{"id":184,"date":"2025-04-21T00:38:11","date_gmt":"2025-04-20T16:38:11","guid":{"rendered":"https:\/\/www.zack.cn\/?p=184"},"modified":"2025-04-21T00:38:53","modified_gmt":"2025-04-20T16:38:53","slug":"nvidiadgxsparkaipc","status":"publish","type":"post","link":"https:\/\/www.zack.cn\/index.php\/2025\/04\/21\/nvidiadgxsparkaipc\/","title":{"rendered":"NVIDIA\u53d1\u5e03DGX Spark\u4e2a\u4ebaAI\u8ba1\u7b97\u673a"},"content":{"rendered":"

\u5728GTC\u5927\u4f1a\u4e0a\uff0cNVIDIA2025\u5e743\u670818\u65e5\u63ed\u6653\u4e86\u57fa\u4e8eNVIDIA Grace Blackwell\u5e73\u53f0\u7684NVIDIA DGX\u2122\u4e2a\u4ebaAI\u8d85\u7ea7\u8ba1\u7b97\u673a\u7cfb\u5217\u3002<\/p>\n

\u4ea7\u54c1\u5730\u5740\uff1ahttps:\/\/www.<\/span>nvidia.com\/en-us\/produc<\/span>ts\/workstations\/dgx-spark\/<\/span><\/p>\n

\n
\"NVIDIA\u53d1\u5e03DGX<\/div>
\u642d\u8f7dNVIDIA Grace Blackwell\u5e73\u53f0\uff0c\u53f0\u5f0f\u8d85\u7ea7\u8ba1\u7b97\u673a\u5c06\u52a0\u901fAI\u80fd\u529b\u8d4b\u4e88\u5f00\u53d1\u8005\u3001\u7814\u7a76\u4eba\u5458\u548c\u6570\u636e\u79d1\u5b66\u5bb6\uff1b\u534e\u7855\u3001\u6234\u5c14\u79d1\u6280\u3001\u60e0\u666e\u548c\u8054\u60f3\u7b49\u9886\u5148\u8ba1\u7b97\u673a\u5236\u9020\u5546\u5373\u5c06\u63a8\u51fa\u76f8\u5173\u7cfb\u7edf<\/figcaption><\/figure>\n

\u573a\u666f\uff1a<\/b><\/h2>\n
\n
\"NVIDIA\u53d1\u5e03DGX<\/div>\n<\/figure>\n

\u82f1\u4f1f\u8fbe\u8bbe\u60f3\u7684\u573a\u666f\u5305\u62ec\uff1a\u539f\u578b\u5f00\u53d1\u3001\u5fae\u8c03\u3001\u63a8\u7406\u3001\u6570\u636e\u79d1\u5b66\u3001\u8fb9\u7f18\u8ba1\u7b97\u7b49\u3002\u5f53\u7136\u8bad\u7ec3\u80af\u5b9a\u8fd8\u662f\u63a8\u8350B100\u3001B200\u5566\u3002<\/p>\n

\n
\"NVIDIA\u53d1\u5e03DGX<\/div>\n<\/figure>\n

\u89c4\u683c<\/h2>\n

\u6458\u8981\uff1aGB10\u82af\u7247\u3001128G\u7edf\u4e00\u5185\u5b58\u3001ConnectX<\/p>\n

\n
\"NVIDIA\u53d1\u5e03DGX<\/div>\n<\/figure>\n

NVIDIA DGX Spark\u00a0\u8be6\u7ec6\u89c4\u683c\uff1a<\/p>\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
Architecture<\/td>\nNVIDIA Grace Blackwell<\/td>\n<\/tr>\n
GPU<\/td>\nBlackwell Architecture<\/td>\n<\/tr>\n
CPU<\/td>\n20 core Arm, 10 Cortex-X925 + 10 Cortex-A725 Arm<\/td>\n<\/tr>\n
CUDA Cores<\/td>\nBlackwell Generation<\/td>\n<\/tr>\n
Tensor Cores<\/td>\n5th Generation<\/td>\n<\/tr>\n
RT Cores<\/td>\n4th Generation<\/td>\n<\/tr>\n
Tensor Performance1<\/td>\n1000 AI TOPS<\/td>\n<\/tr>\n
System Memory<\/td>\n128 GB LPDDR5x, unified system memory<\/td>\n<\/tr>\n
Memory Interface<\/td>\n256-bit<\/td>\n<\/tr>\n
Memory Bandwidth<\/td>\n273 GB\/s<\/td>\n<\/tr>\n
Storage<\/td>\n1 or 4 TB NVME.M2 with self-encryption<\/td>\n<\/tr>\n
USB<\/td>\n4x USB 4 TypeC (up to 40Gb\/s)<\/td>\n<\/tr>\n
Ethernet<\/td>\n1x RJ-45 connector
\n10 GbE<\/td>\n<\/tr>\n
NIC<\/td>\nConnectX-7 Smart NIC<\/td>\n<\/tr>\n
Wi-Fi<\/td>\nWiFi 7<\/td>\n<\/tr>\n
Bluetooth<\/td>\nBT 5.3<\/td>\n<\/tr>\n
Audio-output<\/td>\nHDMI multichannel audio output<\/td>\n<\/tr>\n
Power Consumption<\/td>\n170W<\/td>\n<\/tr>\n
Display Connectors<\/td>\n1x HDMI 2.1a<\/td>\n<\/tr>\n
NVENC | NVDEC<\/td>\n1x | 1x<\/td>\n<\/tr>\n
OS<\/td>\nNVIDIA DGX\u2122 OS<\/td>\n<\/tr>\n
System Dimensions<\/td>\n150 mm L x 150 mm W x 50.5 mm H<\/td>\n<\/tr>\n
System Weight<\/td>\n1.2 kg<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

* preliminary specifications, subject to change<\/p>\n

1 Theoretical FP4 TOPS using the sparsity feature.<\/p>\n

\u4ecb\u7ecd<\/h2>\n

\u4ee5\u4e0b\u5185\u5bb9\u6765\u81eaNvidia\u7684\u65b0\u95fb\u7a3f\uff08https:\/\/<\/span>nvidianews.nvidia.com\/n<\/span>ews\/nvidia-announces-dgx-spark-and-dgx-station-personal-ai-computers<\/span>\uff09<\/p>\n

DGX Spark\uff08\u539fProject DIGITS\uff09\u548cDGX Station\u2122\uff08\u642d\u8f7dNVIDIA Blackwell Ultra\u5e73\u53f0\u7684\u65b0\u578b\u9ad8\u6027\u80fdNVIDIA Grace Blackwell\u53f0\u5f0f\u8d85\u7ea7\u8ba1\u7b97\u673a\uff09\uff0c\u4f7fAI\u5f00\u53d1\u8005\u3001\u7814\u7a76\u4eba\u5458\u3001\u6570\u636e\u79d1\u5b66\u5bb6\u548c\u5b66\u751f\u80fd\u591f\u5728\u53f0\u5f0f\u673a\u4e0a\u5bf9\u5927\u578b\u6a21\u578b\u8fdb\u884c\u539f\u578b\u8bbe\u8ba1\u3001\u5fae\u8c03\u548c\u63a8\u7406\u3002\u7528\u6237\u65e2\u53ef\u5728\u672c\u5730\u8fd0\u884c\u8fd9\u4e9b\u6a21\u578b\uff0c\u4e5f\u53ef\u5c06\u5176\u90e8\u7f72\u81f3NVIDIA DGX Cloud\u6216\u5176\u4ed6\u52a0\u901f\u4e91\u6216\u6570\u636e\u4e2d\u5fc3\u57fa\u7840\u8bbe\u65bd\u3002<\/p>\n

DGX Spark\u548cDGX Station\u5c06\u6b64\u524d\u4ec5\u9002\u7528\u4e8e\u6570\u636e\u4e2d\u5fc3\u7684Grace Blackwell\u67b6\u6784\u80fd\u529b\u5f15\u5165\u53f0\u5f0f\u673a\u9886\u57df\u3002\u534e\u7855\u3001\u6234\u5c14\u3001\u60e0\u666e\u548c\u8054\u60f3\u7b49\u5168\u7403\u7cfb\u7edf\u5236\u9020\u5546\u5c06\u53c2\u4e0e\u5f00\u53d1DGX Spark\u548cDGX Station\u3002<\/p>\n

\u201cAI\u5df2\u91cd\u5851\u8ba1\u7b97\u5806\u6808\u7684\u6bcf\u4e00\u5c42\u3002\u56e0\u6b64\uff0c\u4e13\u4e3aAI\u539f\u751f\u5f00\u53d1\u8005\u8bbe\u8ba1\u3001\u7528\u4e8e\u8fd0\u884cAI\u539f\u751f\u5e94\u7528\u7684\u65b0\u578b\u8ba1\u7b97\u673a\u7c7b\u522b\u5e94\u8fd0\u800c\u751f\uff0c\u8fd9\u5408\u4e4e\u903b\u8f91\u3002\u201dNVIDIA\u521b\u59cb\u4eba\u517c\u9996\u5e2d\u6267\u884c\u5b98\u9ec4\u4ec1\u52cb\u8868\u793a\uff0c\u201c\u51ed\u501f\u8fd9\u4e9b\u65b0\u578bDGX\u4e2a\u4ebaAI\u8ba1\u7b97\u673a\uff0cAI\u7684\u5e94\u7528\u8303\u56f4\u5c06\u4ece\u4e91\u670d\u52a1\u6269\u5c55\u81f3\u684c\u9762\u548c\u8fb9\u7f18\u7aef\u3002\u201d<\/p>\n

DGX Spark\u70b9\u71c3\u521b\u65b0<\/b><\/p>\n

DGX Spark\u662f\u5168\u7403\u6700\u5c0f\u7684AI\u8d85\u7ea7\u8ba1\u7b97\u673a\uff0c\u52a9\u529b\u6570\u767e\u4e07\u7814\u7a76\u8005\u3001\u6570\u636e\u79d1\u5b66\u5bb6\u3001\u673a\u5668\u4eba\u5f00\u53d1\u8005\u53ca\u5b66\u751f\u7a81\u7834\u751f\u6210\u5f0fAI\u548c\u7269\u7406AI\u7684\u8fb9\u754c\uff0c\u63d0\u4f9b\u6781\u81f4\u6027\u80fd\u4e0e\u5f3a\u5927\u529f\u80fd\u3002<\/p>\n

DGX Spark\u7684\u6838\u5fc3\u662f\u4e13\u4e3a\u53f0\u5f0f\u673a\u4f18\u5316\u7684NVIDIA GB10 Grace Blackwell\u8d85\u7ea7\u82af\u7247\u3002GB10\u642d\u8f7d\u9ad8\u6027\u80fdNVIDIA Blackwell GPU\uff0c\u652f\u6301\u7b2c\u4e94\u4ee3Tensor Core\u548cFP4\u7cbe\u5ea6\uff0c\u53ef\u4e3a\u6700\u65b0AI\u63a8\u7406\u6a21\u578b\uff08\u5305\u62ecNVIDIA Cosmos Reason\u4e16\u754c\u57fa\u7840\u6a21\u578b\u548cNVIDIA GR00T N1\u673a\u5668\u4eba\u57fa\u7840\u6a21\u578b\uff09\u63d0\u4f9b\u9ad8\u8fbe\u6bcf\u79d21,000\u4e07\u4ebf\u6b21AI\u8ba1\u7b97\uff0c\u7528\u4e8e\u5fae\u8c03\u548c\u63a8\u7406\u3002<\/p>\n

GB10\u8d85\u7ea7\u82af\u7247\u91c7\u7528NVIDIA NVLink\u2122-C2C\u4e92\u8fde\u6280\u672f\uff0c\u5b9e\u73b0CPU+GPU\u534f\u540c\u5185\u5b58\u6a21\u578b\uff0c\u5e26\u5bbd\u8fbe\u7b2c\u4e94\u4ee3PCIe\u76845\u500d\u3002\u8be5\u6280\u672f\u4f7f\u8d85\u7ea7\u82af\u7247\u80fd\u591f\u5728GPU\u548cCPU\u95f4\u9ad8\u6548\u8bbf\u95ee\u6570\u636e\uff0c\u4f18\u5316\u5185\u5b58\u5bc6\u96c6\u578bAI\u5f00\u53d1\u8005\u5de5\u4f5c\u8d1f\u8f7d\u7684\u6027\u80fd\u3002<\/p>\n

\u501f\u52a9NVIDIA\u5168\u6808AI\u5e73\u53f0\uff0cDGX Spark\u7528\u6237\u53ef\u65e0\u7f1d\u5c06\u6a21\u578b\u4ece\u53f0\u5f0f\u673a\u8fc1\u79fb\u81f3DGX Cloud\u6216\u5176\u4ed6\u52a0\u901f\u4e91\/\u6570\u636e\u4e2d\u5fc3\u57fa\u7840\u8bbe\u65bd\u2014\u2014\u51e0\u4e4e\u65e0\u9700\u4fee\u6539\u4ee3\u7801\uff0c\u4f7f\u539f\u578b\u8bbe\u8ba1\u3001\u5fae\u8c03\u548c\u6d41\u7a0b\u8fed\u4ee3\u524d\u6240\u672a\u6709\u7684\u4fbf\u6377\u3002<\/p>\n

DGX Station\u5168\u901f\u524d\u884c<\/b><\/p>\n

NVIDIA DGX Station\u4e3a\u53f0\u5f0f\u673a\u5e26\u6765\u6570\u636e\u4e2d\u5fc3\u7ea7AI\u5f00\u53d1\u6027\u80fd\u3002\u4f5c\u4e3a\u9996\u6b3e\u642d\u8f7dNVIDIA GB300 Grace Blackwell Ultra\u53f0\u5f0f\u8d85\u7ea7\u82af\u7247\u7684\u7cfb\u7edf\uff0cDGX Station\u63d0\u4f9b\u9ad8\u8fbe784GB\u7684\u534f\u540c\u5185\u5b58\u7a7a\u95f4\uff0c\u53ef\u52a0\u901f\u5927\u89c4\u6a21\u8bad\u7ec3\u548c\u63a8\u7406\u5de5\u4f5c\u8d1f\u8f7d\u3002GB300\u53f0\u5f0f\u8d85\u7ea7\u82af\u7247\u914d\u5907\u6700\u65b0\u4e00\u4ee3Tensor Core\u548cFP4\u7cbe\u5ea6\u7684NVIDIA Blackwell Ultra GPU\uff0c\u901a\u8fc7NVLink-C2C\u4e0e\u9ad8\u6027\u80fdNVIDIA Grace\u2122 CPU\u8fde\u63a5\uff0c\u5b9e\u73b0\u4e1a\u754c\u9886\u5148\u7684\u901a\u4fe1\u6548\u7387\u548c\u6027\u80fd\u3002<\/p>\n

DGX Station\u8fd8\u642d\u8f7dNVIDIA ConnectX\u00ae-8 SuperNIC\uff0c\u4e13\u4e3a\u52a0\u901f\u8d85\u5927\u89c4\u6a21AI\u8ba1\u7b97\u5de5\u4f5c\u8d1f\u8f7d\u4f18\u5316\u3002\u652f\u6301\u9ad8\u8fbe800Gb\/s\u7684\u7f51\u7edc\u901f\u5ea6\uff0cConnectX-8 SuperNIC\u63d0\u4f9b\u6781\u901f\u9ad8\u6548\u7684\u7f51\u7edc\u8fde\u63a5\uff0c\u652f\u6301\u591aDGX Station\u9ad8\u901f\u4e92\u8054\u4ee5\u5904\u7406\u66f4\u5927\u89c4\u6a21\u5de5\u4f5c\u8d1f\u8f7d\uff0c\u5e76\u4e3aAI\u4efb\u52a1\u52a0\u901f\u6570\u636e\u4f20\u8f93\u3002<\/p>\n

\u7ed3\u5408DGX Station\u7684\u5c16\u7aef\u80fd\u529b\u4e0eNVIDIA CUDA-X\u2122 AI\u5e73\u53f0\uff0c\u56e2\u961f\u53ef\u5b9e\u73b0\u5353\u8d8a\u7684\u53f0\u5f0f\u673aAI\u5f00\u53d1\u6027\u80fd\u3002\u6b64\u5916\uff0c\u7528\u6237\u8fd8\u53ef\u901a\u8fc7NVIDIA AI Enterprise\u8f6f\u4ef6\u5e73\u53f0\u8bbf\u95eeNVIDIA NIM\u2122\u5fae\u670d\u52a1\uff0c\u8be5\u5e73\u53f0\u63d0\u4f9b\u9ad8\u5ea6\u4f18\u5316\u3001\u6613\u4e8e\u90e8\u7f72\u4e14\u53d7\u4f01\u4e1a\u652f\u6301\u7684\u63a8\u7406\u5fae\u670d\u52a1\u3002<\/p>\n

\u4e0a\u5e02\u65f6\u95f4<\/b><\/p>\n

DGX Spark\u7cfb\u7edf\u4eca\u65e5\u5f00\u653e\u9884\u8ba2\u3002DGX Station\u9884\u8ba1\u5c06\u4e8e\u4eca\u5e74\u665a\u4e9b\u65f6\u5019\u7531\u534e\u7855\u3001BOXX\u3001\u6234\u5c14\u3001\u60e0\u666e\u3001Lambda\u548cSupermicro\u7b49\u5236\u9020\u5546\u63a8\u51fa\u3002<\/p>\n

\u524d\u4e24\u5929\u521a\u8bf4\uff0c\u82f1\u4f1f\u8fbe\u5e94\u8be5\u4e5f\u8981\u51fa\u7c7b\u4f3cRyzen AI MAX+ 395\u6216\u8005\u82f9\u679cMacStudio\u4e4b\u7c7b\u57fa\u4e8e\u7edf\u4e00\u5927\u5185\u5b58\u7684AI PC\u3002\u73b0\u5728\u5c31\u6765\u4e86\u3002\u636e\u4e00\u4e9b\u5206\u6790\u5458\u4f30\u8ba1\u4ef7\u683c3000\u7f8e\u5143\u5de6\u53f3\uff0c\u5927\u5bb6\u89c9\u5f97\u4ef7\u683c\u5982\u4f55\uff1f<\/p>\n","protected":false},"excerpt":{"rendered":"

\u5728GTC\u5927\u4f1a\u4e0a\uff0cNVIDIA2…<\/p>\n","protected":false},"author":1,"featured_media":185,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[6],"tags":[20,24,2],"class_list":["post-184","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-pcbresearch","tag-ai-pc","tag-dgx-spark","tag-nvidia"],"_links":{"self":[{"href":"https:\/\/www.zack.cn\/index.php\/wp-json\/wp\/v2\/posts\/184","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.zack.cn\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.zack.cn\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.zack.cn\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.zack.cn\/index.php\/wp-json\/wp\/v2\/comments?post=184"}],"version-history":[{"count":2,"href":"https:\/\/www.zack.cn\/index.php\/wp-json\/wp\/v2\/posts\/184\/revisions"}],"predecessor-version":[{"id":190,"href":"https:\/\/www.zack.cn\/index.php\/wp-json\/wp\/v2\/posts\/184\/revisions\/190"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.zack.cn\/index.php\/wp-json\/wp\/v2\/media\/185"}],"wp:attachment":[{"href":"https:\/\/www.zack.cn\/index.php\/wp-json\/wp\/v2\/media?parent=184"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.zack.cn\/index.php\/wp-json\/wp\/v2\/categories?post=184"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.zack.cn\/index.php\/wp-json\/wp\/v2\/tags?post=184"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}