GPU Cloud
GPU cloud: yi hayar NVIDIA GPU servers ta kowace awa
Yi hayar NVIDIA GPU servers don LLM inference, fine-tuning da kirkirar hotuna. Farashin kai tsaye, biyan kudi na kowace awa, templates na PyTorch, vLLM da ComfyUI, tare da damar shiga ta SSH.
- Kayan aiki na kai tsaye, an tabbatar da farashi kafin ku yi oda
- Templates don PyTorch, vLLM, Ollama da ComfyUI
- Servers na kowace awa da aka tsayar ana cajin kudin disk ne kawai
Tare da Cloudj GPU Cloud zaka iya yin hayan sabar GPU daga kayan da ke akwai a yanzu: zabi samfurin GPU, VRAM da disk, fara da samfuri kamar PyTorch, vLLM, Ollama ko ComfyUI, sannan ka hada hanya ta SSH. Ana cajin sabar da ake biya a kowace awa daga kudin asusunka, kuma sabar da aka dakatar ana cajin kudin disk dinta kawai.
Farashin hayar GPU da wadanda ake da su a yanzu
Kayan da ke akwai a yanzu: samuwa da farashi na canzawa bisa kasuwa. Shafin sarrafawa zai tabbatar da ainihin farashin karshe kafin ka yi oda.
| GPU | VRAM a kowace GPU | Ginin fasaha | Lissafin FP32 | Ƙayyadaddun wutar lantarki | Daga | |
|---|---|---|---|---|---|---|
| NVIDIA RTX PRO 6000 Max-Q | 96 GB | Blackwell | ≈ 119 TFLOPS | 300 W | – | Cikakken bayani → |
| NVIDIA RTX PRO 6000 WS | 96 GB | Blackwell | ≈ 119 TFLOPS | 600 W | – | Cikakken bayani → |
| NVIDIA RTX 5090 | 32 GB | Blackwell | ≈ 107.6 TFLOPS | 575 W | – | Cikakken bayani → |
| NVIDIA RTX 5090D | 32 GB | Blackwell | ≈ 107.6 TFLOPS | 575 W | – | Cikakken bayani → |
| NVIDIA RTX PRO 6000 S | 96 GB | Blackwell | ≈ 93.6 TFLOPS | 600 W | – | Cikakken bayani → |
| NVIDIA RTX 4090 | 24 GB | Ada Lovelace | ≈ 81.4 TFLOPS | 450 W | – | Cikakken bayani → |
| NVIDIA RTX 6000Ada | 48 GB | Ada Lovelace | ≈ 81.4 TFLOPS | 300 W | – | Cikakken bayani → |
| NVIDIA RTX 6000D | 84 GB | Blackwell | ≈ 77.6 TFLOPS | 600 W | – | Cikakken bayani → |
| NVIDIA L40S datacenter | 45 GB | Ada Lovelace | ≈ 73.3 TFLOPS | 350 W | – | Cikakken bayani → |
| NVIDIA RTX PRO 5000 | 48 GB | Blackwell | ≈ 69.6 TFLOPS | 300 W | – | Cikakken bayani → |
| NVIDIA H100 SXM datacenter | 80 GB | Hopper | ≈ 53.5 TFLOPS | 700 W | – | Cikakken bayani → |
| NVIDIA RTX 5080 | 16 GB | Blackwell | ≈ 53.4 TFLOPS | 360 W | – | Cikakken bayani → |
| NVIDIA RTX 4080S | 16 GB | Ada Lovelace | ≈ 50.9 TFLOPS | 320 W | – | Cikakken bayani → |
| NVIDIA RTX 4080 | 16 GB | Ada Lovelace | ≈ 48.6 TFLOPS | 285 W | – | Cikakken bayani → |
| NVIDIA H100 NVL datacenter | 94 GB | Hopper | ≈ 48.3 TFLOPS | 400 W | – | Cikakken bayani → |
| NVIDIA H200 NVL datacenter | 140 GB | Hopper | ≈ 48.3 TFLOPS | 600 W | – | Cikakken bayani → |
| NVIDIA RTX 5070 Ti | 16 GB | Blackwell | ≈ 44.3 TFLOPS | 300 W | – | Cikakken bayani → |
| NVIDIA RTX PRO 4000 | 24 GB | Blackwell | ≈ 44.3 TFLOPS | 145 W | – | Cikakken bayani → |
| NVIDIA RTX 4070S Ti | 16 GB | Ada Lovelace | ≈ 42.6 TFLOPS | 285 W | – | Cikakken bayani → |
| NVIDIA H100 PCIE datacenter | 80 GB | Hopper | ≈ 41 TFLOPS | 350 W | – | Cikakken bayani → |
| NVIDIA RTX 4500Ada | 24 GB | Ada Lovelace | ≈ 38.2 TFLOPS | 210 W | – | Cikakken bayani → |
| NVIDIA RTX 3090 Ti | 24 GB | Ampere | ≈ 36.4 TFLOPS | 450 W | – | Cikakken bayani → |
| NVIDIA RTX A6000 | 48 GB | Ampere | ≈ 36.1 TFLOPS | 300 W | – | Cikakken bayani → |
| NVIDIA RTX 4070S | 12 GB | Ada Lovelace | ≈ 35.8 TFLOPS | 220 W | – | Cikakken bayani → |
| NVIDIA RTX 3090 | 24 GB | Ampere | ≈ 35.4 TFLOPS | 350 W | – | Cikakken bayani → |
| NVIDIA RTX 3080 Ti | 12 GB | Ampere | ≈ 35.1 TFLOPS | 350 W | – | Cikakken bayani → |
| NVIDIA A40 datacenter | 45 GB | Ampere | ≈ 29.9 TFLOPS | 300 W | – | Cikakken bayani → |
| NVIDIA RTX 3080 | 10 GB | Ampere | ≈ 29.5 TFLOPS | 380 W | – | Cikakken bayani → |
| NVIDIA RTX 4070 | 12 GB | Ada Lovelace | ≈ 29.3 TFLOPS | 200 W | – | Cikakken bayani → |
| NVIDIA L4 datacenter | 22 GB | Ada Lovelace | ≈ 24.2 TFLOPS | 72 W | – | Cikakken bayani → |
| NVIDIA RTX 5060 Ti | 16 GB | Blackwell | ≈ 22.8 TFLOPS | 180 W | – | Cikakken bayani → |
| NVIDIA RTX 4060 Ti | 16 GB | Ada Lovelace | ≈ 21.6 TFLOPS | 165 W | – | Cikakken bayani → |
| NVIDIA RTX A4000 | 16 GB | Ampere | ≈ 20.6 TFLOPS | 140 W | – | Cikakken bayani → |
| NVIDIA RTX 3070 | 8 GB | Ampere | ≈ 19.8 TFLOPS | 220 W | – | Cikakken bayani → |
| NVIDIA RTX 3060 Ti | 8 GB | Ampere | ≈ 15.7 TFLOPS | 200 W | – | Cikakken bayani → |
| NVIDIA A100 PCIE datacenter | 40 GB | Ampere | ≈ 15.6 TFLOPS | 250 W | – | Cikakken bayani → |
| NVIDIA A100 SXM4 datacenter | 40 GB | Ampere | ≈ 15.6 TFLOPS | 400 W | – | Cikakken bayani → |
| NVIDIA A800 PCIE datacenter | 80 GB | Ampere | ≈ 15.6 TFLOPS | 300 W | – | Cikakken bayani → |
| NVIDIA RTX 2080 Ti | 11 GB | Turing | ≈ 14.6 TFLOPS | 250 W | – | Cikakken bayani → |
| NVIDIA RTX 3060 laptop | 12 GB | Ampere | ≈ 12.9 TFLOPS | 80 W | – | Cikakken bayani → |
| NVIDIA Tesla V100 datacenter | 16 GB | Volta | ≈ 12.5 TFLOPS | 300 W | – | Cikakken bayani → |
| NVIDIA CMP 170HX datacenter | 64 GB | Ampere | ≈ 12.1 TFLOPS | 250 W | – | Cikakken bayani → |
| NVIDIA RTX 3060 | 12 GB | Ampere | ≈ 12 TFLOPS | 170 W | – | Cikakken bayani → |
| NVIDIA Titan Xp | 12 GB | Pascal | ≈ 11.7 TFLOPS | 160 W | – | Cikakken bayani → |
| NVIDIA RTX A2000 | 6 GB | Ampere | ≈ 11.2 TFLOPS | 70 W | – | Cikakken bayani → |
| NVIDIA GTX 1080 Ti | 11 GB | Pascal | ≈ 11 TFLOPS | 250 W | – | Cikakken bayani → |
| NVIDIA GTX 1080 | 8 GB | Pascal | ≈ 7.8 TFLOPS | 130 W | – | Cikakken bayani → |
| NVIDIA GTX 1070 Ti | 8 GB | Pascal | ≈ 7.4 TFLOPS | 180 W | – | Cikakken bayani → |
| NVIDIA GTX 1650 | 4 GB | ≈ undefined TFLOPS | W | – | Cikakken bayani → | |
| NVIDIA Q RTX 8000 | 45 GB | Turing | ≈ 11.9 TFLOPS | 250 W | – | Cikakken bayani → |
| NVIDIA Tesla P40 datacenter | 24 GB | Pascal | ≈ 9.4 TFLOPS | 250 W | – | Cikakken bayani → |
| NVIDIA RTX 2070S | 8 GB | Turing | ≈ 8.8 TFLOPS | 210 W | – | Cikakken bayani → |
| NVIDIA Q RTX 4000 | 8 GB | Turing | ≈ 7.7 TFLOPS | 125 W | – | Cikakken bayani → |
| NVIDIA RTX 2070 | 8 GB | Turing | ≈ 7.7 TFLOPS | 150 W | – | Cikakken bayani → |
| NVIDIA RTX 2060S | 8 GB | Turing | ≈ 7.5 TFLOPS | 175 W | – | Cikakken bayani → |
| NVIDIA GTX TITAN X | 12 GB | Maxwell | ≈ 6.8 TFLOPS | 150 W | – | Cikakken bayani → |
| NVIDIA RTX 2060 | 12 GB | Turing | ≈ 6.5 TFLOPS | 130 W | – | Cikakken bayani → |
| NVIDIA GTX 1660 Ti | 6 GB | Turing | ≈ 5.3 TFLOPS | 120 W | – | Cikakken bayani → |
| NVIDIA GTX 1660 S | 6 GB | Turing | ≈ 3.6 TFLOPS | 125 W | – | Cikakken bayani → |
GPU templates don ayyukan AI, LLM da rendering
Aiwatar da LLM (LLM inference)
vLLM, SGLang ko Ollama tare da Open WebUI domin gudanarwa da gwada samfuran harshe.
Ingantawa da horarwa (Fine-tuning)
Unsloth Studio domin ingantawa da horar da samfuran harshe bisa bayanan kanka.
Kirkirar hotuna da bidiyo
ComfyUI domin ayyukan Flux, SDXL da bidiyo.
Ginin manhaja na PyTorch da CUDA
Hotunan manhajar PyTorch da Ubuntu 24.04 tare da NVIDIA CUDA, masu sauƙin shiga ta SSH.
Desktop ko cikakkiyar VM
Desktop na Ubuntu a cikin burawza, ko cikakkiyar na'ura mai kama-da-wani (KVM virtual machine) lokacin da kake buƙata.
Yadda ake yin hayar GPU server a matakai 4
- 01
Zabi samfuri da girman disk
- 02
Zabi GPU mai dacewa daga kayan da ke akwai a yanzu
- 03
Tabbatar da farashin sannan ka kaddamar da sabar daga kudin asusunka
- 04
Hada hanya ta SSH ko tashar mashigar intanet
Farashi da tsarin biyan kudin GPU cloud
- Biyan kudi a kowace awa: ana cajin GPU a kowace awa yayin da sabar ke aiki. Sabar da aka dakatar ana caje ta kudin disk dinta kawai.
- Zangon biyan kudi: na wata-wata, kowane wata uku, kowane wata shida, ko na shekara-shekara, tare da nuna duk wani rangwame kafin ka biya.
- Ana cajin ma'ajiyar disk a kowace GB tare da farashin GPU.
- Farashin yana cikin USD; zaka iya nuna su a VND bisa canjin kudi na yanzu.
Tambayoyin da aka fi yi game da GPU cloud
Ayyukan da suka dace da ma'ajiyar GPU: LLM inference da fine-tuning, kirkirar hotuna da bidiyo, rendering da ci gaban aikin CUDA. Templates don PyTorch, vLLM, SGLang, Ollama, Unsloth da ComfyUI suna girka muku manhajar kai tsaye.
Jadawalin da ke sama ya lissafa tsare-tsaren da ake da su a halin yanzu, ciki har da NVIDIA RTX 5090 mai 32 GB na VRAM idan har akwai shi a ajiya. Samuwarsu na sauyawa a tsawon yini.
Farashin yana fitowa ne daga ainihin kayan GPU da ake da su a halin yanzu kuma yana iya canzawa. Console tana nuna tabbataccen farashin kafin ku yi oda, kuma ana cajin server dinku ne a kan wannan tabbataccen farashin.
Kwarai kuwa. Baya ga biyan kudi na kowace awa, za ku iya zabar zagayen wata-wata, kowane wata uku, kowane wata shida ko na shekara, tare da duk wani rangwamen da aka nuna kafin ku biya. Ana cajin ma'ajiyar disk a kowace GB tare da farashin GPU.
A tsarin biyan kudi na kowace awa, server din da aka tsayar ana cajin ma'ajiyar disk dinsa ne kawai. Ku goge server din idan ba ku buƙatar bayanansa kuma.
Ta hanyar SSH tare da key dinku, ko ta terminal a browser dinku daga console. Templates na Desktop da VM suma suna ba da damar shiga mai dauke da hoton allon aiki (graphical access).