Rocket Lake Launch OC Event Las Vegas
Intel sent a special chip from the OC Lab in Oregon to be used in testing for a small gathering of overclockers in Las Vegas.
Over the course of two weeks we blasted though 500 liters of LN2 pushing the CPU at maximum speed. We set many world records pre-launch and had the opportunity to test lots of new products. The benchmark selected were PCIe bandwidth, memory bandwidth, and system performance oriented. Most tests were long duration tests like PCMark and SysMark where it can take from 15 minutes to over two hours to complete each run. In short, what we found was this processor design will be setting many new world records in 2D, 3D, encoding, encryption, decryption, and both memory and SSD bandwidth.
|
The 11900K OC Lab Chip
Huge thanks to Intel for a chip to test under extreme conditions, this chip was overclocked LN2 to test the Rocket Lake platform.
The first thing I did when the CPU arrived was lap it to 1000 grit on glass, switching back a few times in grit to find the dips. The IHS quality is good but room for improvement that voids the warranty. This added about +100Mhz usable to the top speed of the CPU.
System Performance
My first goal was to take back PCMark 10 Express from AMD and set the bar out of reach for now. This was one bench that Intel can take the lead leveraging the strengths of Rocket Lake, you will see nice gains in gaming especially in older games.
This benchmark is 15 minutes in duration and chances are you will need to re-run it to get the score you want.
|
Bapco CrossMark is a new benchmark released shortly before the launch of Rocket Lake, Crossmark is a shorter version of Sysmark to measure overall system performance. Rocket Lake had no problem beating out the 16 core 32 thread parts taking first place.
BAPCo - CrossMark Result Detail
BAPCo - CrossMark Result Detail
PCIe Gen4 Bandwidth
I ran the only SSD test on the bot and I posted up new numbers in a new area code.
Memory Bandwidth
Rocket Lake has stepped up IMC performance and IPC to deliver stronger results than previous generations of MSDT platforms and it is safe to say that Rocket Lake delivers performance that was normally found in HEDT quad channel platforms.
PLL voltage
In order to get the CPU to withstand extreme temperatures and even "full pot" the OC Lab chip only only required 1.6v to run full pot with ease. The review sample 11900K was able to hit full pot without touching PLL but was a little touchy on full pot pours. The 11600K did not need PLL at all for full pot. This was a noticeable improvement in this design.
The IMC
Vastly improved and this alone will bring new interest in performance tuning on Intel again. The ability to run 5.4Ghz Memory with tight timing is not only possible but I can do it with Ycruncher 1B stability with super tight timings. Both Hynix DJR and Sansung B die are far more fun on Z590 as you will see. You will see that the dual rank memory modules are very strong on Z590 while the Hynix DJR modules are a bit more fun to tune testing towards 6Ghz. You will find memory ratios have changed a bit with some memory ratios greyed out in some bios.
Built for Speed
If you are not familiar with SVID, let me introduce you to the easiest CPU overclocking you have ever seen to date, the voltages will scale to your cooling performance without moving from "auto" or any funky test cycle. SVID automatically sets voltages on the fly under different load conditions far better and in much wider range of use including overclocking on LN2. You just set the CPU and cache multiplier to what you know should work, if your cooling can handle the setting the machine will do it. No need to fiddle with voltages and just try a lower multiple until you find what works.
Testing on LN2 I found this to work all the way to 6.4Ghz no problem, there was no issues with stability with and without hyperthreading enabled.
Testing on LN2 I found this to work all the way to 6.4Ghz no problem, there was no issues with stability with and without hyperthreading enabled.
Hyperthreading
I test with and without HT enabled as my cascade can only hand a few hundred watts of load. After working with this chip a week I can see that this architecture likes HT enabled a bit more. You will find that your max speed with and without HT enabled is closer together.
BCLK Tolerance
I found that the usable range has increased from Comet Lake and scaling is on point. We will see a lot more BCLK adjustments due to the nice scaling and broader usable range. I tested memory frequencies out of reach and brining the BCLK down below 100 to work fine.
AIO Performance
One thing I did not expect was to be running Rocket Lake to be as AIO friendly as it is. We see a few reviews and social media posts that Rocket Lake is hot or not worth the sand to make but that is fake news to instill doubt. I was sent a few AIO coolers and I really like being able to run at 5.5Ghz 24/7 and I connect the AIO on my bench rig to an AC unit for maximum performance where I can run 5.7Ghz. Lapping the CPU helps a lot, I gained 100+ Mhz usable but this voids the warranty.