After acquiring one of these desktop processors from Intel a long time ago, I came across the bundle consisting of a motherboard and CPU again while rearranging other hardware. I personally have never overclocked on this platform, but according to everything you read in overclocking circles at the time (2007/2008), there should be quite a bit of potential. Poor heat transfer in the heat spreader is not an issue, as processors from this era were soldered.

With these first quad-core processors, Intel took the same approach as it had some time before with the first dual-core processor: two dies together on one carrier. From a technical point of view, the result is therefore more of a 2 + 2 core processor, also in terms of other functionalities (shared bandwidth, shared cache, double power consumption, etc.).

Which processors are we talking about?

The code name Kentsfield tells seasoned overclockers everything they need to know: the first mainstream quad-core CPUs were called Intel Core 2 Quad Q6xxx. The most legendary model, the Q6600 (test from back then), was first available in the “B3” stepping with 105 watts TDP and shortly thereafter in the improved “G0” stepping, then with 95 watts TDP. The newer G0 revision is generally slightly better, especially when overclocking, where the lower power consumption is very noticeable and significantly increases the headroom.

My Q6600 is the G0 stepping, which is ideal for an “old-school FSB overclocking session.” — the multiplier is locked (only the much more expensive Extreme variants available at the time had an unlocked multiplier). The biggest advantage was getting at least equal performance from the cheapest quad-core processors compared to their expensive counterparts.

These are all the Kentsfield models from back then, excluding Xeon:

All of these processors fit into the LGA 775 socket, have 8 (4 + 4) MB L2 cache, and are quad-core. They are all manufactured by Intel with a 65 nm structure size. For serious overclocking to the limits of this generation, an Extreme model with a free multiplier would of course be essential. In my case, however, it’s more about nostalgia – apart from that, the Q6600 G0 in particular is a legend. PCGH users also thought so in 2010.

⧉ Intel

⧉ Intel

⧉ Intel

Hm, Mainboard?

At the time, I got the CPU together with two 2 GB sticks of 667 standard RAM installed on a Gigabyte GA-965P-DS4 in revision 1.0. A usable to good board for overclocking. I replaced the RAM with four XMS2 DDR2-800 modules, each with 2 GB from Corsair – so in total, I still have 8 GB of RAM that is theoretically usable today.

I continued to use the Freezer 7 Pro from Arctic that was already on the board as a cooler, together with new MX-4 thermal paste* – whether this will be sufficient remains to be seen.

⧉ Gigabyte

Initial situation

In good old DIY fashion, I put the motherboard on the benchtable, inserted all four RAM sticks, and connected the cables. For such projects, I like to use a “problem-free” GPU model, in this case an older GeForce 210 from ASUS with 1 GB VRAM – no additional power connection, no fan. PCI Express 2.0 with 16 lanes.

As the system drive, I used an inexpensive Intenso SATA SSD with 256 GB capacity* connected to the first SATA-II port, with Windows 10 22H2 Professional 64-bit installed. The BIOS was already up to date (F12 from June 25, 2009).

This is what it looked like:

Then let’s get started!

First of all, I ran Cinebench R23 for single and multi-core performance, using the default settings. Incidentally, the required VID for my model is a moderate 1.288 volts. The results were 330 / 1251 points:

As was previously common with Gigabyte motherboards, you have to press “CTRL + F1” in the main BIOS menu to display all options. At first, I wondered why I couldn’t enter RAM timings anywhere – but after pressing the key combination, all the options suddenly appeared in the overclocking submenu. You have to know this kind of thing first.

Once I had cleared the first hurdle, I set everything relevant from “Auto” to fixed values, primarily the RAM timings to those stored in the modules (5 – 5 – 5 – 18 at 1.8 volts) and the PCIe speed to “100 MHz”. I wanted to achieve a RAM divider of 1:1, so I always made sure that the final value was at or below the 800 MHz of the RAM modules.

The actual overclocking began here: First, I had to find out what this specific processor was capable of at its original Vcore voltage. According to the BIOS, this was 1.28750 volts, which matched the read VID of 1.288 volts exactly. 266 MHz is the original FSB clock, which has to run in any case.

Front Side Bus…

I slightly increased the clock speed, saved it, restarted, and then briefly checked stability with Cinebench R23. Starting with 300 MHz FSB – as expected, this ran smoothly and already resulted in a clock speed of 2.70 GHz (300 x 9). Next, 333 MHz also ran without any problems (333 x 9). This already resulted in 3.00 GHz instead of the original 2.40 GHz.

At a clock speed of 350 MHz, I had my first blue screen in CB23. No problem, just slightly increase the Vcore in the BIOS – in this case to 1.30 volts to have some leeway. Then everything ran smoothly again, reaching 3.24 GHz!

After further smaller clock jumps, instabilities, and Vcore increases, I had to increase the FSB and MCH voltage by a minimal 0.10 volts for the first time at 372 MHz FSB. Then it ran stably again. I pushed the whole thing much further, up to about 1.55 volts—that’s way too much, especially with air cooling!

I wanted to find out at what point the clock speed no longer scales well with the Vcore. With this very old hardware, I ultimately settled for 380 MHz FSB, as anything above that required really significant voltage increases for marginal gains (from 380 MHz to 386 MHz instead of 0.10 volts, now 0.25 volts on FSB and MCH, plus 1.525 volts Vcore!). Of course, the temperatures also rose disproportionately – I wouldn’t have wanted to go above 1.50 volts permanently anyway. The Arctic Freezer 7 Pro does a good job; thermally and voltage-wise, this is an acceptable range.

The result

With 380 MHz FSB and 1.45 Vcore set in the BIOS, this results in 3.42 GHz at 760 MHz RAM clock speed (1:1 division, 380 x 9). That’s 1.02 GHz more than the original, even with a rather small air cooler. It’s clear why this CPU is so well known among overclockers around the world – my sample is actually only average in terms of silicon quality, as can be seen from the average VID. Under full load, HWiNFO reads only 1.392 volts as Vcore, so the Vdroop is clearly visible.

The processor does get very warm with these settings. After about 45 minutes of Prime95 Smallest FFTs, the hottest core was at 79 °C and the package at 84 °C. These are clearly too warm as permanent values, but in Cinebench R23, nothing ever exceeded 72 °C. Prime95 is a rather unrealistic load in normal operation anyway.

CB23 Comparison

Here, I have once again compared the changes between the original and the newly overclocked Q6600 G0:

This was a very nostalgic overclocking project. All that remains to be said is: you notice the difference in performance immediately. Both in terms of computing speed under Windows 10 and the significantly increased waste heat. Over 40% performance gain with a locked multiplier at only 15% more voltage!

Der Beitrag The first desktop quad-core processors: OC legends erschien zuerst auf flohs blog.

I proceeded as usual when overclocking the system. Requirements:

• BIOS to factory settings
• Clean Windows
• Benchmarking software available
• Preferably without internet
• Plenty of time and patience
• Not exactly midsummer

XMP first

First, I activated the stored 2400 MHz XMP profile in the BIOS. After a few minutes of Memtest86, numerous errors appeared, and the RAM was unstable. After a slight increase in voltage, the RAM was still unstable. So I left all the loaded values as they were and only reduced the clock speed to 2133 MHz.

With these settings, Memtest86 then ran for two full days without any errors. Great, that’s a very high clock speed for a fully equipped four-channel processor.

Then I optimized the timings a little more, everything within reason. 100% stable.

My first overclocked Xeon

Here, too, my goal was to achieve an overclock that is not completely static, so that the power-saving functions could still be used. Even if that might cost a little performance in the end.

After much back and forth, as is often the case with overclocking, I finally settled on the following values:

Ready?

This overclock is not Prime95-AVX-stable, but since these commands are only used very rarely anyway, I don’t see a problem with that. Folding@Home runs, albeit at the thermal limit.

Thermal management is an issue anyway: I would have assumed that the Dark Rock Pro 4 would be the limit for this CPU. That is only true to a certain extent: The “extreme” VRM layout is apparently not extreme enough; the motherboard throttles the processor after excessive load for thermal reasons!

I had to install a super-thin 120 mm PWM fan from Akasa* behind the motherboard to prevent this throttling (the Rampage IV Extreme also has VRMs on the back of the board!).

At maximum, the processor draws a hefty ~230 watts in CB23 Multi, 100 watts more than specified by Intel (130 watts TDP).
I would describe the current settings as “stable for everyday use.”

OK, now the graphics cards

Good joke. Anyone who knows what temperatures Nvidia’s blower-style cooling solutions reach knows that this is pretty much out of the question. At this point, custom water cooling would be necessary, including suitable coolers for the cards. I decided against this: firstly, I want to be able to use the cards individually somewhere else in the future, and secondly, you can’t actually get the necessary parts new anymore.

On top of that, there are all the disadvantages of custom water cooling, such as high costs, space requirements, maintainability, durability, etc.

I set the thermal and performance limits to the maximum in MSI Afterburner under each operating system so that I can boost more freely, and more is simply not possible. But that’s enough, because four GTX TITAN X cards are already very overkill with the original settings.

Der Beitrag <h5>Retro-XP-PC #12: </h5><h3><b>Overclocking</b></h3> erschien zuerst auf flohs blog.

After successfully installing the drivers, I first installed Unigine Heaven (a DirectX 11 benchmark) to see if the 3D acceleration really runs smoothly and also how the power-hungry 55-watt TDP card performs in the relatively small 15.6″ EliteBook 8540p. As always, I monitored everything with HWiNFO64.

First surprise:

The Quadro 2000M reveals a lot more than the NVS 5100M did, namely very interesting VRM values such as temperatures and current/voltage values:

After Heaven had been running for a few hours without any problems and I noticed that I still had about 10-15 °C of thermal headroom before it would become critical, the question naturally arose…

…if the NVS 5100M could be overclocked so easily and significantly with MSI Afterburner, can the Quadro 2000M do the same?

The answer is yes – and the performance boost is nothing short of incredible.
But first, testing is carried out using original clock speeds.

First verify test values, with obstacles:

To maintain consistency with PassMark scores for comparison purposes, I installed and ran Performance Test v11.1 on Windows 10.

The result was only 711 points with original clock speeds (550/900 MHz) – significantly less than previously listed in the technical.city table (this could be due to many factors, such as Windows, drivers, CPU performance, background processes, daily mood, etc.).

It should have been 765 points, but with 711, it is now 7% fewer points.
At first, I suspected that the CPU was not powerful enough or that background processes were preventing the boost to 3.20 GHz on one core from being achieved consistently, meaning that it could not feed the GPU well enough.

Then I thought that it could at least partly be the “penalty” that this benchmark imposes because it cannot run the intended resolution (Full HD, but the panel built into our 8540p can “only” run HD+, i.e., 1600 x 900). This “penalty” is appropriately 7%, so that could well be it.

I verified this by connecting an external monitor via VGA (I had the cable to hand, DP would also have worked) and lo and behold – now I have 740 points. Still not quite there, but better. Only a 3.3% difference, so I left it at that.

Of course, at higher resolutions, the load is increasingly shifted towards the GPU, which in this case could reveal a very slight CPU bottleneck.

Now it’s time to overclock!

After a few hours with MSI Afterburner (slowly increasing the clock speeds one after the other and repeatedly testing with Heaven) and HWiNFO for monitoring, I reached the maximum overclock, almost at the limit—I had to reduce the VRAM clock speed slightly because one of the two cards started to show artifacts.

Since the original NVS 5100M was presumably suffering from dying VRAM components, it’s probably a good idea to go a little easier this time, regardless of stability.

Of course, I wonder what clock speeds would be possible on the GPU core if MSI Afterburner didn’t limit it (I suspect that the real limit, as is always the case, comes from Nvidia’s VBIOS, which is located on the card).

In the subsequent rerun of the PassMark Performance Test on the external Full HD screen, the result was a smooth 1100 points – a whopping 49.1% increase in points!

I am very satisfied with the results achieved; the Quadro 2000M runs much better than expected!

It was definitely worth it – of course, the card now reaches 95°C under prolonged full load, whereas previously it was more in the range of 80–85°C. The VRMs then exceed 100°C significantly, sometimes reaching 115°C or more.

In our case, I can live with that, since the GPU will never be under extreme stress for very long, as in a benchmark. It’s better to have performance reserves that you rarely need than to constantly run at the limit.

Here are the overclocking changes in tabular form:

So, looking at the entire process, these were the increases:

Original GPU, Nvidia NVS 5100M with 1 GB VRAM:
199 PassMark points – that was the baseline.
100% performance

Overclocking to the limit of the Nvidia NVS 5100M with 1 GB VRAM:
240 PassMark points (the stronger of the two)
121% performance

Conversion to Nvidia Quadro 2000M with 2 GB VRAM:
740 PassMark points
371% performance

Overclocking to the limit of the Nvidia Quadro 2000M with 2 GB VRAM:
1100 PassMark points
552% performance!

So the whole operation more than quintupled the original graphics performance of our 8540p and improved a few other things as well (DirectX 11, WDDM, 2 GB VRAM)!

Brief explanation of overclocking graphics cards:

I hope that I have been able to provide some general insights into overclocking (Nvidia) graphics chips. The whole process works in the same way with all Nvidia graphics chips that allow overclocking – you can find this out by installing and launching Afterburner. If everything is grayed out, overclocking is not possible.

Up to and including the Maxwell architecture (GT/GTX 9xx series), you could even freely edit the VBIOS limitations by flashing a modified VBIOS onto the card, which would then remove or extend the limits in Afterburner.

Of course, this is more dangerous than doing it with running software like Afterburner, but it allowed you to do cool things like writing the overclock directly into the VBIOS and thus being able to use it under Linux / Macintosh without special software – or simply passing on “pre-overclocked GPUs” that anyone can use without special knowledge.

Increasing or removing thermal and/or performance and/or voltage limits, programming other VRAM modules, flashing them, then desoldering the existing ones and installing larger ones – all of this is possible. It’s pretty cool.

Or undervolting and underclocking to optimize efficiency, which is also very useful, especially with limited cooling options – but that’s a science in itself and a topic for another post.

The topic of GPUs is now complete, so let’s move on to other hardware changes I’ve made.

Der Beitrag <h5>RetroBook 8540p #9: </h5><h3><b>Overclocking the GPU</b></h3> erschien zuerst auf flohs blog.

After many years, Nvidia graphics cards from the Pascal era are still serving us well in our “main PCs.” This is partly because their performance has not been so poor that replacement would have been unavoidable, and partly because Nvidia graphics cards have since become extremely expensive, large, and power-hungry.

My (Floh PC) has an “ASUS ROG Strix GeForce GTX 1080 Ti 11 GB OC” installed.
Tamy (Tamy PC) runs the smaller sister model “ASUS ROG Strix GeForce GTX 1080 8 GB Advanced” .

The exact model numbers from ASUS are “ROG-STRIX-GTX1080TI-O11G-GAMING” and “ROG-STRIX-GTX1080-A8G-GAMING” – tests of the cards can be found here for the 1080 Ti OC and here for the 1080 Advanced.

⧉ ASUS

⧉ ASUS

Both cards only activate the fans at around 55°C (0 dB mode) and from there on, the speed is increased in parallel with the rising temperature, which is generally a very good feature – it protects the fans and your hearing, and also means that less dust gets into the cooler.
In addition, both cards run with MSI Afterburner, but “only” with the power and temperature limits increased to the maximum so that they can be boosted more “freely.”

I had been wanting to try to improve the cooling for a long time, especially for the 1080 Ti.
Not because it gets too hot, but because the fan noise bothered me more and more, even at medium load. I would say that efficient cooling is always a good thing.

The thermal paste has been changed several times over the years, using Arctic MX-4*.
In general, new thermal paste is always a good idea, especially if it has never been replaced.
It did its job well, but I was interested to see whether liquid metal* would significantly improve the cooling efficiency.

Requirements:

Since I had familiarized myself with the topic of liquid metal (delidding an i7-7700K), I knew what to look out for – for example, the contact surface between the die and the cooler should ideally not be pure copper, but nickel-plated – to prevent the liquid metal from “sinking in” over time.

This absorption is not an unavoidable problem; you simply have to repeat the application process at increasingly longer intervals until the cooling performance remains constant. The indium in the liquid metal simply reacts with the copper, and at some point the surface has reacted so thoroughly that nothing more is absorbed.

After a little research, it became clear that the cooler of the 1080 Ti is nickel-plated and thus provides optimal conditions for liquid metal. In the 1080, the heat pipes appear to make direct contact with the die and also appear to be nickel-plated, but in the contact area they are “smoothly ground,” which partially reveals the copper again—not ideal.

⧉ Hardwareluxx

⧉ Kitguru

Liquid metal is quite expensive compared to thermal paste, so I was initially skeptical about whether it would be worth it. As with the delidding project before, I opted for the market leader, as I still had some left over:

⧉ Thermal Grizzly

Now it’s being dismantled…

You can easily find instructions on how to disassemble these card models on the Internet.
Here is a good video tutorial that can be used for both models:

The connectors for the fans are very fragile, so be very careful.

Once everything has been dismantled, it is of course advisable to thoroughly dust the components.
The cooler, the GPU die, and the surrounding area must be thoroughly cleaned with isopropanol*.

Now it’s time to prepare:

First, you need to make sure that no short circuits can occur later, as liquid metal is electrically conductive. This is easiest and most attractive to do with the 1080 Ti, as the die sits on a circuit board carrier reinforced with a metal frame. The 1080 lacks this metal reinforcement:

I covered the metal bracket on the 1080 Ti and the circuit board on the 1080 directly with insulating tape* to make it easier to apply the following coat of paint for insulation. The red areas were covered with tape:

Afterwards, you can paint the exposed area very well with (nail) polish*, but you must be careful around the die, so please be careful (marked in yellow). No paint should get on the die, and if it does, it must be thoroughly removed. You can also insulate the area with Kapton tape*, which has the advantage of being removable. I found painting to be safer and therefore did that, as it does not have to be reversible.

Once all SMD components in the area now marked in blue have been neatly painted over or masked off, the insulating tape can be carefully removed while the paint is still wet:

Now the paint should be left to dry for a few hours, if it has been painted.
Here is an example of how it might look when working with Kapton tape:

⧉ Phiarc

Finally, the liquid metal:

Now comes the actual task of applying the liquid metal. First, place a small “ball” from the syringe in the center of the die:

⧉ Phiarc

This is best done by carefully rubbing it in a circular motion on the die using one of the cotton swabs provided until it looks like this:

⧉ Phiarc

With time, you will get a feel for when enough has been rubbed in and when the amount is sufficient.
Then reassemble the graphics card backwards and plug it back into the computer.

Observations and conclusion:

The use of liquid metal has really paid off, especially with the 1080 Ti, but also with both cards.
The fans activate a little later because the heat is dissipated more efficiently even passively.

The biggest change is that you can hardly hear the fan until the GPU is at about 75% capacity, and even at full load, HWiNFO shows a temperature that is 5 °C lower than with the MX-4 thermal paste, which was not exactly old or bad.

If you want to divide the fan curve into 10 levels plus “off,” the card previously ran at fan level 8 at 75% load, but now runs at level 4-5 thanks to liquid metal. The temperatures at full load have improved slightly, but above all, the temperatures in the partial load range below 80% have improved significantly.

I no longer hear the 1080 Ti at all in moderately demanding games (e.g. Guild Wars 2 at highest settings, 2560 x 1080), as the load never exceeds 80-85% (this game is heavily CPU-limited) and even then only rarely. Most of the time, the load is in the 40-60% range.

I would immediately recommend this treatment to anyone who owns one of these cards or a similar one. The cards have been running with liquid metal for more than two years now and the cooling performance has not noticeably deteriorated.

Der Beitrag ASUS ROG Strix GTX 1080 und 1080 Ti – Liquid metal erschien zuerst auf flohs blog.

The reason for overclocking the graphics card in this system was purely out of curiosity to see what could be achieved. Almost any GPU less than five years old would be more powerful than the built-in Nvidia GeForce GT 1030 OC with 2 GB and also better connected (only PCIe 3.0 x4!).

The card also does not allow any increase in the power limit, so any overclocking must be satisfied with the original 30 watts. At least this GT 1030 is a version with 2 GB of fast GDDR5; there were also significantly slower variants with DDR4 as VRAM.

As always, overclocking was done with MSI Afterburner, monitored with HWiNFO, and tested for stability with Unigine Heaven.
In the end, these clock speeds were stable:

There’s not much to say, got a little more out of it – good.
We’ll continue in the next post with the conclusion and thoughts.

Der Beitrag <h5>i7-7700K OC #8: </h5><h3><b>What about the GPU?</b></h3> erschien zuerst auf flohs blog.

The following procedure has proven to work well, regardless of the newly built system:
First, check whether there are any BIOS/UEFI updates available from the motherboard manufacturer, in my case ASUS—there was a newer version (v1501). I installed this using a FAT32-formatted USB stick. Even if there had been no newer version, continue with:

Set everything in UEFI to factory settings and then to minimum settings, i.e.:

• RAM at minimum clock speed, i.e., simply leave the existing setting (presumably 2133 MHz for DDR4) unchanged.
• Processor settings all “original”
• Prepare Windows—disable CSM/Legacy, enable TPM, boot order, enable Secure Boot if desired, etc.

First, install a fresh copy of Windows, if you don’t already have one, to rule out software problems later on and avoid unnecessary background processes and legacy issues that could interfere with overclocking. I installed Windows 11 Pro 24H2, bypassing the “incompatibilities.”

Required software

Once Windows is installed, you can let the system install the drivers itself via a working Internet connection—that’s the easiest way. Then install the updates offered and restart the whole thing. Next, check in Device Manager to see if drivers have been installed for everything relevant—usually they have.

However, I always take the “old-school” approach:
Before I even install Windows, I download the latest drivers for the components I am using from the manufacturer’s websites (in this case, ASUS, Nvidia, and Intel). Then I install Windows offline and, once that is complete, I install the drivers I downloaded earlier, also offline.

This has the advantage of usually resulting in newer driver versions than those offered by Windows Update. Sometimes the manufacturer’s websites also contain additional information, such as notes on current or past problems, etc., which you would never see if you used the automatic method.

I used the following software to overclock the CPU and GPU:

• HWiNFO – For monitoring
• Prime95 – For load generation, optionally with or without AVX / AVX2
• Cinebench R23 – For generating faults and testing differences
• ASUS ROG RealBench – To generate load in order to test stability
• Unigine Heaven – To generate load on the GPU in order to test stability
• MSI Afterburner – For overclocking the GPU
• MemTest86 – To generate faults on the RAM and memory controller in order to test stability

My goal is to achieve stability in ASUS ROG RealBench, as I believe it reflects moderate to heavy everyday use equally well.

Processor overclocking

Actually, the whole thing isn’t complicated, but you need to have a lot of time and patience.

I first overclocked the i7-7700K statically, ensured stability, and repeatedly pushed it to its limits (voltage, temperatures). Since my goal was 5.00 GHz on all four cores with hyperthreading enabled, I started with that—the delidding I had done beforehand should have made this feasible in terms of temperature.

So I simply set the core voltage to 1.30 volts in the BIOS and set the “CPU Core Ratio” for “All Cores” to “50”. Then I saved, booted Windows, and started HWiNFO / Prime95 with AVX and AVX2 and let it run.

First observation: It runs stably for at least 1-2 hours, but gets quite warm (around 80-85 °C).
Then back into the UEFI and down with the voltage, now to 1.25 volts – then tested again.
Temperatures have dropped to just under 80 °C, still everything stable.

I did this until I reached 1.18 volts, at which point I had my first “quick” blue screen.
So I increased it again to 1.22 volts. It then ran for many hours, stable with acceptable temperatures (still ~ 80 °C). Without delidding, even that would have been either barely possible or not possible at all; I’m at the limit.

Next, I ensured stability with ROG RealBench, which is a realistic everyday measure for me; it has to be stable in the end. It was, it runs smoothly and, of course, cooler; I saw a maximum of 75 °C on the hottest core.

Great, I would be satisfied with that in principle.
But now, of course, I want to know what will limit me in the end:
The cooling or the voltage?

How far will it go?

Then I dared to increase the clock speed, setting the multiplier to “51,” i.e., 5.10 GHz. I also adjusted the voltage to 1.30 volts to match. Lo and behold: Prime95 with AVX quickly forces the cooler to its knees thermally, the first thermal throttling kicks in, and later the first blue screen appears (well over 100 °C!).

So I continued with 1.28 volts. Same result, thermal throttling—but later. No blue screen.
But you can already see the limit with AVX, 5.00–5.10 GHz. Then the air cooler, which otherwise does a really good job and is also tolerably quiet, reaches its limit.

After only a few minutes of ROG RealBench, a blue screen appears.
The journey with AVX seems to be over here, from now on I’ll optimize the system for “stable everyday use with a little AVX.”

And I’ve still come a long way:

The final result:

First of all: I made it very easy for myself with the RAM. I simply loaded the stored XMP profile for 2400 MHz 16-16-16-39 at 1.20 volts, slightly increased the DRAM voltage to 1.2540 volts, and set the memory frequency to 3000 MHz, nothing else!

The whole thing ran for days with Memtest86 without any errors. That was a 25% performance boost for free.
Above 3000 MHz, errors occurred that would have required very time-consuming fine-tuning to fix, but that wasn’t my focus.

For the processor, I switched from static overclocking at 5.20 GHz with over 1.35 volts to adaptive overclocking. 5.30 GHz on all cores got too hot – it took a few minutes, but throttling kicked in.

In the end, of course, I spent hours experimenting and days testing to see what worked best and how close I could get to the limits.

The CPU now runs as follows:
Load on 1–2 cores = 5.30 GHz at LLC 6 with adaptive auto voltage, slightly above 1.42 volts in HWiNFO under load.
Load on 3–4 cores = 5.20 GHz at LLC 6, also adaptive, then 1.385 volts under load.

Cache = 4.90 GHz, above that it quickly became very unstable.

Without AVX, I am therefore at both the temperature and voltage limits; it would not be possible to go any further without entering even more dangerous voltage regions. I find 1.42 volts to be borderline; my limit was actually 1.40 volts.

I couldn’t use an AVX offset without the turbo and power-saving mechanisms only working halfway afterwards, or the single-core clock speeds never being reached, as Windows 11 seems to use AVX instructions quite often. I wanted to achieve a dynamic overclock that is also capable of downclocking in idle mode to save heat and power – hence the use of an adaptive “auto” offset in the first place.

Conclusion:

This inexpensive cooler packs a punch – without AVX, I can push the CPU to almost its maximum capacity. At the same time, the board’s power supply is so good that I have never experienced any problems in this regard (I didn’t have any additional airflow over the VRM coolers!).

I would describe the clock speeds as very high, achieved at the expense of equally high voltage. Here is a comparison of the changes:

Cinebench and PassMark results:

First, Cinebench R23 with single-core overview – 1423 points.
Then with multi-core overview – 7112 points.

Finally, Cinebench R24 – 86 points or 419 in multi-core.

When compared to other CPU models, it is easy to see how far the i7-7700K has risen, especially in terms of single-core performance. Temperatures were consistently around 90 °C while Cinebench was running.

Click here for the PassMark PerformanceTest v11.1 Baseline.

What about the graphics output?
We’ll continue with that in the next post.

Der Beitrag <h5>i7-7700K OC #7: </h5><h3><b>Now the overclocking can finally begin</b></h3> erschien zuerst auf flohs blog.