Yi hayar NVIDIA Tesla V100 GPU mai 16 GB VRAM
Yi hayar sabar GPU ta gajimare ta Tesla V100 a duk sa'a a kan Cloudj: 16 GB VRAM, NVIDIA Volta da kuma kusan 780 GB/s na memory bandwidth tare da 12.5 TFLOPS FP32. V100 ta kawo Tensor Cores kuma ta kasance ingantacciyar katin datacenter don horar da FP16 mixed-precision, sake gwada sakamakon bincike, da gudanar da inference mai matsakaicin girma a kan iyakar ƙarfin 300 W. Zaɓi samfurin PyTorch, duba buƙatar CUDA dinta, sannan ka haɗa ta hanyar SSH.
Bayanan fasaha na Tesla V100
VRAM a kowace GPU
16 GB
Ginin fasaha
NVIDIA Volta
Ikon lissafi (Compute capability)
7
Lissafin FP32
≈ 12.5 TFLOPS
Gudun ƙwaƙwalwa (Bandwidth)
≈ 780 GB/s
Ƙayyadaddun wutar lantarki
300 W
Ƙimomin da mai samar da kayan aiki ya bayar don na'urorin da aka jera a yanzu; ainihin aikin ya bambanta dangane da na'ura.
Tsare-tsare da farashin Tesla V100 na yanzu
Kayan da ke akwai da farashin kai tsaye, ana sabunta su akai-akai. Ana tabbatar da farashin ƙarshe kafin ka yi oda.
Yi haya yanzuAbubuwan da za ka iya gudanarwa a kan cloud GPU na Tesla V100
Horar da FP16 mixed-precision
Volta Tensor Cores na hanzarta horar da FP16 don CNNs, ƙananan transformers, da samfuran sauti.
Sake fitar da sakamakon da aka buga
An horar da takardun bincike da ma'aunan gwaji da yawa a kan V100s; sake gudanar da su a kan tsari iri ɗaya don samun alkaluma masu daidai.
7B LLMs da aka yi wa quantized
16 GB na iya gudanar da samfuran 7B–8B a 8-bit ko samfuran aji na 13B a 4-bit don gwaji da sauƙaƙan hidimar serving.
HPC da lissafin kimiyya
780 GB/s na bandwidth na taimakawa simulations, solvers, da lambobin kimiyya masu dogaro da ƙwaƙwalwa.
Shin Tesla V100 ita ce GPU da ta dace da aikinku?
16 GB Tesla V100 ta dace da masu bincike da masu haɓaka manhaja waɗanda ke buƙatar babban memory bandwidth don ƙanana zuwa matsakaitan samfura da ayyukan kimiyya ba tare da biyan kuɗin sabbin kayan aikin datacenter ba. Zaɓi ƙaramar GPU ta yau da kullun don ayyukan haɓakawa na asali masu ƙarancin buƙatar ƙwaƙwalwa, ko kuma ka koma zuwa GPU na sabon zamani mai babban VRAM idan samfuranka sun haura 16 GB ko kuma suna buƙatar sabbin fasalullukan tsari fiye da compute capability 7.0.
Tesla V100 idan aka kwatanta da sauran cloud GPU
| GPU | VRAM a kowace GPU | Lissafin FP32 | Ƙayyadaddun wutar lantarki | |
|---|---|---|---|---|
| NVIDIA Tesla V100 | 16 GB | ≈ 12.5 TFLOPS | 300 W | |
| NVIDIA L4 | 22 GB | ≈ 24.2 TFLOPS | 72 W | Cikakken bayani → |
| NVIDIA Tesla P40 | 24 GB | ≈ 9.4 TFLOPS | 250 W | Cikakken bayani → |
| NVIDIA A100 PCIE | 40 GB | ≈ 15.6 TFLOPS | 250 W | Cikakken bayani → |
| NVIDIA A100 SXM4 | 40 GB | ≈ 15.6 TFLOPS | 400 W | Cikakken bayani → |
Cloud GPU na Tesla V100: tambayoyin da aka fi yi
Yi hayar cloud GPU na Tesla V100 a kowace sa'a
Zaɓi samfuri, zaɓi tsari na kai tsaye, sannan ka haɗa ta SSH ko terminal na browser. Caji na tsayawa da zarar ka goge GPU ɗin.