Will this reduce noise on my RTX 5090 during AI workloads?

Yes. Stock RTX 5090 air coolers run at 45-65 dBA under sustained AI inference loads (Ollama, Stable Diffusion, ComfyUI). A full-cover waterblock like this mounted to a 360mm radiator typically reduces noise to under 30 dBA while also lowering VRAM temperatures from 90°C+ to 60-70°C, preventing throttling during long generation runs.

What else do I need for a complete water cooling loop?

A complete custom loop requires: this GPU waterblock, a radiator (360mm recommended for single GPU, 480mm for GPU+CPU), a D5 or DDC pump with reservoir combo, G1/4 fittings, tubing, and premixed coolant. FormulaMod sells complete kits starting at $249 USD including pump, radiator, and fittings.

Does this waterblock cover the VRAM and VRM chips?

Full-cover waterblocks cool the GPU die, VRAM modules, and VRM components simultaneously via a single block with an integrated cold plate. This is essential for AI workloads where sustained 100% VRAM utilization can cause memory junction temperatures to exceed 110°C on air cooling. The full-cover design distributes coolant flow across the entire PCB thermal zone.

1 5

Barrow Galax Gainward 5090 Ta Xue Full Coverage GPU Water Block Cooler BS-GAD5090-PA

$143.15 USD

Sale Sold out

Shipping calculated at checkout.

Lighting System 5v ARGB

Color Black White

SKU:BS-GAD5090-PA-BK

RTX 5090 AI Server Waterblock

A full-cover liquid cooling block for NVIDIA RTX 5090 accelerators deployed in AI inference nodes, training workstations, and HPC benches running at sustained full power. Validated to the card's 575W TDP ceiling with margin for short-term spikes.

Compatible GPU SKUs

Fits RTX 5090 cards built on the NVIDIA reference PCB:

NVIDIA RTX 5090 Founders Edition
ASUS ProArt / TUF RTX 5090 (reference PCB variants)
Gigabyte RTX 5090 Gaming / Aero (reference PCB variants)
MSI RTX 5090 Ventus / Gaming Trio (reference PCB variants)
Palit / Gainward RTX 5090 (reference PCB variants)
Zotac RTX 5090 Trinity (reference PCB variants)

Non-reference PCB cards (extended boards, custom power delivery) are NOT compatible. If you are unsure which PCB your card uses, send a photo of the bare board to customermanager@formulamod.com before ordering.

Technical Specifications

Specification	Value
Supported TDP ceiling	575W sustained
GPU die contact area	~400 mm² polished copper
Coldplate material	Nickel-electroplated C1100 copper
Top material	POM acetal (black)
G1/4" ports	4 (two inlet, two outlet)
O-rings	EPDM, factory leak-tested
Compatible coolants	Water, propylene glycol mixes, non-conductive dielectric fluids

Included In The Box

RTX 5090 waterblock assembly (coldplate + POM top, pre-assembled)
Pre-cut thermal pad set for VRAM, VRM, and inductors (1.0mm, 1.5mm, 2.0mm)
Thermal paste sachet
Mounting hardware (screws, standoffs, backplate)
Installation guide

5-Step Installation Checklist

Power down and remove the stock cooler. Disconnect power, remove the GPU, and unscrew the stock heatsink. Clean the die and VRAM modules with isopropyl alcohol.
Apply thermal paste and pads. Place a rice-grain dot of paste on the die. Apply pre-cut pads to VRAM, VRM, and inductors per the included diagram.
Mount the waterblock. Align the block over the PCB using the mounting holes. Hand-tighten screws in a diagonal cross pattern, then torque to snug. Do not overtighten.
Leak test before installation. Connect the block to your loop. Fill with distilled water only. Run the pump for 30 minutes outside the system. Inspect every port and seam.
Install and verify. Mount the GPU, connect power, run a sustained load (FurMark, ML training, or equivalent). Idle should sit near coolant temperature. Full-load junction should hold below 55C with adequate radiator capacity.

Academic & B2B Research Support

B2B inquiries from universities, research labs, and institutional buyers using a verified institutional email address receive a 10% educational discount and a 48-hour technical response SLA on compatibility and integration questions. Contact customermanager@formulamod.com with your institution name and project scope to initiate a quote.

Frequently Asked Questions

Is this waterblock compatible with my RTX 5090?

This Barrow waterblock is designed for the RTX 5090. Check the product title for specific card brand compatibility — non-reference PCB designs (ASUS Strix, MSI Suprim, etc.) require model-specific blocks. Reference/Founders Edition cards use a different block.

How much quieter will my RTX 5090 be with water cooling?

Stock RTX 5090 coolers run at 45-65 dBA under sustained AI workloads like Ollama or Stable Diffusion. A full-cover waterblock reduces noise to under 30 dBA — quieter than a refrigerator. VRAM temperatures also drop from 90°C+ to 60-70°C, preventing thermal throttling during long inference sessions.

What else do I need for a complete loop?

Besides this waterblock, a complete custom loop needs: a radiator (360mm minimum for 575W TDP), a D5 or DDC pump with reservoir, G1/4 fittings (8-12 pieces), tubing (soft or hard), and coolant. FormulaMod carries all components — browse our AI Workstation Cooling collection for complete setups.

Can I run this 24/7 for AI inference?

Yes. Water cooling is designed for sustained loads, unlike air coolers which rely on fan speeds that create noise. A properly maintained loop can run 24/7 indefinitely with coolant changes every 12-18 months.

Does FormulaMod ship to the USA and Europe?

Yes. We ship worldwide from our Shenzhen warehouse. USA and EU orders typically arrive within 7 days. All products are covered by our return policy.

Why Water Cool the RTX 5090?

575W TDP under AI load

32GB VRAM (throttles at 92°C)

<30dB With water cooling

- 575W TDP is the highest ever in a consumer GPU — stock triple-fan coolers hit 55+ dBA under sustained AI load, unusable in home office
- 32GB GDDR7 runs hot at full utilization; waterblocks with full VRAM coverage prevent throttling during long inference sessions
- Dual-5090 AI workstation (1150W GPU power) is only physically buildable with water cooling — air coolers are too thick

Full-cover water blocks cool GPU die + VRAM + VRM simultaneously — unlike air coolers that only target the GPU die.