Intel Sapphire Rapids-SP Xeon Platinum ES CPUs Fail To Impress Against EPYC Milan (X) & Last-Gen Xeons In Leaked Benchmarks


Intel's Sapphire Rapids-SP Xeon Platinum CPUs that leaked out yesterday have now been tested in a range of benchmarks but fail to impress when compared to AMD's EPYC & even older Intel Xeon chips.

Intel Sapphire Rapids-SP Xeon Platinum ES CPUs Tested In Dual Socket Configuration, Lackluster Performance So Far Against AMD's EPYC CPUs

In yesterday's leak, we reported how the latest Intel Sapphire Rapids-SP Xeon Platinum CPUs in ES state consume 350W of power at PL1 and 420W of power at PL2 limits. The CPU itself, which is a Xeon Platinum 8476 or Xeon Platinum 8480, is a 56 core part with 112 threads based on the Golden Cove core architecture, & and comes with 112 MB of L2 & 105 MB of L3 cache, and features clock speeds of up to 3.7 GHz (3.3 GHz all-core). Now, these aren't the final clocks but the power figures are definitely worrisome for Intel's upcoming part.

Intel Arc Graphics Cards Get ‘VRAM Self-Refresh’ Feature In Latest Linux Drivers

Yuuki_Ans has now provided benchmarks for the same Intel Sapphire Rapids-SP ES CPU against Intel's Cascade Lake Xeon 8280L (8S Config), Ice Lake Xeon 8380 (2S Config), EPYC 7763 (2P config), and EPYC 7773X (2P config). The Sapphire Rapids-SP chips were also running a dual-socket config with 64 GB of DDR5-4800 memory. The rest of the platforms were running DDR4 platforms at the max specs that the platforms can support, respectively.

Intel's upcoming Sapphire Rapids-SP ES CPU has been tested in various benchmarks. (Image Credits: Yuuki_Ans)

As for the benchmarks, In single-core, the Intel Sapphire Rapids-SP CPU did show a little boost over the Ice Lake-SP chips but was beaten by AMD's EPYC Milan parts in 4 out of 6 benchmarks. Despite the new Golden Cove architecture, the CPU was only able to match the Zen 3 server chips in 1T CPU benches. In multi-threaded benchmarks, the AMD EPYC Milan lineup destroyed the Intel Sapphire Rapids-SP chips outright. Just for comparison's sake, in CPU-z, the AMD EPYC Milan CPU ran at up to 2.3x faster performance than the unreleased Intel chip.

Interestingly, the CPU does deliver much higher performance in the CPU-z AVX-512 tests but we know that AVX-512 makes the CPU consume way more power and that could lead to power figures nearing the 700-750W BIOS limit. This is also why the AVX-512 clock speeds are rated much lower due to their intensity on each individual core. AVX-512 has been used in the past by Intel to inflate their benchmark numbers and while there are certain HPC workloads where the tech is useful, it isn't widely adopted and as such, non-AVX-512 workloads will benefit from standard core compute performance more.

Although this is an ES chip and we expect performance to improve significantly for the Sapphire Rapids-SP chip in the QS state, the final chip may still be only competitive against AMD's EPYC Milan parts while it will be launching at the same time when AMD will come out guns blazing with its next-generation EPYC Genoa 7004 CPUs. With such high power draw and not enough performance at hand, it looks like Intel will be getting a serious beating in the server segment by AMD, something that has been the case over the past few generations.

Intel Xeon SP Families (Preliminary):

Family BrandingSkylake-SPCascade Lake-SP/APCooper Lake-SPIce Lake-SPSapphire RapidsEmerald RapidsGranite RapidsDiamond Rapids
Process Node14nm+14nm++14nm++10nm+Intel 7Intel 7Intel 3Intel 3?
Platform NameIntel PurleyIntel PurleyIntel Cedar IslandIntel WhitleyIntel Eagle StreamIntel Eagle StreamIntel Mountain Stream
Intel Birch Stream
Intel Mountain Stream
Intel Birch Stream
Core ArchitectureSkylakeCascade LakeCascade LakeSunny CoveGolden CoveRaptor CoveRedwood Cove?Lion Cove?
IPC Improvement (Vs Prev Gen)10%0%0%20%19%8%?35%?39%?
MCP (Multi-Chip Package) SKUsNoYesNoNoYesYesTBD (Possibly Yes)TBD (Possibly Yes)
SocketLGA 3647LGA 3647LGA 4189LGA 4189LGA 4677LGA 4677TBDTBD
Max Core CountUp To 28Up To 28Up To 28Up To 40Up To 56Up To 64?Up To 120?Up To 144?
Max Thread CountUp To 56Up To 56Up To 56Up To 80Up To 112Up To 128?Up To 240?Up To 288?
Max L3 Cache38.5 MB L338.5 MB L338.5 MB L360 MB L3105 MB L3120 MB L3?240 MB L3?288 MB L3?
Vector EnginesAVX-512/FMA2AVX-512/FMA2AVX-512/FMA2AVX-512/FMA2AVX-512/FMA2AVX-512/FMA2AVX-1024/FMA3?AVX-1024/FMA3?
Memory SupportDDR4-2666 6-ChannelDDR4-2933 6-ChannelUp To 6-Channel DDR4-3200Up To 8-Channel DDR4-3200Up To 8-Channel DDR5-4800Up To 8-Channel DDR5-5600?Up To 12-Channel DDR5-6400?Up To 12-Channel DDR6-7200?
PCIe Gen SupportPCIe 3.0 (48 Lanes)PCIe 3.0 (48 Lanes)PCIe 3.0 (48 Lanes)PCIe 4.0 (64 Lanes)PCIe 5.0 (80 lanes)PCIe 5.0 (80 Lanes)PCIe 6.0 (128 Lanes)?PCIe 6.0 (128 Lanes)?
TDP Range (PL1)140W-205W165W-205W150W-250W105-270WUp To 350WUp To 375W?Up To 400W?Up To 425W?
3D Xpoint Optane DIMMN/AApache PassBarlow PassBarlow PassCrow PassCrow Pass?Donahue Pass?Donahue Pass?
CompetitionAMD EPYC Naples 14nmAMD EPYC Rome 7nmAMD EPYC Rome 7nmAMD EPYC Milan 7nm+AMD EPYC Genoa ~5nmAMD Next-Gen EPYC (Post Genoa)AMD Next-Gen EPYC (Post Genoa)AMD Next-Gen EPYC (Post Genoa)