Blackwell · SXM module # Rent a dedicated B200 SXM6 GPU server. 180 GB of HBM3e at 8,000 GB/s, in a machine that is yours alone. Monthly, paid in crypto, opened without an identity check and delivered in under 5 minutes. - $11,790/mo one card, everything included - 180 GB HBM3e - 8,000 GB/s memory bandwidth - 4×, 8× node sizes available ## Price, by node size and term One number per line, and it is the whole number: no setup fee, no reinstall fee, no bandwidth charge, no support plan. Longer terms are paid up front. **Monthly price of an B200 SXM6 server, by node size and term length** | Node | CPU and memory | Monthly | 3 months | 12 months | | |---|---|---|---|---|---| | 4 × B200 SXM6 720 GB total · NVLink 5 · 1.8 TB/s | 2 × Intel Xeon Gold 6438Y+64 c · 512 GB · 4 × 3.84 TB NVMe | $11,790/mo cancel any time | $11,201/mo save $1,767/yr | $10,375/mo save $16,980/yr | [Configure](https://gpuserver.io/configure?c=b200-x4) | | 8 × B200 SXM6 1440 GB total · NVLink 5 · 1.8 TB/s | 2 × Intel Xeon Platinum 8462Y+64 c · 1 TB · 8 × 3.84 TB NVMe | $22,351/mo cancel any time | $21,233/mo save $3,354/yr | $19,669/mo save $32,184/yr | [Configure](https://gpuserver.io/configure?c=b200-x8) | Every configuration above is available in all 6 data centres, racked and burned in, with an image ready to write — which is why delivery is under 5 minutes rather than a working day. ## Full specification GPU NVIDIA B200 SXM6 Blackwell architecture, SXM module on an HGX board. Memory 180 GB HBM3e 8,000 GB/s of bandwidth — the number that predicts generation speed, far better than any teraflop figure. Interconnect NVLink 5 · 1.8 TB/s All-to-all between every card on the board. Tensor parallelism without the link becoming the bottleneck. [Which one your job needs](https://gpuserver.io/guides/nvlink-vs-pcie). Node sizes 4 × card, 8 × card SXM cards live on an HGX board of four or eight — never two, never ten. Storage 4 × 3.84 TB NVMe Handed over raw beyond the system volume — we impose no RAID level. More NVMe and archive HDD are monthly options. Network 10 Gbit/s, unmetered Both directions, no included volume and no overage at any traffic level. One IPv4, a routed IPv6 /64, DDoS filtering on from delivery. ## What runs on a B200 SXM6 Every model below fits on a *single* card at 8k context — no sharding, no interconnect to think about. The precision shown is the best one that fits; the arithmetic is in the [VRAM sizing guide](https://gpuserver.io/guides/vram-sizing), so you can check it. **Language models that fit on a single B200 SXM6 at 8k context** | Model | Best precision that fits | Memory needed | Estimated tokens/s | |---|---|---|---| | Llama 3.1 8B 8B parameters | BF16 / FP16Reference quality | 20 GBof 180 GB | ~225 | | Llama 3.3 70B 70B parameters | BF16 / FP16Reference quality | 164 GBof 180 GB | ~26 | | Qwen 3 32B 32B parameters | BF16 / FP16Reference quality | 76 GBof 180 GB | ~56 | | Qwen 3 235B-A22B (MoE) 22B active of 235B | 4-bit (AWQ, GPTQ)Smallest weights, cache stays FP16 | 137 GBof 180 GB | ~59 | | Mixtral 8×22B (MoE) 39B active of 141B | FP8Near-reference, Hopper and Blackwell | 163 GBof 180 GB | ~17 | | Mistral Small 24B 24B parameters | BF16 / FP16Reference quality | 57 GBof 180 GB | ~75 | | Gemma 3 27B 27B parameters | BF16 / FP16Reference quality | 67 GBof 180 GB | ~67 | | Phi-4 14B 14B parameters | BF16 / FP16Reference quality | 34 GBof 180 GB | ~129 | On the 8-card node the same models are held across 1440 GB in total, which changes what is possible entirely — up to DeepSeek V3 671B-A37B (MoE). Sharding costs synchronisation at every layer, so a model that fits on one card should stay on one card. ## Throughput to expect Generation is bounded by memory bandwidth: each token requires re-reading the active weights. At 8,000 GB/s, that sets a ceiling no amount of tuning gets past. **Our figures are a floor, not a promise.** The estimates on this page are calibrated against published measurements and are deliberately conservative — they model single-stream generation. With continuous batching under vLLM the *aggregate* across concurrent requests is several times higher. If a number here looks low against a benchmark you have seen, that is usually the difference. ## Compared with nearby cards The three cards closest to this one in price. Memory decides what loads; bandwidth decides how fast it answers. **The B200 SXM6 compared with the three closest cards by price** | Card | Memory | Bandwidth | Interconnect | From | |---|---|---|---|---| | B200 SXM6 This page Blackwell | 180 GB | 8,000 GB/s | NVLink | $11,790/mo | | [H200 SXM5](https://gpuserver.io/gpu/h200) Hopper | 141 GB | 4,800 GB/s | NVLink | $8,405/mo | | [H100 SXM5](https://gpuserver.io/gpu/h100-sxm5) Hopper | 80 GB | 3,350 GB/s | NVLink | $6,061/mo | | [A100 SXM4](https://gpuserver.io/gpu/a100-sxm4) Ampere | 80 GB | 2,039 GB/s | NVLink | $4,395/mo | [See all 12 cards side by side](https://gpuserver.io/gpu), or let the [configurator](https://gpuserver.io/configure) pick from your model and context length instead of from a price. ## Compared with other cards The arbitrations people actually make against a B200 SXM6, each worked out on price, memory, bandwidth and the models that fit. - [B200 SXM6 vs H200 SXM5 Blackwell against Hopper, at the top See the comparison](https://gpuserver.io/compare/b200-vs-h200) - [B200 SXM6 vs H100 SXM5 Is Blackwell worth the step up? See the comparison](https://gpuserver.io/compare/h100-sxm5-vs-b200) ## What is included - ### No identity check No document, no selfie, no phone number, no company registration. Nothing verified · at any spend - ### The whole card The GPU is passed through to one machine, and that machine is yours. No MIG · no vGPU · no hypervisor - ### Root and IPMI Full root from the first minute, plus remote power and virtual media. Out-of-band on its own VLAN - ### Unmetered bandwidth No included volume, no overage tier, nothing to watch on a graph. 1 to 25 Gbit/s · both directions - ### No setup or reinstall fee Reimage as often as you like, to any operating system we offer. $0 · reinstalls unlimited - ### Hardware swapped, fast From spares held in the same suite, at any hour, disks left in place. Four-hour target · 24/7 ## Where you can have it Every B200 SXM6 configuration is available in all 6 data centres. There is no site where a card costs more, and none where it is unavailable. AMS Available ### Amsterdam Netherlands Round trip 7 ms to Frankfurt, 12 ms to London Facility Tier III · 100% wind STO Available ### Stockholm Sweden Round trip 22 ms to Frankfurt, 31 ms to London Facility Tier III · 100% hydro ZRH Available ### Zürich Switzerland Round trip 9 ms to Milan, 15 ms to Frankfurt Facility Tier IV · 96% carbon-free REY Available ### Reykjavík Iceland Round trip 18 ms to London, 40 ms to New York Facility Tier III · 100% geothermal + hydro YUL Available ### Montréal Canada Round trip 12 ms to New York, 19 ms to Toronto Facility Tier III · 99% hydro SIN Available ### Singapore Singapore Round trip 38 ms to Tokyo, 45 ms to Sydney Facility Tier III · Grid + REC offset ## B200 SXM6 server questions How much does an B200 SXM6 server cost per month? A single-card B200 SXM6 server is $11,790 a month, with everything included: 512 GB of system RAM, 4 × 3.84 TB NVMe, unmetered 10 Gbit/s, one IPv4 address and a routed IPv6 /64. There is no setup fee and no bandwidth charge. Multi-GPU nodes run from $22,351 for 8 cards up to $22,351 for 8. How much VRAM does the B200 SXM6 have, and what fits in it? 180 GB of HBM3e per card, at 8,000 GB/s. The largest model it holds on one card at 8k context is DeepSeek V3 671B-A37B (MoE) (4-bit (AWQ, GPTQ)). A 8-card node has 1440 GB in total, enough for DeepSeek V3 671B-A37B (MoE). Is the B200 SXM6 dedicated, or shared with other customers? Dedicated. The card is passed through to a single physical machine that only you have an account on — no MIG partitioning, no vGPU layer and no other tenant on the same silicon. On a multi-GPU node you get the whole node, including the NVLink 5 · 1.8 TB/s fabric between the cards. Do I need to verify my identity to rent an B200 SXM6? No. There is no document upload, no selfie, no phone number and no company registration, at any level of spend — an eight-card node opens on exactly the same terms as the cheapest single card. Opening an account takes an email address, a password and a billing address, and none of it is verified. How do I pay, and how quickly is the server delivered? In cryptocurrency only: Bitcoin, Monero, Ethereum (ERC-20), Tether (ERC-20) and others. Each invoice gets its own address, the amount is fixed at the rate shown when it is issued, and the machine is provisioned on the first confirmation — root access in under 5 minutes, at any hour. Can I rent an B200 SXM6 by the hour instead? Not here. Every price on this site is a monthly price, with no meter and no per-second billing. Against a typical $2.99/hour cloud, a monthly term is cheaper past roughly 3944 hours of the machine simply existing — about 164 days out of thirty. ## A B200 SXM6 of your own, from $11,790 a month. Racked, burned in and waiting in all 6 data centres. Pay in crypto and it is yours in under 5 minutes. [Configure this server](https://gpuserver.io/configure?c=b200-x4) [Read the guides](https://gpuserver.io/guides) ## Related pages - [Hardware policy The twelve NVIDIA GPUs we operate, the 72-hour burn-in before a node is sold, the four-hour replacement target, and how disks are erased between tenants.](https://gpuserver.io/hardware) - [Guides Eight practical guides: VRAM sizing, NVLink versus PCIe, image and video models, monthly versus hourly, self-hosting versus an API, serving Llama 70B, and QLoRA.](https://gpuserver.io/guides) - [Documentation First SSH, verifying the hardware, CUDA containers, serving a model with vLLM, RAID, firewalling, IPMI and reinstalls — the commands, on a real machine.](https://gpuserver.io/docs) - [Network Unmetered ports up to 25 Gbit/s, two carriers and an IX per site, always-on DDoS filtering, routed IPv6 — and the four things we do not offer, stated up front.](https://gpuserver.io/network) --- Source: https://gpuserver.io/gpu/b200/. This file is generated from the same data as the website; if a figure here differs from a page, the page is authoritative and this file is stale — the canonical source is https://gpuserver.io/.