6.9 GHz i9-9900KS cores with all 8 cores active, here's how

6.9 GHz i9-9900KS cores with all 8 cores active, here's how

It's not a secret: the Core i9-9900KS is the fastest consumer CPU ever created by Intel, a version of the 9900K which has been carefully selected to include only the best cores produced in the factory. Every 9900KS is able to work at 5 GHz on all eight cores and generally with overclocking you can reach i 5.2 GHz. A result that may seem a no brainer compared to what can be achieved with the best 9900KS and the liquid nitrogen cooling.

To show what its chip is capable of, Intel sent Tom's Hardware USA colleagues a sample of Core i9-9900KS selected to perform some overclocking tests. The expert hands of the well-known overclocker have ventured into this venture Allen 'Splave' Golibersuch, which managed to reach a frequency of well 6.95 GHz on all cores and 7.3 GHz on a single thread. Below we see how he did it and what obstacles he had to overcome.

NOTE: Intel previously tested the chip internally to test its overclocking potential. Although Intel does not guarantee the frequencies reached in overclocking, the company believes that the results on this selected sample will exceed those of a “generic” sample.

In this test we will see what a selected 9900KS is capable of and what advantages it brings – if it does – that S more than a 9900K. Let's start by analyzing the physical differences. In an overclock, one of the key factors is undoubtedly the temperature: the thermal transfer from the integrated heat spreader (IHS) to the heat sink plays a fundamental role.

Thermal paste fills all the gaps between the heatsink and the IHS, and it is necessary to use it if you want to maintain the guarantee, as the heat generated could permanently damage the processor. If though both surfaces are flattened so that the thermal paste layer is as thin and uniform as possible, tangible improvements in terms of temperatures can be obtained.

Given the rather limited test sample (three Core i9-9900KS and 20 Core i9-9900K), the average die temperature on the Core i9-9900KS seemed to be significantly lower on the 9900KS compared to the tested 9900K. Of the three 9900KS, two were purchased in-store, but nevertheless the selected sample provided by Intel showed the same thermal characteristics.

Read This Now:   Intel CEO visited TSMC to secure more space on its production lines

For a test of this type, the first move is always to go to flatten (lapping) the IHS by bleaching it on a real flat surface. Since the toilet used by Splave is flat, a flat IHS provides a better contact surface for heat exchange. Some CPU waterblocks are arc-shaped and many others are concave, so not optimal for these tests. Sometimes it is a design choice, while other times it is the “fault” of the thicker nickel plating on the edges of the cooling surface.

In the photo below there are 4 samples of 9900K stepping P0 that have the same model on the heat spread. You can see the highest points, where copper emerges, while the nickel parts are the lowest. All four have high edges and are concave in the middle. This involves the use of a thicker layer of thermal paste in the medium, but as already mentioned above, the greater thickness does not help.

The 9900KS stepping R0, as you can see in the photo below, is much more uniform with only a few slightly lower edges. In this regard, it must be said that all three samples were identical from this point of view. Normally, flattening the IHS manages to reduce temperatures up to 4-5 ° C on the warmer core, as well as facilitating a more uniform diffusion of the temperature between the dies.

Below is what the finished product looks like. No advantage emerged when using the 9900KS with a liquid heatsink. As already mentioned, the chip sample is too limited and this only allows speculating that there has been a factory improvement of the heat spreader on the 9900K.

After an assembly test of the toilet, you can see how the layer of thermal paste is thin and very uniform. A perfect result!

A noteworthy point about these chips is the one regarding new microcode which serves as the basis for many CPU features. By installing a CPU stepping P0 (9900KF, 9900K), the BIOS detected the version A0 of the microcode. By inserting the 9900KS, however, the microcode has changed to B4. What does this mean?

Read This Now:   Test - NZXT Grid+ V2: |Specs | CPU | Hashrate | Review | Config

Given that microcode is the basis for the functioning of a processor, companies do not like to talk about how it works and do not share much information about it. However, it immediately became clear, already from the preparations, that something had changed.

Before you can use liquid nitrogen and possibly aim for the world record, there are some objectives to hit, intermediate results to be achieved in tests such as Geekbench3 or Cinebench R15. However, the first scores obtained using the air cooling system with the 9900KS were lower than expected. Where was the problem?

Digging deeper, Splave swapped the 9900KS again for a 9900K stepping P0, finally finding correct scores. Apparently there was something wrong with R0 stepping.

The next step was therefore to disable mitigation from the operating system for Specter is Meltdown for optimal performance, a trick well known to overclockers to – in theory – gain performance. Unfortunately, after doing so, there was no performance increase with the R0 chip stepping. Why? The answer is obvious: now it is the microcode that manages the mitigation, not the operating system.

If on the one hand it is an excellent thing for the end consumer, this creates an additional obstacle for competitive overclockers who play records for tenths of a second. This being the case, the 9900KS benchmark did not start in the best way. Nonetheless, the processor is among the fastest in circulation and therefore capable of achieving more than relevant results.

Intel's selected 9900KS has concluded 5.4 GHz Cinebench R15 with a 1.27V vCore and liquid cooling! Against all initial expectations, it was possible to continue to lower the operating voltage starting from what was thought to be a more than decent setting of 5.4 GHz with a vCore of 1.35V. What is lost due to microcode could be “mitigated” by the extra frequency.

Read This Now:   how to monitor your hardware with this software

The other two 9900KS purchased in the store also did not disappoint. The first one has achieved 5.4 GHz with a vCore of 1.32V, The second stopped at 5.3 GHz with a vCore of 1.23V. A nice show of strength.

Splave then inserted the best chip (the sample supplied by Intel) on the test bench, cooling it with liquid nitrogen. For the occasion, he used a motherboard ASRock Z390 Phantom Gaming X (Amazon) together with memory B-Die SPD 4800 from Team Group.

To obtain better temperatures, thermal paste was used Thermal Grizzly Kryonaut LHE-edition (Amazon) and a custom toilet from bigblock990 (another world famous overclocker). The system was powered by a single 1250 W Enermax Maxtytan (Amazon), more than enough since the system peaked at around 650 W under nitrogen.

The result was extraordinary: Splave obtained a Gold in Cinebench R15 with a frequency of 6952 MHz, getting a score of 3122 points. As for single-thread tests, Splave has come up with a frequency of well 7300 MHz during superpi32m and pifast. You can see all the Splave test results on the 9900KS on the HWBot website, at this address.

The 9900KS is exactly what I thought it was: a carefully selected 9900K, with a large OC margin which adds to an already high stock working frequency “, said Splave. “If I had the opportunity to pay a little more for chips with better performance like this, I would always do so. The fact that Intel has been able to bring this type of performance out of an existing design makes me very curious to see what the company will be able to do with the new 10-core chips coming next year “.


Notice: ob_end_flush(): failed to send buffer of zlib output compression (1) in /home/gamefeve/bitcoinminershashrate.com/wp-includes/functions.php on line 5373

Notice: ob_end_flush(): failed to send buffer of zlib output compression (1) in /home/gamefeve/bitcoinminershashrate.com/wp-includes/functions.php on line 5373