3rd Gen AMD EPYC Processors with AMD 3D V‑Cache Technology Deliver Outstanding Leadership Performance in Technical Computing Workloads

— Newest addi­ti­on to 3rd Gen AMD EPYC fami­ly fea­tures 768MB of L3 cache, drop-in plat­form com­pa­ti­bi­li­ty, and modern secu­ri­ty features —

— The EPYC pro­ces­sor eco­sys­tem for tech­ni­cal com­pu­ting grows with solu­ti­ons from major OEMs, ODMs, SIs, ISVs and the cloud —

SANTA CLARA, Calif., March 21, 2022 (GLOBE NEWSWIRE) — AMD (NASDAQ: AMD) announ­ced the gene­ral avai­la­bi­li­ty of the world’s first data cen­ter CPU using 3D die stack­ing, the 3rd Gen AMD EPYC™ pro­ces­sors with AMD 3D V‑Cache™ tech­no­lo­gy, form­er­ly code­na­med “Milan‑X.” Built on the “Zen 3” core archi­tec­tu­re, the­se pro­ces­sors expand the 3rd Gen EPYC CPU fami­ly and can deli­ver up to 66 per­cent per­for­mance uplift across a varie­ty of tar­ge­ted tech­ni­cal com­pu­ting workloads ver­sus com­pa­ra­ble, non-sta­cked 3rd Gen AMD EPYC pro­ces­sors.1, 2

The­se new pro­ces­sors fea­ture the industry’s lar­gest L3 cache,3 deli­ve­ring the same socket, soft­ware com­pa­ti­bi­li­ty and modern secu­ri­ty fea­tures as 3rd Gen AMD EPYC CPUs while pro­vi­ding out­stan­ding per­for­mance for tech­ni­cal com­pu­ting workloads such as com­pu­ta­tio­nal flu­id dyna­mics (CFD), fini­te ele­ment ana­ly­sis (FEA), elec­tro­nic design auto­ma­ti­on (EDA) and struc­tu­ral ana­ly­sis. The­se workloads are cri­ti­cal design tools for com­pa­nies that must model the com­ple­xi­ties of the phy­si­cal world to crea­te simu­la­ti­ons that test and vali­da­te engi­nee­ring designs for some of the world’s most inno­va­te products.

Buil­ding upon our momen­tum in the data cen­ter as well as our histo­ry of indus­try-firsts, 3rd Gen AMD EPYC pro­ces­sors with AMD 3D V‑Cache tech­no­lo­gy show­ca­se our lea­der­ship design and pack­a­ging tech­no­lo­gy enab­ling us to offer the industry’s first workload-tail­o­red ser­ver pro­ces­sor with 3D die stack­ing tech­no­lo­gy,” said Dan McNa­ma­ra, seni­or vice pre­si­dent and gene­ral mana­ger, Ser­ver Busi­ness Unit, AMD. “Our latest pro­ces­sors with AMD 3D V‑Cache tech­no­lo­gy pro­vi­de breakth­rough per­for­mance for mis­si­on-cri­ti­cal tech­ni­cal com­pu­ting workloads lea­ding to bet­ter desi­gned pro­ducts and fas­ter time to market.”

Cus­to­mers’ increased adop­ti­on of data-rich appli­ca­ti­ons requi­res a new approach to data cen­ter infra­struc­tu­re. Micron and AMD share a visi­on of deli­ve­ring full capa­bi­li­ty of lea­ding DDR5 memo­ry to high-per­for­mance data cen­ter plat­forms,” said Raj Haz­ra, seni­or vice pre­si­dent and gene­ral mana­ger of the Com­pu­te and Net­wor­king Busi­ness Unit at Micron. “Our deep col­la­bo­ra­ti­on with AMD includes rea­dy­ing AMD plat­forms for Micron’s latest DDR5 solu­ti­ons as well as brin­ging 3rd Gen AMD EPYC pro­ces­sors with AMD 3D V‑Cache tech­no­lo­gy into our own data cen­ters, whe­re we are alre­a­dy see­ing up to a 40% per­for­mance impro­ve­ment over 3rd Gen AMD EPYC pro­ces­sors wit­hout AMD 3D V‑Cache on sel­ect EDA workloads.”

Lea­ding Pack­a­ging Innovations
Cache size increa­ses have been at the fore­front of per­for­mance impro­ve­ment, par­ti­cu­lar­ly for tech­ni­cal com­pu­ting workloads rely­ing hea­vi­ly on lar­ge data sets. The­se workloads bene­fit from increased cache size, howe­ver 2D chip designs have phy­si­cal limi­ta­ti­ons on the amount of cache that can effec­tively be built on the CPU. AMD 3D V‑Cache tech­no­lo­gy sol­ves the­se phy­si­cal chal­lenges by bon­ding the AMD “Zen 3” core to the cache modu­le, incre­asing the amount of L3 while mini­mi­zing laten­cy and incre­asing through­put. This tech­no­lo­gy repres­ents an inno­va­ti­ve step for­ward in CPU design and pack­a­ging and enables breakth­rough per­for­mance in tar­ge­ted tech­ni­cal com­pu­ting workloads.

Breakth­rough Performance
The world’s hig­hest per­for­mance ser­ver pro­ces­sors for tech­ni­cal com­pu­ting,4 the 3rd Gen AMD EPYC pro­ces­sors with AMD 3D V‑Cache tech­no­lo­gy deli­ver fas­ter time-to-results on tar­ge­ted workloads, such as:

  • EDA – The 16-core, AMD EPYC7373X CPU can deli­ver up to 66 per­cent fas­ter simu­la­ti­ons on Syn­op­sys VCS™, when com­pared to the EPYC 73F3 CPU.5
  • FEA – The 64-core, AMD EPYC 7773X pro­ces­sor can deli­ver, on avera­ge, 44 per­cent more per­for­mance on Alta­ir® Radioss® simu­la­ti­on appli­ca­ti­ons com­pared to the competition’s top of stack pro­ces­sor.6
  • CFD – The 32-core AMD EPYC 7573X pro­ces­sor can sol­ve an avera­ge of 88 per­cent more CFD pro­blems per day than a com­pa­ra­ble com­pe­ti­ti­ve 32-core count pro­ces­sor, while run­ning Ansys® CFX®.7

The­se per­for­mance capa­bi­li­ties ulti­m­ate­ly enable cus­to­mers to deploy fewer ser­vers and redu­ce power con­sump­ti­on in the data cen­ter, hel­ping to lower total cost of owner­ship (TCO), redu­ce car­bon foot­print and address their envi­ron­men­tal sus­taina­bi­li­ty goals. For ins­tance, in a typi­cal data cen­ter sce­na­rio run­ning 4600 jobs per day of the Ansys® CFX® test case cfx-50, using 2P 32-core AMD EPYC 7573X CPU based ser­vers can redu­ce the esti­ma­ted num­ber of ser­vers requi­red from 20 to 10 and lower power con­sump­ti­on by 49 per­cent, when com­pared to the competition’s latest 2P 32-core pro­ces­sor-based ser­ver. This ends up pro­vi­ding a pro­jec­ted 51 per­cent lower TCO over three-years.

In other words, choo­sing 3rd Gen AMD EPYC pro­ces­sors with AMD 3D V‑Cache tech­no­lo­gy in this deploy­ment would have the envi­ron­men­tal sus­taina­bi­li­ty bene­fit of more than 81 acres of US forest per year in car­bon seques­te­red equi­va­lents.8

3rd Gen AMD EPYC pro­ces­sor with AMD 3D V‑Cache Tech­no­lo­gy Pro­duct Chart

Cores Model # CCD TDP (W) cTDP ran­ge (W) Base Freq (GHz) Max Boost Freq (Up to GHz)* L3 Cache
(MB)
DDR
Channels
Pri­ce
(1KU)
64 7773X 8 280 225 – 280 2.20 3.50 768 8 $ 8,800
32 7573X 8 280 225 – 280 2.80 3.60 768 8 $ 5,590
24 7473X 8 240 225 – 280 2.80 3.70 768 8 $ 3,900
16 7373X 8 240 225 – 280 3.05 3.80 768 8 $ 4,185

*Max boost for AMD EPYC pro­ces­sors is the maxi­mum fre­quen­cy achie­va­ble by any sin­gle core on the pro­ces­sor under nor­mal ope­ra­ting con­di­ti­ons for ser­ver systems.

Indus­try-wide Eco­sys­tem Support
3rd Gen AMD EPYC pro­ces­sors with AMD 3D V‑Cache tech­no­lo­gy are available today from a wide array of OEM part­ners, inclu­ding, Atos, Cis­co, Dell Tech­no­lo­gies, Giga­byte, HPE, Leno­vo, QCT, and Supermicro.

3rd Gen AMD EPYC pro­ces­sors with AMD 3D V‑Cache tech­no­lo­gy are also broad­ly sup­port­ed by AMD soft­ware eco­sys­tem part­ners, inclu­ding, Alta­ir, Ansys, Cadence, Das­sault Sys­tè­mes, Sie­mens, and Synopsys.

Micro­soft Azu­re HBv3 vir­tu­al machi­nes (VMs) have now been ful­ly upgraded to 3rd Gen AMD EPYC with AMD 3D V‑Cache tech­no­lo­gy. Accor­ding to Micro­soft, HBv3 VMs are the fas­test adopted addi­ti­on to the Azu­re HPC plat­form ever and have seen per­for­mance gains of up to 80 per­cent in key HPC workloads from the addi­ti­on of AMD 3D V‑Cache com­pared to the pre­vious HBv3 series VMs.

Watch the video announce­ment here and visit the landing page for 3rd Gen AMD EPYC pro­ces­sors with AMD 3D V‑Cache tech­no­lo­gy to learn more and read about what AMD cus­to­mers have to say, here.

Sup­port­ing Resources

About AMD
For more than 50 years AMD has dri­ven inno­va­ti­on in high-per­for­mance com­pu­ting, gra­phics and visua­liza­ti­on tech­no­lo­gies. Bil­li­ons of peo­p­le, lea­ding For­tu­ne 500 busi­nesses and cut­ting-edge sci­en­ti­fic rese­arch insti­tu­ti­ons around the world rely on AMD tech­no­lo­gy dai­ly to impro­ve how they live, work and play. AMD employees are focu­sed on buil­ding lea­der­ship high-per­for­mance and adap­ti­ve pro­ducts that push the boun­da­ries of what is pos­si­ble. For more infor­ma­ti­on about how AMD is enab­ling today and inspi­ring tomor­row, visit the AMD (NASDAQ: AMDweb­siteblogLin­ke­dIn and Twit­ter pages.

AMD, the AMD Arrow logo, EPYC, AMD 3D V‑Cache, and com­bi­na­ti­ons the­reof are trade­marks of Advan­ced Micro Devices, Inc.

1 MLNX-021B: AMD inter­nal test­ing as of 02/14/2022 on 2x 64C EPYC 7773X com­pared to 2x 64C EPYC 7763 using cumu­la­ti­ve avera­ge of each of the fol­lo­wing benchmark’s maxi­mum test result score: ANSYS® Flu­ent® 2022.1 (max is flu­ent-pump2 82%), ANSYS® CFX® 2022.1 (max is cfx_10 61%), and Alta­ir® Radioss® 2021.2 (max is rad-neon 56%) plus 1x 16C EPYC 7373X com­pared to 1x 16C EPYC 75F3 on Syn­op­sys VCS 2020 (max is AMD gra­phics core 66%). Results may vary.
2 “Tech­ni­cal Com­pu­ting” or “Tech­ni­cal Com­pu­ting Workloads” as defi­ned by AMD can include: elec­tro­nic design auto­ma­ti­on, com­pu­ta­tio­nal flu­id dyna­mics, fini­te ele­ment ana­ly­sis, seis­mic tomo­gra­phy, wea­ther fore­cas­ting, quan­tum mecha­nics, cli­ma­te rese­arch, mole­cu­lar mode­ling, or simi­lar workloads. GD-204
3 EPYC-024A: 3rd Gen AMD EPYC™ CPUs with AMD 3D V‑Cache™ tech­no­lo­gy have 768MB total L3 cache com­pared to a maxi­mum L3 cache size of 60MB on only one 3rd Gen Intel Xeon pro­ces­sor (Pla­ti­num 8380) and com­pared to all other com­mer­cial CPUs in the mar­ket. Other L3 cache sizes:
Ampere Alt­ra Max 16MB SLC
SPARC64 XII 32MB
POWER10 120MB
4 MLNX-032: World’s hig­hest per­for­mance x86 ser­ver CPU for tech­ni­cal com­pu­ting com­pa­ri­son based on AMD inter­nal test­ing as of 2/14/2022 mea­su­ring the score, rating or jobs/day for each of esti­ma­ted SPECrate®2017_fp_base, Ansys Flu­ent, Alta­ir Radioss and Ansys LS-Dyna appli­ca­ti­on test case simu­la­ti­ons avera­ge spee­dup on 2P ser­vers run­ning 32-core EPYC 7573X to 2P ser­vers run­ning 32-core Intel Xeon Pla­ti­num 8362 for per-core per­for­mance lea­der­ship and on 2P ser­vers run­ning top-of-stack 64-core EPYC 7773X to 2P ser­vers run­ning top-of-stack 40-core Intel Xeon Pla­ti­num 8380 for den­si­ty per­for­mance lea­der­ship. See www.spec.org for more infor­ma­ti­on. Results may vary based on fac­tors inclu­ding sili­con ver­si­on, hard­ware and soft­ware con­fi­gu­ra­ti­on and dri­ver ver­si­ons. SPEC®, SPE­Cra­te® and SPEC CPU® are regis­tered trade­marks of the Stan­dard Per­for­mance Eva­lua­ti­on Corporation.
5 MLNX-001A: EDA RTL Simu­la­ti­on com­pa­ri­son based on AMD inter­nal test­ing com­ple­ted on 9/20/2021 mea­su­ring the avera­ge time to com­ple­te a test case simu­la­ti­on. com­pa­ring: 1x 16C EPYC7373X with AMD 3D V‑Cache Tech­no­lo­gy ver­sus 1x 16C AMD EPYC73F3 on the same AMD “Day­to­na” refe­rence plat­form. Results may vary based on fac­tors inclu­ding sili­con ver­si­on, hard­ware and soft­ware con­fi­gu­ra­ti­on and dri­ver versions.
6 MLNX-016: Alta­ir® Radioss® 2021.2 com­pa­ri­son based on AMD inter­nal test­ing as of 02/14/2022 mea­su­ring the time to run the drop­san­der, neon, and t10m test case simu­la­ti­ons. Con­fi­gu­ra­ti­ons: 2x 64C AMD EPYC 7773X with AMD 3D V‑Cache™ ver­sus 2x 40C Intel® Xeon® Pla­ti­num 8380. neon is the max result. Results may vary based on fac­tors inclu­ding sili­con ver­si­on, hard­ware and soft­ware con­fi­gu­ra­ti­on and dri­ver versions.
7 MLNX-010A: ANSYS® CFX® 2022.1 com­pa­ri­son based on AMD inter­nal test­ing as of 02/14/2022 mea­su­ring the avera­ge time to run the cfx_10, cfx_50, cfx_100, cfx_lmans, and cfx_pump test case simu­la­ti­ons. Con­fi­gu­ra­ti­ons: 2x 32C AMD EPYC7573X with AMD 3D V‑Cache tech­no­lo­gy™ ver­sus 2x 32C Intel Xeon Pla­ti­num 8362. Cfx_10 is the max result. Results may vary based on fac­tors inclu­ding sili­con ver­si­on, hard­ware and soft­ware con­fi­gu­ra­ti­on and dri­ver versions.
8 MLNXTCO-007: To run 4600 airfoil_50M bench­marks per day with Ansys® CFX® it takes an esti­ma­ted 10 2P AMD EPYC7573X powered ser­vers or 20 2P Intel® Pla­ti­num 8362 based ser­vers. The EPYC 7573X solu­ti­on has an esti­ma­ted 50% fewer ser­vers; 50% less RU space; 49% less power, with an esti­ma­ted 50% lower 3‑year TCO which includes both OS and appli­ca­ti­on soft­ware. The EPYC 7573X solu­ti­on saves an esti­ma­ted 203.19 Metric Tons of CO2 which is an esti­ma­ted equi­va­lent car­bon sequestra­ti­on of 81 acres of US forests annually.