Wednesday, September 8th 2021
Intel Intros Xeon E-2300 Series Based on 14 nm "Rocket Lake-E"
Intel today released the Xeon E-2300 series enterprise processors for entry level servers, based on the 14 nm "Rocket Lake-E" silicon. These are slightly different from the Xeon W-1300 series processors targeting workstations. The E-2300 has a more server-relevant feature-set, and is designed for high uptime. You get ECC memory support, as well as vPro, SGX, and MPX (memory encryption). You still get only client-relevant AVX-512 instructions found in 11th Gen Core processors, as well as DLBoost AI acceleration.
The "Rocket Lake-E" silicon comes with up to 8 "Cypress Cove" CPU cores, each with 512 KB of dedicated L2 cache, and 16 MB of shared L3 cache. The processor features a 2-channel DDR4 memory interface that supports up to 128 GB of DDR4-3200 memory. It puts out 20 PCI-Express 4.0 lanes that can be segmented in a number of ways; as well as the Intel C250 series chipset puts out 24 PCI-Express 3.0 lanes. Chips in this series come with TDP of up to 95 W.
Source:
ServeTheHome
The "Rocket Lake-E" silicon comes with up to 8 "Cypress Cove" CPU cores, each with 512 KB of dedicated L2 cache, and 16 MB of shared L3 cache. The processor features a 2-channel DDR4 memory interface that supports up to 128 GB of DDR4-3200 memory. It puts out 20 PCI-Express 4.0 lanes that can be segmented in a number of ways; as well as the Intel C250 series chipset puts out 24 PCI-Express 3.0 lanes. Chips in this series come with TDP of up to 95 W.
14 Comments on Intel Intros Xeon E-2300 Series Based on 14 nm "Rocket Lake-E"
Bearing in mind that Alder Lake is supposedly released soon - how long until the Alder Lake Xeon's get released? 3-6 Months? I think you would have to be really desperate to buy these...
But apparently not:shadedshu:
And in DC's you pay for power.
Of course it all depends on the motherboard's power limits settings. Gaming / consumer motherboards will typically have high default values and sometimes unlimited short-term limits, but these can be tweaked. Server motherboards will be more likely to have defaults within Intel [recommended] specifications. You could also set the limits manually to 95W or disable PL2 so that it will not go over 95W no matter the load.
I always find it funny when motherboard vendors in consumers come up with the most ridiculous designs in relation of VRM power; while 99% of those users running such a board won't be even close to the peak power of what that VRM is able to provide. In my experience a VRM switching frequency is far more important (higher the better) then a massively overbuild VRM if you relate to overclocking. For normaly daily use any board is sufficient. Hence why the 50$ boards reviewed on TPU here can run a 3950x or so even in overclocked conditions.
Any board vendor has to comply to AMD's minimum recommendations. So you'll find the same from low to high-end, and youll notice that the 50$ boards are just as capable of running a CPU as a high end board. Overclocking is indeed a different story, VRM temps too, but not worth the 250 up to 400$ extra price lol.
If I need a "Real" server CPU I would buy a server class CPU with tons of PCIE lanes and SAS.
If I could live up with less PCIE and no SAS in a server, I would buy 16 core Ryzen and plug it in a server MB.
That's just seeing a very small part of the problem, since Intel Rocket Lake processors are inefficient compared to the AMD offerings also at nominal TDP (PL1), although desktop-part reviews have not focused much on this.