RENDA™ Nova8 Servers
Eight full-power NVIDIA RTX PRO 6000 Blackwell GPUs in one 4U server, liquid-cooled end-to-end by Nexalus. Built to serve models at scale, on your premises or in your data centre.
The Nova8: liquid-cooled 8-GPU servers for AI inference and training. Built and tuned by RENDA™, cooled end-to-end by Nexalus, and optimised for your business to maximise your tokenomics. Whether it’s operating as a single server, full rack or data centre, the RENDA™ Nova8 is flexible and scalable around your business’ needs.
RENDA™ Nova8 Benchmarks
See how the Nova8 performs over the course of a typical hour, based on 175 back-to-back runs with all eight GPUs pushed to their limits. Throughput held at ~54,000 tokens per second and the hottest GPU stayed at 80°C or below.
Server throughput Replay
GPU Output
Every card averaged more than 6,600 tokens a second across the hour.
| GPU | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
|---|---|---|---|---|---|---|---|---|
| Hour avg, tok/s | 6,623 | 6,914 | 6,975 | 6,627 | 6,759 | 6,640 | 6,719 | 6,726 |
| Avg temp | 77 °C | 71 °C | 74 °C | 72 °C | 73 °C | 77 °C | 74 °C | 74 °C |
| Peak temp | 80 °C | 74 °C | 78 °C | 75 °C | 76 °C | 80 °C | 77 °C | 76 °C |
Benchmark Comparisons
We tested the RENDA™ Nova8 against a similar tier-1 server. Here is the side-by-side comparison of hardware, LLM model, and token performance:
| RENDA™ Nova8 (5th October 2026) | Tier-1 air-cooled reference (published) | |
|---|---|---|
| GPUs | 8x RTX PRO 6000 Blackwell Workstation Edition | 8x RTX PRO 6000 Blackwell Server Edition |
| CPU | 1x AMD EPYC | 2x AMD EPYC |
| Cooling | Water cooled | Air cooled |
| LLM Model | Llama 3.3 70B Instruct, NVFP4 | Llama 3.3 70B, FP4 |
| Engine | TensorRT-LLM 1.3.0rc29 | TensorRT (version not stated) |
| Layout | 1 Model replica per GPU, 512 concurrent requests each (4,096 total) | 2 Replicas per GPU, 8,192 concurrent users |
| Tokens | 256 input / 256 output | Not stated |
The RENDA™ Nova8 achieved the following results when compared to the tier-1 reference above in a variety of scenarios.
| Scenario | Clocks | Engine | 8-GPU Output (Tokens per Second) | vs. Reference |
|---|---|---|---|---|
| Out of the box (Cold start) | Stock | rc20 | 49,301 | +8% |
| Out of the box (Cold start) | Stock | rc29 | 51,705 | +14% |
| Tuned | Core +600 MHz, lock 2,550, RAM +1,400MHz | rc29 | 54,564 | +20% |
| One-hour run | Tuned | rc29 | 53,984 avg / 54,968 peak | +19% / +21% |
| Reference | Published | – | 45,527 | – |
All scenarios tested with eight GPUs configured with a 600W power limit. Tuned runs set GPU clock offsets through NVIDIA’s management tools, with no further hardware modifications. Single-scenario tests captured using an average of 5 runs.
-
01
Eight full-power Blackwell GPUs
RTX PRO 6000, 96GB each, at the full 600 W. Single-slot liquid blocks put eight in one 4U chassis.
-
02
Jet-impingement CPU block
Coolant aimed at the hot spots for the lowest thermal resistance at the chip.
-
03
Balanced manifold
Equal-length runs so every GPU gets the same flow. No weak slot.
-
04
Redundant micropumps
10 aerospace-grade micropumps in 5 loops. Lose one and the server keeps running.
-
05
High-density radiator and 16 fans
The largest radiator the 4U frontal area can accomodate, with many small fans for redundancy.
-
06
Medical-grade tubing
Long-life tubing for low-leaching and leak-proof, low-pressure operation.
Running Cost & ROI Calculator
See how the cost of running a RENDA™ Nova8 compares to traditional air-cooled solutions. Simply set your fleet size and energy price below to see how the RENDA™ Nova8 can improve your tokenomics.
tokens served per year
less electricity per year, same tokens
energy cost saved per year
CO₂ avoided per year
air-cooled servers needed for the same output
heat available for reuse per year
Estimate. Uses the one-hour average (53,984 tokens/s), an estimated 6.4 kW per server, the published air-cooled reference (45,527 tokens/s, 7.03 tokens/s per W) and 90% heat capture (design value).¹
Interactive Tools
Eight tools to help you decide which RENDA Nova8 configuration and setup works best for you.
Hardware & Specifications
Here is a full breakdown of the hardware featured within the RENDA Nova8:
| GPUs | 8 × NVIDIA RTX PRO 6000 Blackwell, 96GB GDDR7 with ECC each, full 600W, no de-rating |
| GPU memory | 768GB total (1.8TB/s per GPU) |
| CPU | AMD EPYC |
| Memory | Up to 2TB DDR5 RDIMM |
| Storage | 8 × U.2 PCIe Gen5 NVMe |
| Power | 4 × 3kW 80 PLUS Titanium, 2+2 redundant · ~6.4kW typical at full load (estimate) |
| Cooling | Direct liquid cooling on every GPU and the CPU · jet-impingement cold plates · 10 aerospace-grade micropumps in 5 redundant loops · 16 fans · balanced flow · medical-grade tubing |
| Form factor | 4U, standard 19-inch rack · optional rear quick-connects for heat export |
| Software | Linux with NVIDIA drivers and TensorRT-LLM, pre-installed and benchmarked · compatible with vLLM, NVIDIA NIM and other CUDA inference stacks |
| Performance | Up to 55,000 tokens/s · 53,984 tokens/s one-hour average · Llama 3.3 70B NVFP4 |
Book A Demo
Pick a day and time, and a member of our team will be in touch.
Why Choose A RENDA™ Nova8 Server
Personalised Peak Performance
The RENDA™ team harnesses precise needs analysis and expert knowledge to design professional computers. This process means we can optimise from a price-to-performance standpoint for your specific software needs. Leveraging our expertise and in-depth knowledge of the complete computer hardware universe, we hand-pick each component, ensuring they are optimal for your needs and function in perfect harmony within your bespoke system. With further performance increases driven by advanced water-cooling and firmware fine-tuning, we deliver systems that exceed expectations, all within your budget.
Improved Productivity
When faced with relentless competition, RENDA™ propels your business forward by supercharging productivity. Our professional servers and workstation PCs are engineered for performance and reliability, turning downtime into uptime and delay into delivery. This means client projects are returned faster, amplifying satisfaction and service acclaim. Moreover, RENDA™ understands that people are at the heart of every business. By equipping them with our lightning-fast systems, we ignite a sense of workplace fulfilment, anchoring talent and nurturing a culture that thrives on continual improvement and innovation.
Enhanced Reliability and Longevity
Experience unwavering reliability with a RENDA™ server or workstation PC, each designed to thrive under pressure. Our systems feature hand-picked components, complemented by advanced water-cooling solutions and bespoke firmware optimisations, to set the new standard in operational reliability. These same innovations bolster component longevity, as the synergy of optimal hardware, cooling efficiency, and tailored firmware significantly extends the durable lifespan of each unit. Maximise the return on your hardware investments.
Holistic Resource Optimisation
In the realm of your enterprise, significant annual investments in personnel and software licenses are a given. RENDA™ addresses this challenge by providing bespoke systems that turbo-charge the productivity of your team and the software they wield. This strategic enhancement is not about cost. It’s about investment with a panoramic impact. The real merit of purchasing a RENDA™ system unfolds when you assess the combined costs of your workforce, technology, and software. It’s here that the value of RENDA™ truly shines, elevating the collective output and efficiency of every operational facet you’ve already financially committed to.
RENDA™ Nova8 Frequently Asked Questions
Can I customise the RENDA™ Nova8?
What can I use a RENDA™ Nova8 2U for?
The RENDA™ Nova8 is designed for those looking to maximise their AI computing power, whether that be on the premises to run your own models, or in a data centre to power your client’s projects in the cloud.
How do I get in contact with the RENDA™ expert team?
Contact Us Today
There has never been a better time to turbo-charge your business technology so drop us a call or email today. We’re always happy to talk about how we can improve and streamline your operations.