---
title: "O‘zbekistonda GPU ijarasi: xususiyatlar, dastlabki kirish | UZ Dynamics"
description: "Ariza orqali dastlabki kirish: NVIDIA B300, B200, H200, H100 va boshqa GPU xususiyatlari rasmiy manbalardan. Mavjudlik va form-faktor aniqlanmoqda."
url: "https://uzdynamics.ai/uz/gpus"
lang: "uz"
image: "https://uzdynamics.ai/uz/opengraph-image/card?1237d0f845bd82b9"
alternates:
  ru: "https://uzdynamics.ai/ru/gpus"
  uz: "https://uzdynamics.ai/uz/gpus"
  en: "https://uzdynamics.ai/en/gpus"
---

> For an index of this site's pages, see [llms.txt](https://uzdynamics.ai/llms.txt). A markdown copy of each page is at its address with `.md` appended.

Dastlabki kirish uchun arizalar

# GPU ijarasi: katalog va xususiyatlar

Xususiyatlar NVIDIA rasmiy materiallaridan olingan. Har bir modelning mavjudligi va form-faktorini aniqlashtirmoqdamiz.

[GPU tanlashni ochish](https://uzdynamics.ai/uz/sizing)

Xotira har bir GPU uchun ko‘rsatilgan. Bir nechta GPU xotirasi alohida: katta model uchun mos dasturiy ta’minot va modelni kartalar orasida taqsimlash kerak. NVLink ularni bitta GPU qilmaydi.

Narxlarni hozircha e’lon qilmaymiz: shartlar va mavjudlikni arizaga javobda yuboramiz.

**GPU katalogi va xususiyatlar**

| GPU | Har bir GPU xotirasi | Xotira o‘tkazuvchanligi | FP16/FP8, TFLOPS siyrakliksiz | FP64, TFLOPS | Kartalar orasidagi aloqa | Mos vazifalar (kalkulyator bahosi) | Amallar |
| --- | --- | --- | --- | --- | --- | --- | --- |
| 96 GB va undan ko‘p |  |  |  |  |  |  |  |
| B300 Blackwell, SXM (aniqlanmoqda) 270 GB HBM3e 123 mlrd parametrgacha LLM (FP8), bitta kartada, bir nechta kartada o‘qitish (NVLink) | 270 GB HBM3e | 7,7 TB/s | 2 250/4 500 | 1,2 | NVLink 1,8 TB/s | 123 mlrd parametrgacha LLM (FP8), bitta kartada, bir nechta kartada o‘qitish (NVLink) | [Tafsilotlarni ochish](https://uzdynamics.ai/uz/gpus/b300) [Ariza berish](https://uzdynamics.ai/uz/request?source=catalog%3Ab300&gpu=b300) |
| B200 Blackwell, SXM (aniqlanmoqda) 180 GB HBM3e 123 mlrd parametrgacha LLM (FP8), bitta kartada, bir nechta kartada o‘qitish (NVLink), FP64 hisob-kitoblari | 180 GB HBM3e | 7,7 TB/s | 2 250/4 500 | 37 | NVLink 1,8 TB/s | 123 mlrd parametrgacha LLM (FP8), bitta kartada, bir nechta kartada o‘qitish (NVLink), FP64 hisob-kitoblari | [Tafsilotlarni ochish](https://uzdynamics.ai/uz/gpus/b200) [Ariza berish](https://uzdynamics.ai/uz/request?source=catalog%3Ab200&gpu=b200) |
| H200 Hopper, SXM (aniqlanmoqda) 141 GB HBM3e 70 mlrd parametrgacha LLM (FP8), bitta kartada, bir nechta kartada o‘qitish (NVLink), FP64 hisob-kitoblari | 141 GB HBM3e | 4,8 TB/s | 989,4/1 978,9 | 34, tenzor yadrolarida 67 | NVLink 900 GB/s | 70 mlrd parametrgacha LLM (FP8), bitta kartada, bir nechta kartada o‘qitish (NVLink), FP64 hisob-kitoblari | [Tafsilotlarni ochish](https://uzdynamics.ai/uz/gpus/h200) [Ariza berish](https://uzdynamics.ai/uz/request?source=catalog%3Ah200&gpu=h200) |
| RTX PRO 6000 Blackwell, Server Edition (aniqlanmoqda) 96 GB GDDR7 70 mlrd parametrgacha LLM (FP8), bitta kartada, 3D-render, videoni kodlash | 96 GB GDDR7 | 1,6 TB/s | 467,8/935,6 | ko‘rsatilmagan | PCIe 5, NVLink yo‘q | 70 mlrd parametrgacha LLM (FP8), bitta kartada, 3D-render, videoni kodlash | [Tafsilotlarni ochish](https://uzdynamics.ai/uz/gpus/rtx-pro-6000) [Ariza berish](https://uzdynamics.ai/uz/request?source=catalog%3Artx-pro-6000&gpu=rtx-pro-6000) |
| 80 GB |  |  |  |  |  |  |  |
| H100 Hopper, SXM (aniqlanmoqda) 80 GB HBM3 32 mlrd parametrgacha LLM (BF16), bitta kartada, bir nechta kartada o‘qitish (NVLink), FP64 hisob-kitoblari | 80 GB HBM3 | 3,35 TB/s | 989,4/1 978,9 | 34, tenzor yadrolarida 67 | NVLink 900 GB/s | 32 mlrd parametrgacha LLM (BF16), bitta kartada, bir nechta kartada o‘qitish (NVLink), FP64 hisob-kitoblari | [Tafsilotlarni ochish](https://uzdynamics.ai/uz/gpus/h100) [Ariza berish](https://uzdynamics.ai/uz/request?source=catalog%3Ah100&gpu=h100) |
| A100 Ampere, SXM (aniqlanmoqda) 80 GB HBM2e 32 mlrd parametrgacha LLM (BF16), bitta kartada, bir nechta kartada o‘qitish (NVLink), FP64 hisob-kitoblari | 80 GB HBM2e | 2,04 TB/s | 312/yo‘q | 9,7, tenzor yadrolarida 19,5 | NVLink 600 GB/s | 32 mlrd parametrgacha LLM (BF16), bitta kartada, bir nechta kartada o‘qitish (NVLink), FP64 hisob-kitoblari | [Tafsilotlarni ochish](https://uzdynamics.ai/uz/gpus/a100) [Ariza berish](https://uzdynamics.ai/uz/request?source=catalog%3Aa100&gpu=a100) |
| 48 GB |  |  |  |  |  |  |  |
| L40S Ada Lovelace, PCIe (aniqlanmoqda) 48 GB GDDR6 32 mlrd parametrgacha LLM (FP8), bitta kartada, 3D-render, videoni kodlash | 48 GB GDDR6 | 0,86 TB/s | 362,05/733 | ko‘rsatilmagan | PCIe 4, NVLink yo‘q | 32 mlrd parametrgacha LLM (FP8), bitta kartada, 3D-render, videoni kodlash | [Tafsilotlarni ochish](https://uzdynamics.ai/uz/gpus/l40s) [Ariza berish](https://uzdynamics.ai/uz/request?source=catalog%3Al40s&gpu=l40s) |
| RTX 6000 Ada Ada Lovelace, Workstation (aniqlanmoqda) 48 GB GDDR6 32 mlrd parametrgacha LLM (FP8), bitta kartada, 3D-render, videoni kodlash | 48 GB GDDR6 | 0,96 TB/s | 364/728,5 | ko‘rsatilmagan | PCIe 4, NVLink yo‘q | 32 mlrd parametrgacha LLM (FP8), bitta kartada, 3D-render, videoni kodlash | [Tafsilotlarni ochish](https://uzdynamics.ai/uz/gpus/rtx-6000-ada) [Ariza berish](https://uzdynamics.ai/uz/request?source=catalog%3Artx-6000-ada&gpu=rtx-6000-ada) |
| RTX A6000 Ampere, Workstation (aniqlanmoqda) 48 GB GDDR6 14 mlrd parametrgacha LLM (BF16), bitta kartada, 3D-render, videoni kodlash | 48 GB GDDR6 | 0,77 TB/s | 154,8/yo‘q | ko‘rsatilmagan | 2 ta karta uchun NVLink ko‘prigini qo‘llab-quvvatlaydi | 14 mlrd parametrgacha LLM (BF16), bitta kartada, 3D-render, videoni kodlash | [Tafsilotlarni ochish](https://uzdynamics.ai/uz/gpus/rtx-a6000) [Ariza berish](https://uzdynamics.ai/uz/request?source=catalog%3Artx-a6000&gpu=rtx-a6000) |
| 24 GB |  |  |  |  |  |  |  |
| L4 Ada Lovelace, PCIe (aniqlanmoqda) 24 GB GDDR6 14 mlrd parametrgacha LLM (FP8), bitta kartada, 3D-render, videoni kodlash | 24 GB GDDR6 | 0,3 TB/s | 121/242 | ko‘rsatilmagan | PCIe 4, NVLink yo‘q | 14 mlrd parametrgacha LLM (FP8), bitta kartada, 3D-render, videoni kodlash | [Tafsilotlarni ochish](https://uzdynamics.ai/uz/gpus/l4) [Ariza berish](https://uzdynamics.ai/uz/request?source=catalog%3Al4&gpu=l4) |
| RTX A5000 Ampere, Workstation (aniqlanmoqda) 24 GB GDDR6 8 mlrd parametrgacha LLM (BF16), bitta kartada, 3D-render, videoni kodlash | 24 GB GDDR6 | 0,77 TB/s | 111,1/yo‘q | ko‘rsatilmagan | 2 ta karta uchun NVLink ko‘prigini qo‘llab-quvvatlaydi | 8 mlrd parametrgacha LLM (BF16), bitta kartada, 3D-render, videoni kodlash | [Tafsilotlarni ochish](https://uzdynamics.ai/uz/gpus/rtx-a5000) [Ariza berish](https://uzdynamics.ai/uz/request?source=catalog%3Artx-a5000&gpu=rtx-a5000) |

† NVIDIA ma’lumotlariga ko‘ra foydalanish mumkin bo‘lgan xotira. Chip xotirasining eng katta hajmi — 288 GB.

‡ NVIDIA ma’lumotlariga ko‘ra foydalanish mumkin bo‘lgan xotira. Chip xotirasining eng katta hajmi — 192 GB.

\* Siyrakliksiz FP16/BF16 va FP8 — NVIDIA texnik hujjatidagi H100 SXM5 ustunidan: H200 uchun NVIDIA faqat siyraklik bilan olingan qiymatlarni beradi, ular H100 SXM qiymatlari bilan bir xil.

\*\* Hisob: NVIDIA texnik hujjatidagi har bir SM uchun tenzor yadrolari unumdorligi × mahsulot tavsifidagi 2 430 MHz Boost chastotasi. Server Edition uchun NVIDIA FP16/FP8/FP4 ko‘rsatkichlarini faqat yaxlitlangan 1/2/4 PFLOPS maksimal qiymatlari sifatida keltiradi. Bu siyraklikdan foydalanilgandagi qiymatlar darajasiga mos; siyrakliksiz qiymatlar e’lon qilinmagan.

\*\*\* Hisob: NVIDIA RTX A5000 uchun ko‘rsatgan siyraklik bilan FP16 qiymatining yarmi (222,2 TFLOPS); siyrakliksiz qiymatni NVIDIA keltirmaydi.

FP64 «ko‘rsatilmagan»: bu kartalarning spetsifikatsiyalarida NVIDIA qiymatni keltirmaydi; NVIDIA texnik hujjatlariga ko‘ra GA102, AD102 va GB202 chiplarida FP64 tezligi — FP32 ning 1/64 qismi.

Xususiyatlar NVIDIA rasmiy materiallari bo‘yicha, 26-sentabr, 2026 sanasida tekshirilgan.

[GPU tanlashni ochish](https://uzdynamics.ai/uz/sizing)

## GPU tanlash bo‘yicha qo‘llanmalar

- [LLM inferensi uchun GPU qanday tanlanadi](https://uzdynamics.ai/uz/guides/llm-inference)
- [LLM uchun VRAM: vaznlar, KV kesh va kontekst](https://uzdynamics.ai/uz/guides/vram-kv-cache)
- [LoRA va QLoRA uchun GPU konfiguratsiyasini tayyorlash](https://uzdynamics.ai/uz/guides/lora-qlora)
- [Bulutli GPU, ajratilgan server yoki API: qaysi birini tanlash kerak?](https://uzdynamics.ai/uz/guides/gpu-cloud-vs-dedicated-vs-api)
- [GPU ijarasi xarajatini hisoblash: hisoblash resurslari, saqlash va bekor turish](https://uzdynamics.ai/uz/guides/gpu-rental-cost)
- [RTX 4090 yoki L40S: AI uchun qaysi GPUni ijaraga olish kerak?](https://uzdynamics.ai/uz/guides/rtx-4090-vs-l40s)

[Barcha qo‘llanmalar](https://uzdynamics.ai/uz/guides)

[Dastlabki kirish uchun ariza](https://uzdynamics.ai/uz/request?source=gpus)

```json
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"UZ Dynamics","item":"https://uzdynamics.ai/uz"},{"@type":"ListItem","position":2,"name":"GPU katalogi","item":"https://uzdynamics.ai/uz/gpus"}]}
```