The Intel Xeon W-3200 series spans nine SKUs ranging from eight to twenty-eight cores. Like server counterparts, the SKUs utilize the Intel LGA3647 socket, with a key difference: they are single socket only. The Xeon W line is designed for workstations as well as some server applications that require higher clock speeds and lower densities. In our SKU list and value analysis piece, we are going to look at how the lineup scales with price.
Intel Xeon W-3200 SKU List and Value Analysis
The LGA3647 chips have six DDR4 memory channels and up to 28 cores like the Intel Xeon Scalable series but have 64 PCIe lanes exposed to the motherboard. As a result, there are no UPI paths for socket-to-socket communication making these higher PCIe count chips than their mainstream Xeon counterparts. Here is the SKU list and value analysis:
As a quick note, there are “M” parts that support up to 2TB of memory. One can get over 1TB using DDR4 DIMMs in up to twelve DIMMs per socket so Intel has high memory parts with their associated premiums.
We are going to quickly note that Intel offered the Xeon W-3175X which is unlocked for overclocking but only has a 512GB memory capacity and 48x PCIe 3.0 lanes. These chips are an interesting value proposition for $2,999 requiring special motherboards. The Xeon W-3200 series is based on the newer Cascade Lake generation of cores, while the Xeon W-3175X is based on the older Skylake generation.
Final Words
Overall, these are cool chips and we can certainly see the appeal of a single socket 28-core, high clock speed, Intel Xeon W-3275 solution as an example. Further, the 28-core high-memory SKUs are some of the best combinations of cores, clock speeds, and memory channels on the market today.
Compared to the 7002P series it looks extremely expensive.
https://www.servethehome.com/wp-content/uploads/2019/08/AMD-EPYC-7002-SKU-List-and-Value-Comparison-1P.jpg
Intel going nuts on price. AMD is only choice.
Very disappointing article!
No mention/analysis of the real value proposition compared against the EPYC 7002 series makes it feel like it was written by the Intel marketing department. I expected better from STH.
I come here for enterprise coverage because I am a professional. So far the comment section has been pretty nice. Please dear baby jesus don’t let this comment section turn into Anandtech’s. People yelling about how the article was unfair toward a company they have an emotional attachment to for whatever reason. It’s not like enterprise customers are paying the list prices for these parts.
Thank you for the article, it was brief and delivered the information needed to address a new product release. I personally like the fact the article speaks specifically to the product of topic. I’d like to think that given the facts, most readers here are intelligent enough to make product decisions based on their needs. I also appreciate the info on the W-3175X, as I was unaware it had the memory restriction and fewer PCIe lanes, always nice to learn something new…
Didn’t Intel release some single socket Gen 2 scalable SKUs recently that were much better value then these?
I guess they don’t have 4.4ghz turbo, but paying 50% premium for 10% turbo clock is a bit nuts.
Intel keep moving the single socket workstation line up and up the price ladder every release, i would have expected a little less this time going up against AMD, guess HP & Dell will build them anyway and the pro market will buy them.
Nice to see them finally kill off the 4/6 core parts, along with their price points though…
95% of workstations are only going to use the entry level chip though, 4ghz is a bit shitty….
They’ve done this kind of article for a half decade or more. Now people say it is disappointing?
If you’re looking for higher clocks per core, these are a better option than EPYC still. 64 lanes is interesting too
Desperetely agree with Ryan Metz. Don’t turn into Anandtech. :(
Fully agree with both Ryan Metz and Frick. Anandtech has gone nuts; please do your best to keep the comments here sane.
Hi guys. BinkyTO – I get it. This article has been pushed and was scheduled to go live well before the EPYC 7002 series launched. Still, these are targeting higher clock speed applications. If you are per-core licensed on AVX-512 enabled hardware, they make quite a bit of sense. William will be testing the W-3275 starting this week. I actually think this is a really interesting platform proposition.
It’s getting to complicated and to hard to manage. UP Xeon can now be placed in: LGA 3647, FCBGA 2518, LGA 2066, LGA 1151, LGA 2011. It gets way to complicated to balance things and upgrade servers, keeping spare RAM etc.
Getting the processor you choose from intel is problematic becouse intel is facing the same problems with diferent SKU.
And the price/performance Intel/AMD is somthing diferent.
I like that the WS parts went to LGA 3647, for me it was strange that the previous parts were LGA 2066. I liked a lot that the server E5-26xx were interchangeable with the E5-16xx.
This also looks good for a down level to HEDT line up