439:, a new refrigerant-based cooling system, and a reduced number of accelerators compared to the corresponding rackmount DGX A100 of the same generation. The price for the DGX Station A100 320G is $ 149,000 and $ 99,000 for the 160G model, Nvidia also offers Station rental at ~$ 9000 USD per month through partners in the US (rentacomputer.com) and Europe (iRent IT Systems) to help reduce the costs of implementing these systems at a small scale.
99:
38:
242:
The DGX-1 was first available in only the Pascal based configuration, with the first generation SXM socket. The later revision of the DGX-1 offered support for first generation Volta cards via the SXM-2 socket. Nvidia offered upgrade kits that allowed users with a Pascal based DGX-1 to upgrade to a
373:
A higher performance variant of the DGX-2, the DGX-2H, was offered as well. The DGX-2H replaced the DGX-2's dual Intel Xeon
Platinum 8168's with upgraded dual Intel Xeon Platinum 8174's. This upgrade does not increase core count per system, as both CPUs are 24 cores, nor does it enable any new
369:
The DGX-2 differs from other DGX models in that it contains two separate GPU daughterboards, each with eight GPUs. These boards are connected by an NVSwitch system that allows for full bandwidth communication across all GPUs in the system, without additional latency between boards.
567:
Announced May 2023, the DGX GH200 connects 32 Nvidia Hopper
Superchips into a singular superchip, that consists totally of 256 H100 GPUs, 32 Grace Neoverse V2 72-core CPUs, 32 OSFT single-port ConnectX-7 VPI of with 400 Gb/s InfiniBand and 16 dual-port
329:
to better manage the heat of almost 1500W of total system components, this allows it to keep a noise range under 35 dB under load. This, among other features, made this system a compelling purchase for customers without the infrastructure to run
605:
Announced March 2024, GB200 NVL72 connects 36 Grace
Neoverse V2 72-core CPUs and 72 B100 GPUs in a rack-scale design. The GB200 NVL72 is a liquid-cooled, rack-scale solution that boasts a 72-GPU NVLink domain that acts as a single massive GPU
434:
solution that can be purchased, leased, or rented by smaller companies or individuals who want to utilize machine learning. It follows many of the design choices of the original DGX station, such as the tower orientation, single socket CPU
541:, double the bandwidth of the DGX A100. The DGX H100 uses new 'Cedar Fever' cards, each with four ConnectX-7 400 GB/s controllers, and two cards per system. This gives the DGX H100 3.2 Tb/s of fabric bandwidth across Infiniband.
640:
list for most powerful supercomputers at the time of its completion, and has continued to remain high in performance. This same integration is available to any customer with minimal effort on their behalf, and the new
149:
x16 slot, facilitating flexible integration within the system architecture. To manage the substantial thermal output, DGX units are equipped with heatsinks and fans designed to maintain optimal operating temperatures.
580:. Nvidia DGX GH200 is designed to handle terabyte-class models for massive recommender systems, generative AI, and graph analytics, offering 19.5 TB of shared memory with linear scalability for giant AI models.
350:-based V100 32 GB (second generation) cards in a single unit. It was announced on 27 March in 2018. The DGX-2 delivers 2 Petaflops with 512 GB of shared memory for tackling massive datasets and uses
671:
Switches, that allow Eos to deliver 18 EFLOPs of FP8 compute, and 9 EFLOPs of FP16 compute, making Eos the 5th fastest AI supercomputer in the world, according to TOP500 (November 2023 edition).
255:
E5-2698 V3, and one with a 20 core E5-2698 V4. Pricing for the variant equipped with an E5-2698 V4 is unavailable, the Pascal based DGX-1 with an E5-2698 V3 was priced at launch at $ 129,000
366:
cards and 30.72 TB of SSD storage, all enclosed within a massive 10U rackmount chassis and drawing up to 10 kW under maximum load. The initial price for the DGX-2 was $ 399,000.
3405:
127:
674:
As Nvidia does not produce any storage devices or systems, Nvidia SuperPods rely on partners to provide high performance storage. Current storage partners for Nvidia
Superpods are
502:-based H100 accelerators, for a total of 32 PFLOPs of FP8 AI compute and 640 GB of HBM3 Memory, an upgrade over the DGX A100s 640GB HBM2 memory. This upgrade also increases
197:. The DGX-1 was announced on the 6th of April in 2016. All models are based on a dual socket configuration of Intel Xeon E5 CPUs, and are equipped with the following features.
1998:
1974:
1781:
2169:
2144:
2403:
2305:
430:
As the successor to the original DGX Station, the DGX Station A100, aims to fill the same niche as the DGX station in being a quiet, efficient, turnkey
2842:
2821:
2677:
609:. Nvidia DGX GB200 offers 13.5 TB HBM3e of shared memory with linear scalability for giant AI models, less than its predecessor DGX GH200.
2107:
2213:
48:
2758:
2623:
224:
1760:
153:
This framework makes DGX units suitable for computational tasks associated with artificial intelligence and machine learning models.
67:
663:, designed, built, and operated by Nvidia, was constructed of 18 H100 based SuperPods, totaling 576 DGX H100 systems, 500 Quantum-2
85:
279:
computer that can function completely independently without typical datacenter infrastructure such as cooling, redundant power, or
3335:
2794:
549:
2744:
2198:
2082:
334:
DGX systems, which can be loud, output a lot of heat, and take up a large area. This was Nvidia's first venture into bringing
3664:
3265:
1895:
498:
Announced March 22, 2022 and planned for release in Q3 2022, The DGX H100 is the 4th generation of DGX servers, built with 8
621:
solution provided by Nvidia using DGX hardware. This system combines DGX compute nodes with fast storage and high bandwidth
588:
Announced May 2023, the DGX Helios supercomputer features 4 DGX GH200 systems. Each is interconnected with Nvidia
Quantum-2
3570:
3340:
964:
867:
2449:
2039:
1745:
538:
408:
1907:
578:
3752:
3305:
3280:
3270:
630:
141:
GPU modules, which are housed on an independent system board. These GPUs can be connected either via a version of the
1850:
649:
CPUs. This gives the complete SuperPod 20 TB of HBM3 memory, 70.4 TB/s of bisection bandwidth, and up to 1
63:
1858:
NVIDIA DGX-1 Delivers 75X Faster
Training...Note: Caffe benchmark with AlexNet, training 1.28M images with 90 epochs
3330:
3325:
3320:
3310:
3300:
3275:
1643:
1352:
1255:
1158:
1061:
683:
642:
503:
499:
447:
388:
335:
248:
176:
2253:
2227:
3345:
3315:
3295:
3290:
3285:
3218:
3130:
3000:
1546:
1449:
654:
576:
347:
291:
259:
180:
1836:
374:
functions of the system, but it does increase the base frequency of the CPUs from 2.7 GHz to 3.1 GHz.
3747:
3737:
510:
size to 8U to accommodate the 700W TDP of each H100 SXM card. The DGX H100 also has two 1.92 TB SSDs for
2703:
2651:
2597:
2377:
452:
thus giving a total of 160 GB or 320 GB resulting either in DGX Station A100 variants 160G or 320G.
3543:
3449:
3236:
1924:
411:. The DGX A100 is in a much smaller enclosure than its predecessor, the DGX-2, taking up only 6 Rack units.
3624:
3248:
3043:
626:
660:
592:
networking to supercharge data throughput for training large AI models. Helios includes 1,024 H100 GPUs.
3689:
2526:
Every NVIDIA DGX benchmarked & power efficiency & value compared, including the latest DGX H100.
573:
534:
355:
299:
387:
Announced and released on May 14, 2020. The DGX A100 was the 3rd generation of DGX server, including 8
2951:
2787:
553:
526:
2351:
1822:
3172:
3018:
2474:
679:
227:
using specific features for deep learning workloads. The initial Pascal based DGX-1 delivered 170
3428:
3393:
3023:
3013:
3008:
2989:
2984:
2979:
2974:
2969:
111:
3614:
3494:
3062:
3057:
3052:
3033:
3028:
2932:
2422:
2324:
515:
211:
59:
3674:
3669:
3509:
3194:
3160:
3141:
3113:
2964:
2959:
2942:
2937:
2927:
2922:
2900:
2895:
2279:
1805:
622:
569:
522:
511:
431:
2745:"NVIDIA Ampere Unleashed: NVIDIA Announces New GPU Architecture, A100 GPU, and Accelerator"
2730:"NVIDIA Hopper GPU Architecture and H100 Accelerator Announced: Working Smarter and Harder"
2199:"NVIDIA Ampere Unleashed: NVIDIA Announces New GPU Architecture, A100 GPU, and Accelerator"
3484:
3466:
3212:
2780:
2571:
2434:
2336:
2014:
2404:"NVIDIA Announces DGX H100 Systems – World's Most Advanced Enterprise AI Infrastructure"
2306:"NVIDIA Announces DGX H100 Systems – World's Most Advanced Enterprise AI Infrastructure"
657:
AI compute. These SuperPods can then be further joined to create larger supercomputers.
607:
450:-based A100 accelerators, configured with 40 GB (HBM) or 80 GB (HBM2e) memory,
3742:
3551:
3442:
3437:
3206:
1870:
276:
130:(GPGPU). These systems typically come in a rackmount format featuring high-performance
2546:
2131:
3731:
3694:
3528:
3489:
3454:
3200:
2910:
2499:
1755:
618:
326:
183:
123:
3699:
3619:
3609:
3523:
3413:
3224:
3107:
3101:
2914:
1750:
645:
based SuperPod can scale to 32 DGX H100 nodes, for a total of 256 H100 GPUs and 64
507:
442:
The DGX Station A100 comes with two different configurations of the built in A100.
331:
295:
280:
194:
142:
138:
2053:
1975:"Intel® Xeon® Processor E5-2698 v4 (50M Cache, 2.20 GHz) - Product Specifications"
262:
based DGX-1 is equipped with an E5-2698 V4 and was priced at launch at $ 149,000.
3499:
3379:
2890:
472:
436:
392:
115:
354:
for high-bandwidth internal communication. DGX-2 has a total of 512 GB of
3629:
3242:
3166:
2624:"Nvidia reveals H100 GPU for AI and teases 'world's fastest AI supercomputer'"
2524:
2108:"The NVIDIA DGX-2 is the world's first 2-petaflop single server supercomputer"
664:
589:
530:
466:
404:
363:
359:
338:
deskside, which has since remained a prominent marketing strategy for Nvidia.
202:
98:
2875:
2628:
2214:"Nvidia's first Ampere GPU is designed for data centers and AI, not your PC"
695:
1949:
346:
The successor of the Nvidia DGX-1 is the Nvidia DGX-2, which uses sixteen
2870:
2865:
675:
400:
236:
228:
2729:
2280:"Nvidia Refreshes Expensive, Powerful DGX Station 320G and DGX Superpod"
1761:
Page on high performance computing with 4x and 8x A100 per computer node
3679:
3562:
3518:
2816:
286:
The DGX station was first available with the following specifications.
272:
3388:
3230:
2803:
691:
668:
637:
625:
to provide a solution to high demand machine learning workloads. The
351:
191:
119:
17:
2228:"Boston Labs welcomes the DGX A100 to our remote testing portfolio!"
3684:
3659:
3533:
3461:
3177:
3152:
3146:
650:
559:
The DGX H100 was priced at £379,000 or ~$ 482,000 USD at release.
187:
97:
2083:"NVIDIA's DGX-2: Sixteen Tesla V100s, 30 TB of NVMe, only $ 400K"
3654:
3136:
2704:"Nvidia Eos AI supercomputer will need a monster storage system"
2652:"Nvidia Eos AI supercomputer will need a monster storage system"
545:
459:
419:
415:
396:
309:
252:
235:
processing, while the Volta-based upgrade increased this to 960
232:
223:
The product line is intended to bridge the gap between GPUs and
146:
3591:
3366:
3083:
2814:
2776:
2856:
2851:
687:
646:
422:
CPU. The initial price for the DGX A100 Server was $ 199,000.
172:
131:
31:
2352:"NVIDIA H100: Overview, Specs, & Release Date | SeiMaxim"
2598:"Nvidia Reveals Hopper H100 GPU With 80 Billion Transistors"
2378:"Nvidia Reveals Hopper H100 GPU With 80 Billion Transistors"
1823:"NVIDIA Unveils the DGX-1 HPC Server: 8 Teslas, 3U, Q2 2016"
137:
The core feature of a DGX system is its inclusion of 4 to 8
2678:"Nvidia announces Eos, "world's fastest AI supercomputer""
2772:
2450:"NVIDIA Cedar Fever 1.6Tbps Modules Used in the DGX H100"
636:
Selene, built from 280 DGX A100 nodes, ranked 5th on the
391:-based A100 accelerators. Also included is 15 TB of
521:
One more notable addition is the presence of two Nvidia
2759:"NVIDIA Tesla V100 tested: near unbelievable GPU power"
2040:"Nvidia launches the DGX-2 with two petaFLOPS of power"
418:
7742 CPU, the first DGX server to not be built with an
55:
128:
general-purpose computing on graphics processing units
506:
bandwidth to 3 TB/s. The DGX H100 increases the
1950:"CompecTA | NVIDIA DGX Station Deep Learning System"
66:, and by adding encyclopedic content written from a
3647:
3602:
3560:
3542:
3508:
3477:
3427:
3404:
3377:
3258:
3187:
3123:
3094:
3042:
2998:
2950:
2909:
2841:
2834:
1810:
Eight GPU hybrid cube mesh architecture with NVLink
2254:"Nvidia will let you rent its mini supercomputers"
633:, is one example of a DGX SuperPod based system.
275:deskside AI supercomputer, the DGX Station is a
251:base DGX-1 has two variants, one with a 16 core
27:Line of Nvidia produced servers and workstations
617:The DGX Superpod is a high performance turnkey
2788:
2192:
2190:
2009:
2007:
8:
1925:"NVIDIA Ships First Volta-based DGX Systems"
298:V100 accelerators, each with 16 GB of
102:A rack containing five DGX-1 supercomputers
3599:
3588:
3374:
3363:
3091:
3080:
2838:
2831:
2811:
2795:
2781:
2773:
2212:Tom Warren; James Vincent (14 May 2020).
2054:"NVIDIA DGX -2 for Complex AI Challenges"
362:. Also present are eight 100 Gb/sec
216:3200W of combined power supply capability
86:Learn how and when to remove this message
717:
706:Comparison of accelerators used in DGX:
548:Platinum 8480C Scalable CPUs (Codenamed
1773:
2430:
2420:
2332:
2322:
47:contains content that is written like
2076:
2074:
414:The DGX A100 also moved to a 64 core
399:storage, 1 TB of RAM, and eight
122:, primarily geared towards enhancing
7:
1918:
1916:
1763:, also showing switch topology dumps
529:, and the upgrade to 400 Gb/s
358:memory, a total of 1.5 TB of
25:
2622:Vincent, James (22 March 2022).
2376:Walton, Jarred (22 March 2022).
134:server CPUs on the motherboard.
126:applications through the use of
36:
2702:Mellor, Chris (31 March 2022).
2650:Mellor, Chris (31 March 2022).
2596:Jarred Walton (22 March 2022).
2278:Jarred Walton (12 April 2021).
2252:Mayank Sharma (13 April 2021).
478:1 x 7.68 TB U.2 NVMe Drive
3245:(framebuffer in system memory)
2448:servethehome (14 April 2022).
514:storage, and 30.72 TB of
484:Single port 1 Gb BMC port
1:
2728:Smith, Ryan (22 March 2022).
1896:Volta architecture whitepaper
1837:"deep learning supercomputer"
481:Dual port 10 Gb Ethernet
1851:"DGX-1 deep learning system"
1746:Deep Learning Super Sampling
2743:Smith, Ryan (14 May 2020).
2572:"NVIDIA SuperPOD Datasheet"
2475:"NVIDIA DGX H100 Datasheet"
631:Argonne National Laboratory
403:-powered 200 GB/s HDR
3769:
2197:Ryan Smith (14 May 2020).
572:VPI with 200 Gb/s of
336:high performance computing
207:Dual 10 Gb networking
186:with 128 GB of total
3598:
3587:
3450:RSX 'Reality Synthesizer'
3373:
3362:
3090:
3085:Software and technologies
3079:
2830:
2810:
2676:Comment, Sebastian Moss.
2408:NVIDIA Newsroom Newsroom
2350:Albert (24 March 2022).
2310:NVIDIA Newsroom Newsroom
2170:"Product Specifications"
2145:"Product Specifications"
321:Dual 10 Gb Ethernet
190:memory, connected by an
171:DGX-1 servers feature 8
3237:Scalable Link Interface
3203:(variable refresh rate)
3095:Multimedia acceleration
2112:www.hardwarezone.com.sg
1999:Supercomputer datasheet
1426:20736 KB (192 KB × 108)
1329:20736 KB (192 KB × 108)
1232:25344 KB (192 KB × 132)
1135:25344 KB (192 KB × 132)
110:represents a series of
3259:GPU microarchitectures
3251:(live video upscaling)
3249:Video Super Resolution
3169:(ray tracing platform)
2999:Unified shaders &
1620:10240 KB (128 KB × 80)
1523:10240 KB (128 KB × 80)
518:for application data.
103:
3690:Mellanox Technologies
2015:"NVIDIA DGX Platform"
694:, Pavilion Data, and
552:) and 2 Terabytes of
544:The DGX H100 has two
101:
68:neutral point of view
2835:Fixed pixel pipeline
2765:. 17 September 2017.
2682:Data Center Dynamics
1717:1344 KB (24 KB × 56)
627:Selene Supercomputer
318:4x 1.92 TB SSDs
219:3U Rackmount Chassis
3173:Nvidia System Tools
516:Solid state storage
325:The DGX station is
243:Volta based DGX-1.
60:promotional content
3753:Parallel computing
3180:(video decode API)
3133:(shading language)
3046:& tensor cores
2547:"NVIDIA DGX GH200"
2433:has generic name (
2402:Newsroom, NVIDIA.
2335:has generic name (
2304:Newsroom, NVIDIA.
667:switches, and 360
104:
62:and inappropriate
3725:
3724:
3721:
3720:
3717:
3716:
3615:Chris Malachowsky
3583:
3582:
3579:
3578:
3495:Shield Android TV
3358:
3357:
3354:
3353:
3149:(ray tracing API)
3075:
3074:
3071:
3070:
2885:
2884:
2825:
2500:"NVIDIA DGX H100"
2087:www.anandtech.com
1929:www.anandtech.com
1736:
1735:
715:
714:
661:Eos supercomputer
471:1 x 1.92 TB
210:4 x 1.92 TB
96:
95:
88:
16:(Redirected from
3760:
3675:Cumulus Networks
3670:Bright Computing
3655:3dfx Interactive
3600:
3589:
3470:
3458:
3446:
3375:
3364:
3195:Nvidia 3D Vision
3161:Nvidia Omniverse
3142:Nvidia GameWorks
3116:(video decoding)
3110:(video decoding)
3104:(video encoding)
3092:
3081:
2839:
2832:
2819:
2812:
2797:
2790:
2783:
2774:
2767:
2766:
2755:
2749:
2748:
2740:
2734:
2733:
2725:
2719:
2718:
2716:
2714:
2708:Blocks and Files
2699:
2693:
2692:
2690:
2688:
2673:
2667:
2666:
2664:
2662:
2656:Blocks and Files
2647:
2641:
2640:
2638:
2636:
2619:
2613:
2612:
2610:
2608:
2593:
2587:
2586:
2584:
2582:
2568:
2562:
2561:
2559:
2557:
2543:
2537:
2536:
2535:
2533:
2521:
2515:
2514:
2512:
2510:
2496:
2490:
2489:
2487:
2485:
2471:
2465:
2464:
2462:
2460:
2445:
2439:
2438:
2432:
2428:
2426:
2418:
2416:
2414:
2399:
2393:
2392:
2390:
2388:
2373:
2367:
2366:
2364:
2362:
2356:www.seimaxim.com
2347:
2341:
2340:
2334:
2330:
2328:
2320:
2318:
2316:
2301:
2295:
2294:
2292:
2290:
2275:
2269:
2268:
2266:
2264:
2249:
2243:
2242:
2240:
2238:
2232:www.boston.co.uk
2224:
2218:
2217:
2209:
2203:
2202:
2194:
2185:
2184:
2182:
2180:
2166:
2160:
2159:
2157:
2155:
2141:
2135:
2129:
2123:
2122:
2120:
2118:
2104:
2098:
2097:
2095:
2093:
2078:
2069:
2068:
2066:
2064:
2050:
2044:
2043:
2042:. 28 March 2018.
2036:
2030:
2029:
2027:
2025:
2011:
2002:
1996:
1990:
1989:
1987:
1985:
1971:
1965:
1964:
1962:
1960:
1954:www.compecta.com
1946:
1940:
1939:
1937:
1935:
1920:
1911:
1905:
1899:
1893:
1887:
1886:
1884:
1882:
1867:
1861:
1860:
1855:
1847:
1841:
1840:
1833:
1827:
1826:
1819:
1813:
1812:
1808:. 5 April 2016.
1802:
1796:
1795:
1793:
1791:
1786:
1778:
1569:1.75 Gbit/s HBM2
1472:1.75 Gbit/s HBM2
1278:3.2 Gbit/s HBM2e
1084:6.3 Gbit/s HBM3e
718:
709:
708:
512:Operating System
432:cluster-in-a-box
426:DGX Station A100
315:256 GB DDR4
91:
84:
80:
77:
71:
49:an advertisement
40:
39:
32:
21:
3768:
3767:
3763:
3762:
3761:
3759:
3758:
3757:
3748:Nvidia products
3738:AI accelerators
3728:
3727:
3726:
3713:
3643:
3637:Ranga Jayaraman
3634:Debora Shoquist
3594:
3575:
3556:
3538:
3504:
3485:Shield Portable
3473:
3467:Nintendo Switch
3464:
3452:
3440:
3423:
3400:
3369:
3350:
3254:
3227:(module/socket)
3221:(module/socket)
3215:(multi-monitor)
3213:Nvidia Surround
3209:(GPU switching)
3183:
3119:
3086:
3067:
3038:
2994:
2952:Unified shaders
2946:
2905:
2881:
2880:
2861:
2826:
2806:
2801:
2771:
2770:
2757:
2756:
2752:
2742:
2741:
2737:
2727:
2726:
2722:
2712:
2710:
2701:
2700:
2696:
2686:
2684:
2675:
2674:
2670:
2660:
2658:
2649:
2648:
2644:
2634:
2632:
2621:
2620:
2616:
2606:
2604:
2595:
2594:
2590:
2580:
2578:
2570:
2569:
2565:
2555:
2553:
2545:
2544:
2540:
2531:
2529:
2523:
2522:
2518:
2508:
2506:
2498:
2497:
2493:
2483:
2481:
2473:
2472:
2468:
2458:
2456:
2447:
2446:
2442:
2429:
2419:
2412:
2410:
2401:
2400:
2396:
2386:
2384:
2375:
2374:
2370:
2360:
2358:
2349:
2348:
2344:
2331:
2321:
2314:
2312:
2303:
2302:
2298:
2288:
2286:
2277:
2276:
2272:
2262:
2260:
2251:
2250:
2246:
2236:
2234:
2226:
2225:
2221:
2211:
2210:
2206:
2196:
2195:
2188:
2178:
2176:
2168:
2167:
2163:
2153:
2151:
2143:
2142:
2138:
2132:DGX2 User Guide
2130:
2126:
2116:
2114:
2106:
2105:
2101:
2091:
2089:
2080:
2079:
2072:
2062:
2060:
2052:
2051:
2047:
2038:
2037:
2033:
2023:
2021:
2013:
2012:
2005:
1997:
1993:
1983:
1981:
1973:
1972:
1968:
1958:
1956:
1948:
1947:
1943:
1933:
1931:
1922:
1921:
1914:
1906:
1902:
1894:
1890:
1880:
1878:
1869:
1868:
1864:
1853:
1849:
1848:
1844:
1839:. 5 April 2016.
1835:
1834:
1830:
1821:
1820:
1816:
1806:"inside pascal"
1804:
1803:
1799:
1789:
1787:
1784:
1780:
1779:
1775:
1770:
1742:
1737:
1666:1.4 Gbit/s HBM2
1375:2.4 Gbit/s HBM2
1181:5.2 Gbit/s HBM3
855:
835:
830:
825:
823:
818:
813:
805:
797:
792:
787:
785:
780:
778:
770:
765:
760:
755:
750:
745:
743:
738:
733:
731:
704:
615:
603:
598:
586:
565:
550:Sapphire Rapids
496:
494:DGX H100 Server
491:
458:Single 64 Core
455:2.5 PFLOPS FP16
451:
428:
385:
383:DGX A100 Server
380:
344:
305:480 TFLOPS FP16
269:
225:AI accelerators
201:512 GB of
169:
164:
159:
92:
81:
75:
72:
53:
41:
37:
28:
23:
22:
15:
12:
11:
5:
3766:
3764:
3756:
3755:
3750:
3745:
3740:
3730:
3729:
3723:
3722:
3719:
3718:
3715:
3714:
3712:
3711:
3708:
3705:
3702:
3697:
3692:
3687:
3682:
3677:
3672:
3667:
3662:
3657:
3651:
3649:
3645:
3644:
3642:
3641:
3640:Jonah M. Alben
3638:
3635:
3632:
3627:
3622:
3617:
3612:
3610:Jen-Hsun Huang
3606:
3604:
3596:
3595:
3592:
3585:
3584:
3581:
3580:
3577:
3576:
3574:
3573:
3567:
3565:
3558:
3557:
3555:
3554:
3552:Project Denver
3548:
3546:
3540:
3539:
3537:
3536:
3531:
3526:
3521:
3515:
3513:
3506:
3505:
3503:
3502:
3497:
3492:
3487:
3481:
3479:
3475:
3474:
3472:
3471:
3459:
3447:
3434:
3432:
3425:
3424:
3422:
3421:
3416:
3410:
3408:
3402:
3401:
3399:
3398:
3397:
3396:
3385:
3383:
3371:
3370:
3368:Other products
3367:
3360:
3359:
3356:
3355:
3352:
3351:
3349:
3348:
3343:
3338:
3333:
3328:
3323:
3318:
3313:
3308:
3303:
3298:
3293:
3288:
3283:
3278:
3273:
3268:
3262:
3260:
3256:
3255:
3253:
3252:
3246:
3240:
3234:
3228:
3222:
3216:
3210:
3207:Nvidia Optimus
3204:
3198:
3191:
3189:
3185:
3184:
3182:
3181:
3175:
3170:
3164:
3158:
3157:
3156:
3150:
3139:
3134:
3127:
3125:
3121:
3120:
3118:
3117:
3111:
3105:
3098:
3096:
3088:
3087:
3084:
3077:
3076:
3073:
3072:
3069:
3068:
3066:
3065:
3060:
3055:
3049:
3047:
3040:
3039:
3037:
3036:
3031:
3026:
3021:
3016:
3011:
3005:
3003:
2996:
2995:
2993:
2992:
2987:
2982:
2977:
2972:
2967:
2962:
2956:
2954:
2948:
2947:
2945:
2940:
2935:
2930:
2925:
2920:
2918:
2907:
2906:
2904:
2903:
2898:
2893:
2886:
2883:
2882:
2879:
2878:
2873:
2868:
2862:
2860:
2859:
2854:
2848:
2847:
2845:
2836:
2828:
2827:
2815:
2808:
2807:
2802:
2800:
2799:
2792:
2785:
2777:
2769:
2768:
2750:
2735:
2720:
2694:
2668:
2642:
2614:
2602:Tom's Hardware
2588:
2563:
2538:
2516:
2491:
2479:www.nvidia.com
2466:
2440:
2394:
2382:Tom's Hardware
2368:
2342:
2296:
2284:Tom's Hardware
2270:
2244:
2219:
2204:
2186:
2161:
2136:
2124:
2099:
2081:Cutress, Ian.
2070:
2045:
2031:
2003:
1991:
1966:
1941:
1912:
1900:
1888:
1862:
1842:
1828:
1814:
1797:
1782:"nvidia dgx-1"
1772:
1771:
1769:
1766:
1765:
1764:
1758:
1753:
1748:
1741:
1738:
1734:
1733:
1730:
1727:
1724:
1721:
1718:
1715:
1712:
1709:
1706:
1703:
1700:
1697:
1694:
1691:
1688:
1685:
1682:
1679:
1676:
1673:
1670:
1667:
1664:
1661:
1658:
1655:
1652:
1649:
1646:
1641:
1637:
1636:
1633:
1630:
1627:
1624:
1621:
1618:
1615:
1612:
1609:
1606:
1603:
1600:
1597:
1594:
1591:
1588:
1585:
1582:
1579:
1576:
1573:
1570:
1567:
1564:
1561:
1558:
1555:
1552:
1549:
1544:
1540:
1539:
1536:
1533:
1530:
1527:
1524:
1521:
1518:
1515:
1512:
1509:
1506:
1503:
1500:
1497:
1494:
1491:
1488:
1485:
1482:
1479:
1476:
1473:
1470:
1467:
1464:
1461:
1458:
1455:
1452:
1447:
1443:
1442:
1439:
1436:
1433:
1430:
1427:
1424:
1421:
1418:
1415:
1412:
1409:
1406:
1403:
1400:
1397:
1394:
1391:
1388:
1385:
1382:
1379:
1376:
1373:
1370:
1367:
1364:
1361:
1358:
1355:
1350:
1346:
1345:
1342:
1339:
1336:
1333:
1330:
1327:
1324:
1321:
1318:
1315:
1312:
1309:
1306:
1303:
1300:
1297:
1294:
1291:
1288:
1285:
1282:
1279:
1276:
1273:
1270:
1267:
1264:
1261:
1258:
1253:
1249:
1248:
1245:
1242:
1239:
1236:
1233:
1230:
1227:
1224:
1221:
1218:
1215:
1212:
1209:
1206:
1203:
1200:
1197:
1194:
1191:
1188:
1185:
1182:
1179:
1176:
1173:
1170:
1167:
1164:
1161:
1156:
1152:
1151:
1148:
1145:
1142:
1139:
1136:
1133:
1130:
1127:
1124:
1121:
1118:
1115:
1112:
1109:
1106:
1103:
1100:
1097:
1094:
1091:
1088:
1085:
1082:
1079:
1076:
1073:
1070:
1067:
1064:
1059:
1055:
1054:
1051:
1048:
1045:
1042:
1039:
1036:
1033:
1030:
1027:
1024:
1021:
1018:
1015:
1012:
1009:
1006:
1003:
1000:
997:
994:
991:
988:
987:8 Gbit/s HBM3e
985:
982:
979:
976:
973:
970:
967:
962:
958:
957:
954:
951:
948:
945:
942:
939:
936:
933:
930:
927:
924:
921:
918:
915:
912:
909:
906:
903:
900:
897:
894:
891:
890:8 Gbit/s HBM3e
888:
885:
882:
879:
876:
873:
870:
865:
861:
860:
857:
852:
849:
846:
843:
840:
837:
832:
827:
822:TensorFloat-32
820:
815:
810:
807:
802:
799:
794:
789:
782:
775:
772:
767:
762:
757:
752:
747:
740:
739:(excl. tensor)
735:
728:
725:
722:
716:
713:
712:
703:
700:
614:
611:
602:
599:
597:
594:
585:
582:
564:
561:
495:
492:
490:
487:
486:
485:
482:
479:
476:
469:
463:
456:
453:
427:
424:
384:
381:
379:
376:
343:
340:
323:
322:
319:
316:
313:
306:
303:
271:Designed as a
268:
265:
264:
263:
256:
233:half precision
221:
220:
217:
214:
208:
205:
184:daughter cards
168:
165:
163:
162:Pascal - Volta
160:
158:
155:
94:
93:
64:external links
44:
42:
35:
26:
24:
14:
13:
10:
9:
6:
4:
3:
2:
3765:
3754:
3751:
3749:
3746:
3744:
3741:
3739:
3736:
3735:
3733:
3709:
3706:
3703:
3701:
3698:
3696:
3695:Mental Images
3693:
3691:
3688:
3686:
3683:
3681:
3678:
3676:
3673:
3671:
3668:
3666:
3663:
3661:
3658:
3656:
3653:
3652:
3650:
3646:
3639:
3636:
3633:
3631:
3628:
3626:
3623:
3621:
3618:
3616:
3613:
3611:
3608:
3607:
3605:
3601:
3597:
3590:
3586:
3572:
3569:
3568:
3566:
3564:
3559:
3553:
3550:
3549:
3547:
3545:
3541:
3535:
3532:
3530:
3527:
3525:
3522:
3520:
3517:
3516:
3514:
3511:
3507:
3501:
3498:
3496:
3493:
3491:
3490:Shield Tablet
3488:
3486:
3483:
3482:
3480:
3478:Nvidia Shield
3476:
3468:
3463:
3460:
3456:
3455:PlayStation 3
3451:
3448:
3444:
3439:
3436:
3435:
3433:
3430:
3426:
3420:
3417:
3415:
3412:
3411:
3409:
3407:
3403:
3395:
3392:
3391:
3390:
3389:Nvidia Quadro
3387:
3386:
3384:
3381:
3376:
3372:
3365:
3361:
3347:
3344:
3342:
3339:
3337:
3334:
3332:
3329:
3327:
3324:
3322:
3319:
3317:
3314:
3312:
3309:
3307:
3304:
3302:
3299:
3297:
3294:
3292:
3289:
3287:
3284:
3282:
3279:
3277:
3274:
3272:
3269:
3267:
3264:
3263:
3261:
3257:
3250:
3247:
3244:
3241:
3238:
3235:
3232:
3229:
3226:
3223:
3220:
3217:
3214:
3211:
3208:
3205:
3202:
3201:Nvidia G-Sync
3199:
3196:
3193:
3192:
3190:
3186:
3179:
3176:
3174:
3171:
3168:
3165:
3163:(3D graphics)
3162:
3159:
3155:(physics SDK)
3154:
3151:
3148:
3145:
3144:
3143:
3140:
3138:
3135:
3132:
3129:
3128:
3126:
3122:
3115:
3112:
3109:
3106:
3103:
3100:
3099:
3097:
3093:
3089:
3082:
3078:
3064:
3061:
3059:
3056:
3054:
3051:
3050:
3048:
3045:
3041:
3035:
3032:
3030:
3027:
3025:
3022:
3020:
3017:
3015:
3012:
3010:
3007:
3006:
3004:
3002:
2997:
2991:
2988:
2986:
2983:
2981:
2978:
2976:
2973:
2971:
2968:
2966:
2963:
2961:
2958:
2957:
2955:
2953:
2949:
2944:
2941:
2939:
2936:
2934:
2931:
2929:
2926:
2924:
2921:
2919:
2916:
2912:
2908:
2902:
2899:
2897:
2894:
2892:
2888:
2887:
2877:
2874:
2872:
2869:
2867:
2864:
2863:
2858:
2855:
2853:
2850:
2849:
2846:
2844:
2840:
2837:
2833:
2829:
2823:
2818:
2813:
2809:
2805:
2798:
2793:
2791:
2786:
2784:
2779:
2778:
2775:
2764:
2760:
2754:
2751:
2746:
2739:
2736:
2731:
2724:
2721:
2709:
2705:
2698:
2695:
2683:
2679:
2672:
2669:
2657:
2653:
2646:
2643:
2631:
2630:
2625:
2618:
2615:
2603:
2599:
2592:
2589:
2577:
2573:
2567:
2564:
2552:
2548:
2542:
2539:
2528:
2527:
2520:
2517:
2505:
2501:
2495:
2492:
2480:
2476:
2470:
2467:
2455:
2451:
2444:
2441:
2436:
2424:
2409:
2405:
2398:
2395:
2383:
2379:
2372:
2369:
2357:
2353:
2346:
2343:
2338:
2326:
2311:
2307:
2300:
2297:
2285:
2281:
2274:
2271:
2259:
2255:
2248:
2245:
2233:
2229:
2223:
2220:
2215:
2208:
2205:
2200:
2193:
2191:
2187:
2175:
2174:www.intel.com
2171:
2165:
2162:
2150:
2149:www.intel.com
2146:
2140:
2137:
2133:
2128:
2125:
2113:
2109:
2103:
2100:
2088:
2084:
2077:
2075:
2071:
2059:
2055:
2049:
2046:
2041:
2035:
2032:
2020:
2016:
2010:
2008:
2004:
2000:
1995:
1992:
1980:
1976:
1970:
1967:
1955:
1951:
1945:
1942:
1930:
1926:
1919:
1917:
1913:
1909:
1904:
1901:
1897:
1892:
1889:
1876:
1872:
1866:
1863:
1859:
1852:
1846:
1843:
1838:
1832:
1829:
1824:
1818:
1815:
1811:
1807:
1801:
1798:
1783:
1777:
1774:
1767:
1762:
1759:
1757:
1756:Supercomputer
1754:
1752:
1749:
1747:
1744:
1743:
1739:
1731:
1728:
1725:
1722:
1719:
1716:
1713:
1710:
1707:
1704:
1701:
1698:
1695:
1692:
1689:
1686:
1683:
1680:
1677:
1674:
1671:
1668:
1665:
1663:1480 MHz
1662:
1659:
1656:
1653:
1650:
1647:
1645:
1642:
1639:
1638:
1634:
1631:
1628:
1625:
1622:
1619:
1616:
1613:
1610:
1607:
1604:
1601:
1598:
1595:
1592:
1589:
1586:
1583:
1580:
1577:
1574:
1571:
1568:
1566:1530 MHz
1565:
1562:
1559:
1556:
1553:
1550:
1548:
1545:
1542:
1541:
1537:
1534:
1531:
1528:
1525:
1522:
1519:
1516:
1513:
1510:
1507:
1504:
1501:
1498:
1495:
1492:
1489:
1486:
1483:
1480:
1477:
1474:
1471:
1469:1530 MHz
1468:
1465:
1462:
1459:
1456:
1453:
1451:
1448:
1445:
1444:
1440:
1437:
1434:
1431:
1428:
1425:
1422:
1419:
1416:
1413:
1410:
1407:
1404:
1401:
1398:
1395:
1392:
1389:
1386:
1383:
1380:
1377:
1374:
1372:1410 MHz
1371:
1368:
1365:
1362:
1359:
1356:
1354:
1351:
1348:
1347:
1343:
1340:
1337:
1334:
1331:
1328:
1325:
1322:
1319:
1316:
1313:
1310:
1307:
1304:
1301:
1298:
1295:
1292:
1289:
1286:
1283:
1280:
1277:
1275:1410 MHz
1274:
1271:
1268:
1265:
1262:
1259:
1257:
1254:
1251:
1250:
1246:
1243:
1240:
1237:
1234:
1231:
1228:
1225:
1222:
1219:
1216:
1213:
1210:
1207:
1204:
1201:
1198:
1195:
1192:
1189:
1186:
1183:
1180:
1178:1980 MHz
1177:
1174:
1171:
1168:
1165:
1162:
1160:
1157:
1154:
1153:
1149:
1146:
1143:
1140:
1137:
1134:
1131:
1128:
1125:
1122:
1119:
1116:
1113:
1110:
1107:
1104:
1101:
1098:
1095:
1092:
1089:
1086:
1083:
1080:
1077:
1074:
1071:
1068:
1065:
1063:
1060:
1057:
1056:
1052:
1049:
1046:
1043:
1040:
1037:
1034:
1031:
1028:
1025:
1022:
1019:
1016:
1013:
1010:
1007:
1004:
1001:
998:
995:
992:
989:
986:
983:
980:
977:
974:
971:
968:
966:
963:
960:
959:
955:
952:
949:
946:
943:
940:
937:
934:
931:
928:
925:
922:
919:
916:
913:
910:
907:
904:
901:
898:
895:
892:
889:
886:
883:
880:
877:
874:
871:
869:
866:
863:
862:
858:
853:
850:
847:
844:
841:
838:
833:
828:
821:
816:
811:
808:
803:
800:
795:
790:
783:
776:
773:
768:
763:
758:
753:
748:
741:
736:
729:
726:
723:
720:
719:
711:
710:
707:
701:
699:
697:
693:
689:
685:
681:
677:
672:
670:
666:
662:
658:
656:
652:
648:
644:
639:
634:
632:
628:
624:
620:
619:supercomputer
612:
610:
608:
600:
595:
593:
591:
583:
581:
579:
577:
575:
571:
562:
560:
557:
555:
554:System Memory
551:
547:
542:
540:
536:
532:
528:
524:
519:
517:
513:
509:
505:
501:
493:
488:
483:
480:
477:
474:
470:
468:
464:
461:
457:
454:
449:
445:
444:
443:
440:
438:
433:
425:
423:
421:
417:
412:
410:
406:
402:
398:
394:
390:
382:
377:
375:
371:
367:
365:
361:
357:
353:
349:
341:
339:
337:
333:
328:
320:
317:
314:
311:
307:
304:
301:
297:
293:
289:
288:
287:
284:
282:
281:19 inch racks
278:
274:
266:
261:
257:
254:
250:
246:
245:
244:
240:
238:
234:
230:
226:
218:
215:
213:
209:
206:
204:
200:
199:
198:
196:
193:
189:
185:
182:
178:
175:based on the
174:
166:
161:
156:
154:
151:
148:
144:
140:
135:
133:
129:
125:
124:deep learning
121:
117:
113:
109:
100:
90:
87:
79:
69:
65:
61:
57:
51:
50:
45:This article
43:
34:
33:
30:
19:
3700:PortalPlayer
3648:Acquisitions
3620:Curtis Priem
3512:and embedded
3462:Tegra NX-SoC
3418:
3414:Nvidia Tesla
3336:Ada Lovelace
3188:Technologies
2822:List of GPUs
2762:
2753:
2747:. AnandTech.
2738:
2732:. AnandTech.
2723:
2711:. Retrieved
2707:
2697:
2685:. Retrieved
2681:
2671:
2659:. Retrieved
2655:
2645:
2633:. Retrieved
2627:
2617:
2605:. Retrieved
2601:
2591:
2579:. Retrieved
2575:
2566:
2554:. Retrieved
2550:
2541:
2530:, retrieved
2525:
2519:
2507:. Retrieved
2503:
2494:
2482:. Retrieved
2478:
2469:
2457:. Retrieved
2454:ServeTheHome
2453:
2443:
2411:. Retrieved
2407:
2397:
2385:. Retrieved
2381:
2371:
2359:. Retrieved
2355:
2345:
2313:. Retrieved
2309:
2299:
2287:. Retrieved
2283:
2273:
2261:. Retrieved
2257:
2247:
2235:. Retrieved
2231:
2222:
2216:. The Verge.
2207:
2201:. AnandTech.
2177:. Retrieved
2173:
2164:
2152:. Retrieved
2148:
2139:
2127:
2115:. Retrieved
2111:
2102:
2090:. Retrieved
2086:
2061:. Retrieved
2057:
2048:
2034:
2022:. Retrieved
2018:
1994:
1982:. Retrieved
1978:
1969:
1957:. Retrieved
1953:
1944:
1932:. Retrieved
1928:
1903:
1891:
1879:. Retrieved
1874:
1871:"DGX Server"
1865:
1857:
1845:
1831:
1817:
1809:
1800:
1788:. Retrieved
1776:
1751:Nvidia Tesla
1093:141 GB HBM3e
996:192 GB HBM3e
899:192 GB HBM3e
834:Interconnect
831:dense tensor
826:dense tensor
819:dense tensor
814:dense tensor
806:dense tensor
798:dense tensor
793:(non-tensor)
724:Architecture
705:
702:Accelerators
673:
659:
635:
616:
613:DGX SuperPod
604:
587:
566:
558:
543:
520:
497:
465:512 GB
441:
429:
413:
386:
372:
368:
345:
327:water-cooled
324:
285:
270:
241:
222:
195:mesh network
170:
152:
145:socket or a
139:Nvidia Tesla
136:
118:designed by
116:workstations
107:
105:
82:
76:January 2024
73:
58:by removing
54:Please help
46:
29:
3500:GeForce Now
3394:Quadro Plex
3380:Workstation
3239:(multi-GPU)
3197:(stereo 3D)
3044:Ray tracing
3009:GeForce 600
2891:GeForce 256
2843:Pre-GeForce
2581:15 November
2431:|last=
2333:|last=
2024:15 November
1881:7 September
1790:15 November
1732:TSMC 16FF+
1726:610 mm
1696:21.2 TFLOPS
1678:10.6 TFLOPS
1635:TSMC 12FFN
1629:815 mm
1599:31.4 TFLOPS
1581:15.7 TFLOPS
1538:TSMC 12FFN
1532:815 mm
1502:31.4 TFLOPS
1484:15.7 TFLOPS
1435:826 mm
1417:19.5 TFLOPS
1387:19.5 TFLOPS
1381:1.52 TB/sec
1338:826 mm
1320:19.5 TFLOPS
1290:19.5 TFLOPS
1287:80 GB HBM2e
1284:1.52 TB/sec
1241:814 mm
1187:3.35 TB/sec
1144:814 mm
1023:1.98 PFLOPS
1020:1.98 PFLOPS
926:2.25 PFLOPS
923:2.25 PFLOPS
570:BlueField-3
537:ConnectX-7
407:ConnectX-6
267:DGX Station
3732:Categories
3630:Bill Dally
3625:David Kirk
3603:Key people
3431:components
3266:Fahrenheit
3243:TurboCache
3233:(protocol)
3167:Nvidia RTX
3053:GeForce 20
2134:nvidia.com
2001:nvidia.com
1923:Oh, Nate.
1910:nvidia.com
1898:nvidia.com
1875:DGX Server
1768:References
1711:160 GB/sec
1681:5.3 TFLOPS
1675:16 GB HBM2
1672:720 GB/sec
1614:300 GB/sec
1602:125 TFLOPS
1584:7.8 TFLOPS
1578:16 GB HBM2
1575:900 GB/sec
1543:V100 16GB
1517:300 GB/sec
1505:125 TFLOPS
1487:7.8 TFLOPS
1481:32 GB HBM2
1478:900 GB/sec
1446:V100 32GB
1420:600 GB/sec
1414:156 TFLOPS
1411:312 TFLOPS
1408:312 TFLOPS
1390:9.7 TFLOPS
1384:40 GB HBM2
1349:A100 40GB
1323:600 GB/sec
1317:156 TFLOPS
1314:312 TFLOPS
1311:312 TFLOPS
1293:9.7 TFLOPS
1252:A100 80GB
1226:900 GB/sec
1220:495 TFLOPS
1217:990 TFLOPS
1214:990 TFLOPS
1190:80 GB HBM3
1129:900 GB/sec
1123:495 TFLOPS
1120:990 TFLOPS
1117:990 TFLOPS
1090:4.8 TB/sec
1032:1.8 TB/sec
1026:989 TFLOPS
935:1.8 TB/sec
929:1.2 PFLOPS
854:Transistor
744:INT32/FP32
737:FP64 cores
665:InfiniBand
623:networking
590:InfiniBand
584:DGX Helios
531:InfiniBand
420:Intel Xeon
405:InfiniBand
364:InfiniBand
312:E5-2698 v4
310:Intel Xeon
253:Intel Xeon
108:Nvidia DGX
56:improve it
3561:Computer
3378:Graphics
3341:Blackwell
3114:PureVideo
2960:GeForce 8
2923:GeForce 3
2763:TweakTown
2629:The Verge
2361:22 August
2258:TechRadar
1984:19 August
1908:Use Guide
1593:15.7 TOPS
1496:15.7 TOPS
1405:78 TFLOPS
1399:19.5 TOPS
1308:78 TFLOPS
1302:19.5 TOPS
1223:67 TFLOPS
1202:1.98 POPS
1196:34 TFLOPS
1193:67 TFLOPS
1126:67 TFLOPS
1105:1.98 POPS
1099:34 TFLOPS
1096:67 TFLOPS
1053:TSMC 4NP
1029:30 TFLOPS
965:Blackwell
956:TSMC 4NP
932:40 TFLOPS
868:Blackwell
786:precision
779:precision
771:bandwidth
766:bus width
696:VAST Data
629:, at the
601:DGX GB200
596:Blackwell
563:DGX GH200
523:Bluefield
508:rackmount
437:mainboard
332:rackmount
237:teraflops
229:teraflops
203:DDR4-2133
3563:chipsets
3124:Software
2871:RIVA TNT
2866:RIVA 128
2713:29 April
2607:24 March
2556:24 March
2509:24 March
2484:2 August
2459:19 April
2423:cite web
2413:19 April
2387:24 March
2325:cite web
2315:24 March
2289:28 April
2263:31 March
2237:24 March
2179:28 April
2154:28 April
2117:24 March
2092:28 April
2063:24 March
1959:24 March
1934:24 March
1877:. Nvidia
1740:See also
1669:4096-bit
1648:SXM/SXM2
1572:4096-bit
1475:4096-bit
1441:TSMC N7
1429:40960 KB
1396:624 TOPS
1378:5120-bit
1344:TSMC N7
1332:40960 KB
1299:624 TOPS
1281:5120-bit
1247:TSMC 4N
1235:51200 KB
1184:5120-bit
1150:TSMC 4N
1138:51200 KB
1087:6144-bit
1081:1980 MHz
1014:7 PFLOPS
1008:3.5 POPS
993:8 TB/sec
990:8192-bit
917:9 PFLOPS
911:4.5 POPS
896:8 TB/sec
893:8192-bit
859:Process
851:Die size
845:L2 Cache
842:L1 Cache
836:(NVLink)
817:bfloat16
676:Dell EMC
574:Mellanox
535:Mellanox
475:OS drive
460:AMD EPYC
416:AMD EPYC
401:Mellanox
352:NVSwitch
3680:DeepMap
3593:Company
3519:GoForce
3429:Console
3306:Maxwell
3281:Rankine
3271:Celsius
2917:shaders
2817:GeForce
2532:1 March
1720:4096 KB
1623:6144 KB
1587:62 TOPS
1526:6144 KB
1490:62 TOPS
651:ExaFLOP
308:Single
294:-based
273:turnkey
112:servers
3710:Stexar
3707:MediaQ
3704:Exluna
3571:nForce
3529:Jetson
3331:Hopper
3326:Ampere
3321:Turing
3311:Pascal
3301:Kepler
3276:Kelvin
3231:NVLink
2911:Vertex
2889:
2804:Nvidia
2687:21 May
2661:21 May
2635:16 May
2576:NVIDIA
2551:NVIDIA
2504:NVIDIA
2058:NVIDIA
2019:NVIDIA
1729:15.3 B
1644:Pascal
1632:21.1 B
1535:21.1 B
1438:54.2 B
1353:Ampere
1341:54.2 B
1256:Ampere
1159:Hopper
1141:1000 W
1062:Hopper
947:1000 W
824:(TF32)
788:(FP64)
784:Double
781:(FP32)
777:Single
769:Memory
764:Memory
759:Memory
727:Socket
692:NetApp
669:NVLink
643:Hopper
638:Top500
500:Hopper
489:Hopper
448:Ampere
395:gen 4
389:Ampere
378:Ampere
302:memory
249:Pascal
192:NVLink
177:Pascal
157:Models
120:Nvidia
3743:GPGPU
3685:Icera
3660:Ageia
3534:Tegra
3524:Drive
3406:GPGPU
3382:cards
3346:Rubin
3316:Volta
3296:Fermi
3291:Tesla
3286:Curie
3178:VDPAU
3153:PhysX
3147:OptiX
3108:NVDEC
3102:NVENC
2915:pixel
1979:Intel
1854:(PDF)
1785:(PDF)
1723:300 W
1714:GP100
1640:P100
1626:300 W
1617:GV100
1547:Volta
1529:350 W
1520:GV100
1450:Volta
1432:400 W
1423:GA100
1335:400 W
1326:GA100
1238:700 W
1229:GH100
1172:16896
1166:16896
1155:H100
1132:GH100
1075:16896
1069:16896
1058:H200
1050:208 B
1044:700 W
1035:GB100
961:B100
953:208 B
938:GB100
864:B200
856:count
801:INT32
761:clock
756:clock
754:Boost
751:cores
749:INT32
746:cores
742:Mixed
734:cores
721:Model
446:Four
348:Volta
342:DGX-2
296:Tesla
292:Volta
290:Four
277:tower
260:Volta
181:Volta
167:DGX-1
18:DGX-1
3544:CPUs
3510:SoCs
3443:Xbox
3438:NV2A
3137:CUDA
3019:800M
3001:NUMA
2928:4 Ti
2913:and
2901:4 MX
2876:TNT2
2715:2022
2689:2022
2663:2022
2637:2022
2609:2022
2583:2023
2558:2022
2534:2023
2511:2022
2486:2023
2461:2022
2435:help
2415:2022
2389:2022
2363:2022
2337:help
2317:2022
2291:2022
2265:2022
2239:2022
2181:2022
2156:2022
2119:2022
2094:2022
2065:2022
2026:2023
1986:2023
1961:2022
1936:2022
1883:2017
1792:2023
1657:3584
1654:1792
1563:5120
1557:2560
1554:5120
1551:SXM2
1466:5120
1460:2560
1457:5120
1454:SXM3
1366:6912
1363:3456
1360:6912
1357:SXM4
1269:6912
1266:3456
1263:6912
1260:SXM4
1244:80 B
1169:4608
1163:SXM5
1147:80 B
1072:4608
1066:SXM5
969:SXM6
872:SXM6
829:FP64
812:FP16
809:FP16
796:INT8
791:INT8
774:VRAM
732:CUDA
730:FP32
546:Xeon
539:NICs
533:via
527:DPUs
504:VRAM
473:NVMe
467:DDR4
462:7742
409:NICs
397:NVMe
393:PCIe
360:DDR4
356:HBM2
300:HBM2
258:The
247:The
212:SSDs
188:HBM2
173:GPUs
147:PCIe
114:and
106:The
3665:ULi
3419:DGX
3225:SXM
3219:MXM
3024:900
3014:700
2990:500
2985:400
2980:300
2975:200
2970:100
2857:NV2
2852:NV1
1708:N/A
1705:N/A
1702:N/A
1699:N/A
1693:N/A
1690:N/A
1687:N/A
1684:N/A
1660:N/A
1651:N/A
1611:N/A
1608:N/A
1605:N/A
1596:N/A
1590:N/A
1560:N/A
1514:N/A
1511:N/A
1508:N/A
1499:N/A
1493:N/A
1463:N/A
1402:N/A
1393:N/A
1369:N/A
1305:N/A
1296:N/A
1272:N/A
1211:N/A
1208:N/A
1205:N/A
1199:N/A
1175:N/A
1114:N/A
1111:N/A
1108:N/A
1102:N/A
1078:N/A
1047:N/A
1041:N/A
1038:N/A
1017:N/A
1011:N/A
1005:N/A
1002:N/A
999:N/A
984:N/A
981:N/A
978:N/A
975:N/A
972:N/A
950:N/A
944:N/A
941:N/A
920:N/A
914:N/A
908:N/A
905:N/A
902:N/A
887:N/A
884:N/A
881:N/A
878:N/A
875:N/A
848:TDP
839:GPU
804:FP4
688:IBM
684:HPE
680:DDN
655:FP8
653:of
647:x86
231:of
179:or
143:SXM
132:x86
3734::
3131:Cg
3063:40
3058:30
3034:16
3029:10
2933:FX
2761:.
2706:.
2680:.
2654:.
2626:.
2600:.
2574:.
2549:.
2502:.
2477:.
2452:.
2427::
2425:}}
2421:{{
2406:.
2380:.
2354:.
2329::
2327:}}
2323:{{
2308:.
2282:.
2256:.
2230:.
2189:^
2172:.
2147:.
2110:.
2085:.
2073:^
2056:.
2017:.
2006:^
1977:.
1952:.
1927:.
1915:^
1873:.
1856:.
698:.
690:,
686:,
682:,
678:,
556:.
525:3
283:.
239:.
3469:)
3465:(
3457:)
3453:(
3445:)
3441:(
2965:9
2943:7
2938:6
2896:2
2824:)
2820:(
2796:e
2789:t
2782:v
2717:.
2691:.
2665:.
2639:.
2611:.
2585:.
2560:.
2513:.
2488:.
2463:.
2437:)
2417:.
2391:.
2365:.
2339:)
2319:.
2293:.
2267:.
2241:.
2183:.
2158:.
2121:.
2096:.
2067:.
2028:.
1988:.
1963:.
1938:.
1885:.
1825:.
1794:.
89:)
83:(
78:)
74:(
70:.
52:.
20:)
Text is available under the Creative Commons Attribution-ShareAlike License. Additional terms may apply.