Test hardware and methodology
But first I would like to talk a little bit about the used test hardware and the experiences we could gather in the last 3 weeks with the i7 11700K CPU already before the launch. Although existing motherboards with the LGA1200 socket and for example the Z490 chipset are compatible with Rocket Lake, of course new microcode releases and BIOS optimizations always reach the current generation boards first. That’s why I would like to thank ASRock for providing us with a latest generation motherboard in the form of the Z590 PG Velocita, which is used for all tests in today’s article.
This board is a mid-range motherboard from ASRock’s Z590 overclocking segment, with two PCI-Express 4.0 16x slots to the CPU, two 2.5GbE NICs, three M.2 slots, integrated Wi-Fi 6E, plenty of USB and SATA IO. Futuristic design and a large portion of RGB, both on-board and in the form of addressable headers of both variants, may not be missing, of course. The start and reset buttons along with the postcode display make fine tuning RAM overclocks much easier and there’s even a ClearCMOS button on the board just in case.
The power supply consists of 14 SIC654 Dr.MOS 50 A powerstages from Vishay divided into 12 phases for CPU cores and cache, and 2 phases for the iGPU. The PWM controller is an ISL69269 with six ISL6617A doublers. So our Intel rocket won’t be lacking in power supply even when overclocking. Equally important for today’s test are the 4 daisy-chained DIMM slots, which should be representative of most boards in this feature and price segment, and with which we will test the various configuration options.
Speaking of configurations, we used common XMP timings and clock speeds from the DDR4 midrange in our selection, starting with DDR4-3200 CL14 all the way up to DDR4-4000 CL18. This should allow us to cover most actual usecases in the context of a Rocket Lake system, because below DDR4-3200 is JEDEC territory where a Rocket Lake (hopefully) has no business being, and above that we get into price segments for RAM kits that would exceed the acquisition cost of the CPU.
To test the influence of gearing, both DDR4-3200 and DDR4-3600 clock speeds were tested in Gear 1 and Gear 2 to draw a direct comparison between the modes. Furthermore, for DDR4-3200 and DDR4-4000, in addition to the usual single-rank, dual-rank setups were also run through our tests in order to make any effects of the gearing visible. All configurations are of course dual-channel and the rank count refers to the ranks per channel:
- DDR4-3200 14-14-14-34, Gear 1, Single-Rank
- DDR4-3200 14-14-14-34, Gear 1, Dual-Rank
- DDR4-3200 14-14-14-34, Gear 2, Single-Rank
- DDR4-3200 14-14-14-34, Gear 2, Dual-Rank
- DDR4-3600 16-16-36, Gear 1, Single-Rank
- DDR4-3600 16-16-36, Gear 2, Single-Rank
- DDR4-4000 18-18-18-38, Gear 2, Single Rank
- DDR4-4000 18-18-18-38, Gear 2, Dual-Rank
All settings were run with XMP-usual 1.5 V DRAM voltage. For better clarity the “DDR4-” prefix of the clock rates was omitted and only Dual-Rank with “DR” was shown for the ranks. To test as representative as possible of an XMP Usecase, only clock rate, primary timings and voltage were set, secondary and tertiary timings were left on auto. If you were hoping for a high-end configuration at the limit of what is currently possible with Rocket Lake, you won’t go away empty-handed either though, see the conclusion on the last page. 😉
Some of you may have noticed that there is no Command Rate listed. For good reason, because 2T is automatically set in Gear 1 and 1T is automatically set in Gear 2, this can’t be changed either or has no effect on performance, at least with microcode and BIOS versions at this point. Resourceful math geniuses will have noticed: 1 x 2T = 2 x 1T = 2T. So no matter which gear we are in, the effective command rate remains 2T. Of course, this makes testing easier and more consistent, but the enthusiast in me misses a true 1T command rate already, especially because of the sometimes much better benchmark performance.
As always, a complete listing of the test hardware follows:
Test system | |
---|---|
Hardware: |
|
Cooling: |
|
Housing: |
|
Periphery: |
|
Since my particular CPU is closer to a potato than a rocket, a static all-core OC would have resulted in performance degradation. Accordingly, the CPU is operated in the standard Turbo Boost Max 3.0 mode and all boost limits are removed. This then results in an effective 4.6 – 4.9 GHz depending on the load on all cores or 5.0 GHz on 2 cores, both permanently. The cache or ring was also run at the standard clock rate, namely 4.0 GHz.
65 Antworten
Kommentar
Lade neue Kommentare
Moderator
Mitglied
Moderator
Veteran
Mitglied
Mitglied
Veteran
Mitglied
Mitglied
Veteran
Mitglied
Veteran
Neuling
Urgestein
Mitglied
Mitglied
Urgestein
Alle Kommentare lesen unter igor´sLAB Community →