Sierra having that big of shipments is a surprise. And Emerald being less than Sapphire is also a surprise. After all, it's basically the same chip but better.
Course the real question is why are they still buying Intel at this point.
Course the real question is why are they still buying Intel at this point.
Price SPR/EMR are cheap very cheap.
AS GNR/SRF Ramps and also 18A Comes online Foundry should break even by 2027.
Last edited:
DrMrLordX
Lifer
- Apr 27, 2000
- 22,898
- 12,963
- 136
That is very strange. Maybe Intel had a backlog of chips to sell and moved inventory at low prices, essentially undercutting Emerald Rapids. Also there may be some lower core-count SKUs that are only available as Sapphire Rapids.
I think these are estimates. I seem to recall another breakdown that showed had SRF and GNR lower than this chart.
If someone didn't notice Ice Lake is outselling GNR/SRF lmfao.
Turin as well
Turin as well
Last edited:
Ice lake must have lots of stock and Intel is selling them out for cheap or maybe they had some contracts so they are pumping ICL.
SPR is fine for less than 48C and EMR UpTo 64C after that It's Zen 4/GNR/Turin/SRF.
Too lazy for thatNot helpful
I thought you had meticulously prepared spreadsheets.
I work for a Fortune 500 and our decision to purchase -> deciding what server to order -> delivery is about 8 weeks at most.
Ice Lake is popular because most low to mid range Ice Lake Xeons actually exceed the computation requirements of these servers.
Very few corporations need massive on-prem compute. Ice Lake Xeons are cheap, plentiful, and good enough.
I wish they had been when I had nothing but a stupid Cascade Lake server available in my company's tolerable price range
What I mean is that if a Corpo buys a server, they might buy more of the exact same server instead of buying something with a newer processor for a period of time.
I'm sure you could replace a bunch of low-mid Ice Lake servers with a midrange Turin. And save a bunch of power too.
You could do that with GNR as well it's just Corpo being Dumb 🥴.
You could, but depends on requirements too. Our requirements are lots of lower-end, geographically dispersed servers to serve large files over LAN to branch offices because they always complain about when they need to open 500MB files when hosted off-site.
Our two primary DCs run just fine on 2x 8 core Ice Lake servers running in pairs. We've also been spinning up a lot of VMs in Azure. A lot easier because we want all authentication to filter through Entra Conditional Access policies. We'd never downscale to a single sever in our primary DCs because then we'd lose redundancy. Goal is to have servers in pairs never exceeding 50% utilization. I'll provision 8 vCPUs per physical core no problem and can easily run 2 dozen VMs on 16 cores.
Point being, even with fairly low end Ice Lake, our primary bottleneck continues to be RAM and storage IO.
I feel like Mid-Range server is the weak spot of the market. Either you get hyperscalers wanting to maximize compute density per rack-U, or you get corpos looking for low cost local compute options, filling the volume segment.
What really matters most, at least from my personal experience, is cost, delivery time frame, and OEM options. Turin might be better, but my supplier has dozens of Ice Lake servers of varying RAM/storage/physical size configs at a low cost with sub-1 week delivery times.
Last edited:
Supermicro A+ Server 1015A-MT (AS-1015A-MT)
Supermicro A+ Server 1015A-MT is a 1U Rack Server with up to 128GB DDR5 ECC UDIMM memory, and 4x 2.5" SATA Drives.
www.thinkmate.com
Pretty cheap with 128GB RAM and 1.92TB SSD
DrMrLordX
Lifer
- Apr 27, 2000
- 22,898
- 12,963
- 136
About that . . .
Yeah I was thinking the same thing. And even if the many Ice Lake-SP systems are dispersed over large geographical areas, there are lower core-count/lower clockspeed Turin products that could slot in just as easily at a better TCO. TCO matters, especially if these systems are going to be in place for long periods of time. The only advantage for the Ice Lake-SP systems are availability and cost of entry.
See above, in terms of TCO, a newer/better uarch + process should yield superior results.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 22K
-
News NVIDIA and Intel to Develop AI Infrastructure and Personal Computing Products- Started by poke01
- Replies: 384
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
Facebook
Twitter