrated completely inside the CPU die, the significance of FSB:DRAM ratio has changed. What we used to call FSB is now QPI(Quick path interconnect) on Intel systems and for AMD systems we call it HyperTransport(HT). Now its just a ratio of Base Clock (BCLK) to Memory Clock.

SECTION 4 : Ram Voltages

This is the last thing you wanna mess with , after it being configured according to the EPP or XMP profile. Generally voltage increase must be limited to 8% to 10% of the max supported voltage according to EPP or XMP Profile and you must make sure that the chips on the RAM modues supports this level of voltage increment else you will end up damaging it. For DDR2 for Mobos which support Core2 architecture 2.1/2.2/2.3 V is safest maximum but this doesn’t mean all RAMs have this as their max safe limit. The safest max voltage depends on the chips used by the manufacturer. For DDR2 generally the default is 1.8V while for DDR3 its 1.5V . It again majorly depends on the chips used for the modules. But i’d say never-ever do this until you have enough experience and know about various chips. Instead set the RAM’s voltage according to EPP/XMP profile if not set.

Standard JEDEC & Oveclocked Voltages :

• For DDR1 – Jedec = 2.5V | OC = 3.2v Max
• For DDR2 – Jedec = 1.8V | OC = 2.3V Max , 1.8V~2.2V is ‘just safe’ for 24/7
• For DDR3 – Jedec = 1.5V | OC = 2.1V Max , 1.5~2.0V is ‘just safe’for 24/7 while on some systems 1.65V+ may be fatal (old corei7 LGA1366 systems).
• For DDR4 – Jedec = 1.2V | OC = 1.35v-1.40v Max , 1.2~1.35V is ‘just safe’for 24/7 while it seems that going over1.5V might be serious risky stuff.

SECTION 5 : A Word On SPD & XMP/EPP Profiles

SPD is the acronym for Serial Presence Detect. SPD chips are now commonly found on the SDRAM DIMM Modules. Its job is to store the RAM settings for different frequency and voltages. SPD makes it easier for bios to configure RAM for the system. Apart from the JEDEC standard profiles SPD also contains EPP or XMP profiles. EPP is Enhanced Performance Profile which can be read by some Nvidia and AMD chipsets and configures the RAM according to it. If not supported then BIOS just loads the default configuration. EPP is present on RAM modules which are marketed as “SLI ready” or “Crossfire ready” which is just a marketing thing then any thing else. If the Motherboard cannot read EPP it does not mean that the RAM cannot be configured as per the EPP. In this case we just have to configure it manually. XMP is Intel’s substitute for EPP which is short for eXtreme Memory Profile.

The use of EPP/XMP is just to get the RAM configured for max performance at its rate specifications which are often beyond the standard Jedec specs for Overclocking grade RAMs. When you first install High-speed memory sticks, they will run at ‘default’ Jedec speeds. To configure your RAM sticks at the rated speed you will have enter BIOS and apply the XMP profile. This setting is usually present under the Overclocking/Tweaking section of your BIOS. Many RAM sticks have multiple XMP profiles; so just choose the best one available, which will provide a baseline setting for overclocking. Screenshot-1 shows SPD profiles of my RAM.

Here is my BIOS screenshot showing XMP profiles available to select from:

SECTION 6 : How do we do it?

After getting equipped with the knowledge presented above its time to do some RAM Overclocking which is straight forward.

I like to do it in 2 cycle fashion:
CYCLE A: Find your Overclock settings at rated Voltage.
CYCLE B: Find your max Overclock at bumped up Voltages past your Overclock at rated Voltages.

The following Steps Integrate both these Cycles:

Pre-requisite : If applicable, First apply the XMP/EPP profile for your RAM from BIOS.
• 1) Now, Increase the frequency to next available increment from the list. This setting is usually called “DRAM frequency”. On Some systems it may be called “Memory multiplier” where you will need to select the next multiplier value in the list. You can fine tune the values of these increments by changing the FSB or BaseClock as mentioned in STEP 8.
  
• 2) If the system is stable then, run all the necessary stability tests.
• 3) If tests are successful repeat step 1.
• 4) If tests fail i.e. system crashes(unstable) during testing then loosen(increase) the timings. You can increase the 4 timings mentioned above by 1-1-1-2 or 1-1-1-3 receptively and repeat until you get stability.
  
• 5) Now start again from step 2.
• 5.1) Some where here you must get a stable overclock. Now if you want you can repeat the process again from Step 1.
• 6) If you want to Overclock memory further and/or if system crashes after step 5 its time to increase voltage – AT YOUR OWN RISK!
• 7) Increment DRAM voltage in steps of 0.05v or 0.02v depending on your memory module [BEWARE].
• 7.1) If system is unstable or if sometimes you get: “Overclocking failed” in BIOS during boot then, it may be time to increase IO voltage. This is essentially the memory controller voltage. On Current systems(CPU with IMC) its generally called VCCIO in most BIOSes. On Legacy systems its the Northbridge Voltage or NB Voltage. Initially try with 1% to 2% bumps. Consult the Motherboard manual and/or CPU datasheet for absolute max IO Voltage. Its better to ask experienced fellow Overclockers who own the same(nearly) system as your.
• 7.2) You can stop after step 7 or proceed further for some more insanity.
• 8) You can adjust your FSB or Base Clock(BCLK) and repeat from step 2 =or= step 1 for memory frequency fine tuning. On some systems (before Skylake) the Base Clock is tied to PCIE & DMI, hence you don’t have much headroom to overclock your Base Clock(BLCK). If not obvious, please note that increasing your base clock will also increase your CPU speed.
• 9) After this you would have pushed your RAM to its limits and a little more OCing would seriously damage your RAM or even worse your CPU with integrated memory controller..

RAM stability test and other Utilities:

• http://www.memtest86.com/
• http://hcidesign.com/memtest/
• Prime95 – for CPU Benchmarking & stability test | Link : http://mersenne.org
• CPUZ – Gives you all your hardware info | Link : http://cpuid.org

The post RAM Overclocking Guide / Tutorial appeared first on OCFreaks!.