GeForce RTX 5090 Build
The last time I completed a major upgrade of my primary PC was in 2021, when I switched to an AMD 3950X from an Intel Core i7-6700K.
I documented the build process and subsequent upgrades in the following four articles.
The full specification of my current (soon-to-be old) PC can be found below:
- MSI MAG X570 Tomahawk WiFi
- AMD Ryzen 9 3950X 3.5GHz Base / 4.7GHz Boost (16C/32T)
- Noctua NH-D15 CPU Cooler
- 64GB Corsair Vengeance LPX DDR4 PC4-28800C18 3600MHz RAM
- 1TB Samsung 980 Pro M.2 PCI-e 4.0 NVM-e SSD
- 1TB Samsung PM981 M.2 PCI-e 3.0 NVM-e SSD
- NVIDIA GeForce RTX 3090 FE 24GB GDDR6X
- EVGA SuperNova P2 1000W ‘80 Plus Platinum’ PSU
- Fractal Torrent Compact
This PC has served me extremely well over the past four years, delivering great performance across a wide range of workloads.
The use of high-performance parts also allowed me to skip several generational upgrades, specifically the AMD Ryzen 5000 and 7000 series (Zen 4), as well as the NVIDIA GeForce RTX 40 series.
However, at CES 2025, NVIDIA announced the highly anticipated GeForce RTX 50 series of consumer graphics cards. Specifically, the GeForce RTX 5090, based on the NVIDIA Blackwell architecture.
Although not yet available the specification of the GeForce RTX 5090 guarantees a significant performance increase over my GeForce RTX 3090. This is thanks to the increase in CUDA cores, high clock speeds, and faster memory, as well as the introduction of new features such as 4th Generation RT cores, 5th Generation Tensor cores and DLSS4.
The combination of improved specification and new features, was enough to spark my interest, resulting in the decision to upgrade my entire system.
This article outlines the details of the upgrade, including the rationale behind the decisions.
Working in IT, I utilise my PC for a wide range of different workloads, including productivity, collaboration, software development, local artificial intelligence, photo editing, video editing, video effects, virtual labs, gaming and game development.
Similar to my last build, these requirements demand high-performance components and my goal is to design a balanced system that will perform well in all scenarios.
As a result, I have selected the following hardware. Any hardware not listed (e.g., case) will be reused from my current build.
- MSI MAG X870 Tomahawk WiFi
- AMD Ryzen 9 9950X 4.3GHz Base / 5.7GHz Boost (16C/32T)
- Noctua NH-D15 G2 LBC CPU Cooler
- 64GB Corsair Vengeance EXPO DDR5 PC5-48000C30 6000MHz RAM
- 2TB Crucial T705 M.2 PCI-e Gen5 NVM-e SSD
- Corsair HX1200i 1200W ‘80 Plus Platinum’ ATX PSU
- NVIDIA GeForce RTX 5090 FE 32GB GDDR7
Motherboard
My previous MSI MAG X570 Tomahawk WiFi had served me well, delivering reliability and high performance, even when overclocked. I was also impressed by the price/feature ratio, build quality, and BIOS support, which all compare favourably to the competition (e.g., Asus, etc.).
The latest chipsets from AMD are the X870 and X870E. On paper, these chipsets are a little underwhelming, as they are iterations of the previous B650E and X670 chipsets, with the main difference being the addition of USB 4 (40Gb/s). With that said, iteration is not always bad, hopefully allowing the AMD product team to focus on refinement.
Unfortunately, since I last purchased a motherboard, the prices have increased significantly, with X870 motherboards ranging from £200 to £600.
I selected the MSI MAG X870 Tomahawk WiFi, which I believe delivers the best price/feature ratio.
The MSI MAG X870 Tomahawk WiFi is an X870 chipset, not an X870E. The X870E is a dual-chip architecture (incorporating two Promontory21 chipsets), which delivers additional PCI-e lanes and USB I/O. However, the dual-chip architecture also adds complexity, heat and cost. Therefore, unless you have a specific purpose for the PCI-e lanes and/or USB I/O, the X870E is likely not worth the additional investment.
Outlined below are the key features of the MSI MAG X870 Tomahawk WiFi motherboard.
- AMD X870 Chipset
- 14+2+1 Duet Rail Power System (14-phase, 80A Power Stages)
- AM5 Socket
- 4x DDR5 UDIMM
- 1x PCIe Gen5 x16 (CPU)
- 1x PCIe Gen4 x4 (Chipset)
- 1x PCIe Gen3 x1 (Chipset)
- 2x M.2 Gen5 x4 (CPU)
- 1x M.2 Gen4 x4 (Chipset)
- 1x M.2 Gen4 x2 (Chipset)
- 4x SATA (6Gb/s)
- 2x USB 4 (40Gb/s)
- 1x USB 3.2 (20Gb/s)
- 3x USB 3.2 (10Gb/s)
- 7x USB 3.2 (5Gb/s)
- 8x USB 2.0
- 1x 5Gb/s LAN
- WiFi 7
- Bluetooth 5.4
Needless to say, that is a lot of connectivity, which can be confusing. Therefore, I would recommend reviewing the motherboard manual, specifically details regarding the bifurcation of the PCI-e lanes, which can limit/disable certain features based on the hardware installed (e.g., USB 4).
To see how the MSI MAG X870 Tomahawk WiFi performs against the competition, I recommend reviewing the TechSpot X870/X870E Motherboard Roundup.
The image below from TechSpot highlights the testing results of the MSI MAG X870 Tomahawk WiFi. As stated in their article, “performance-wise, the Tomahawk excelled, averaging a clock speed of 5,110 MHz for a score of 2,306 points, one of the highest scores in our testing.”
The image below shows the integrated rear I/O panel of the MSI MAG X870 Tomahawk WiFi. Note, the accessible Clear/Flash CMOS buttons, which can be very useful when tuning or overclocking a system.
At £279.98 (Scan) the MSI MAG X870 Tomahawk WiFi is not the cheapest X870 motherboard but based on my experience with MSI, will deliver a high quality, reliable foundation, with critically reviewed features and performance that rivals more expensive motherboards from Asus, Gigabyte, etc.
Processor and Cooler
If I were building a PC specifically for gaming, I would select the new AMD Ryzen 9800X3D, which thanks to the second-generation AMD 3D V-Cache, delivers best-in-class gaming performance.
With that said, the Ryzen 9800X3D is an eight-core, sixteen-thread processor, with a maximum boost clock of 5.2 GHz. Therefore, although the Ryzen 9800X3D is the gaming champion, it falls behind in productivity or specialist tasks (e.g., development, video editing) compared to processors with a higher core count and/or frequency.
As a result, I have selected the AMD Ryzen 9 9950X, which is a sixteen-core, thirty-two-thread processor, with a maximum boost clock of 5.7 GHz. Although the AMD Ryzen 9 9950X received mixed critical reviews, this was mostly driven by inflated expectations associated with the Zen 5 architecture, which many had unrealistically hoped would deliver significant (> 30%) performance benefits over Zen 4.
In my case, the AMD Ryzen 9 9950X delivers a massive performance increase over my current AMD Ryzen 9 3950X, which is based on the Zen 3 architecture. In addition, it successfully balances performance across a wide range of workloads, unlike the AMD Ryzen 9800X3D, which would compromise productivity.
Arguably, I should wait for the recently announced AMD Ryzen 9950X3D to be released. However, I am not convinced by the benefits of 3D V-Cache for processors with dual Core Chiplet Dies (CCDs), specifically the AMD Ryzen 9900X/9950X.
These processors only have a 3D V-Cache on one die, which requires the software (usually Microsoft Windows) to accurately schedule tasks to see the desired performance benefits.
This additional complexity has been proven to cause issues, requiring more “human effort” to maximise performance across workloads. In addition, 3D V-Cache commands a price premium of approximately £200, making it a significant investment for questionable value.
Finally, following recent discounts, the AMD Ryzen 9 9950X is currently the same price as the AMD Ryzen 9800X3D (approximately £529.99), making it cost-competitive.
To manage the thermals of the AMD Ryzen 9 9950X, I have selected the Noctua NH-D15 G2 LBC CPU Cooler.
This will be my third Noctua NH-D15 and I have always been impressed with the build quality and thermal results. The recent release of the NH-D15 G2 in theory delivers even better results, although I suspect the real-world difference will be minimal.
Either way, outside of custom water cooling (which I have no desire to implement), I continue to put my trust in Noctua.
I fully expect to be able to leverage AMD Precision Boost Overdrive (PBO) and Curve Optimiser to deliver a modest 24x7 overclock, whilst still maintaining acceptable thermals and noise.
Memory
I do not have extreme overclocking expectations. However, I do plan to leverage AMD Extended Profiles for Overclocking (EXPO). To ensure system stability, AMD recommends a maximum of two DIMMs.
As a result, I have selected 64GB (2x32GB) Corsair Vengeance DD5 PC5-48000C30 RAM, which is EXPO (AMD) configured, operating at 6000MHz with 30-36-36-76 timings.
This is not the highest frequency memory, but my research indicates that 6000MHz/6400MHz with tight timings (< 32 CAS Latency) is the sweet spot for optimal performance with AMD Ryzen 9000 series processors.
Higher frequency memory commonly comes with less aggressive timings, which can offset the performance benefits. It also places greater stress on the system, which can impact stability or require a lot of additional tuning.
64GB offers decent headroom, but I would have preferred 96GB (2x48GB). Specialist workloads such as development, local artificial intelligence, and virtual labs can easily consume 60GB+.
Unfortunately, 96GB options are currently limited and costly. Thankfully, memory is the easiest component to upgrade. Therefore, I will keep an eye on the market and potentially upgrade in the future.
Storage
The MSI MAG X570 Tomahawk WiFi has two M.2 PCI-e Gen5 x4 slots, both of which are connected to the processor. I have decided to occupy one slot with the Crucial T705 M.2 PCI-e Gen5 NVM-e SSD.
The Crucial T705 is one of the highest-performing consumer-grade drives available today, with the 2TB version capable of reaching up to 14,500MB/s for sequential reads and 12,700MB/s for sequential writes, as well as up to 1,550K random read IOPS and 1,800K random write IOPS.
It uses the Phison E26 SSD controller (based on the Max14um platform) and Micron 232-Layer TLC NAND flash, similar to what is found on competing products, such as the Corsair MP700 Pro SE. This combination delivers an impressive bus speed of 2,400MT/s, which unlocks the incredible sequential read performance.
In day-to-day scenarios, the impact of this cutting-edge performance is likely negligible. However, for certain storage-heavy workloads, the additional performance offered by the Crucial T705 will reduce the risk of bottlenecks. The drive is also optimised for Microsoft DirectStorage, making it future-proof for gaming.
Power Supply Unit (PSU)
I have selected the Corsair HX1200i 1200W ‘80 Plus Platinum’ ATX PSU, which is ATX 3.1 and PCI-e 5.1 certified, delivering 1200W via a single rail. It also comes equipped with three EPS12Va connectors and a PCI-e 5.1 12V-2x6 GPU cable.
The NVIDIA GeForce RTX 5090 specification states a 575W Thermal Design Power (TDP), which is up from my current 350W of the NVIDIA GeForce RTX 3090.
This extreme power requirement demands a high-performance, high-quality power supply unit to ensure safe/reliable power delivery.
When calculating my maximum system wattage, I assumed the following worst-case scenario.
- Processor Overclocked (25%)
- Graphics Card Overclocked (10%)
- 2x SSD
- 2x RAM
- 5x Fans
- USB 3.2 Gen2 Power Transmission
These assumptions resulted in a maximum system power requirement of approximately 1060W.
NVIDIA officially recommend a minimum 1000W PSU for the GeForce RTX 5090. However, I would always recommend selecting a PSU with plenty of headroom.
Conclusion
With my research complete, I have procured the core components and now await the official launch of the NVIDIA GeForce RTX 5090.
In the interim, I plan to build and tune the system using my current NVIDIA GeForce RTX 3090.
As with previous builds, I plan to document the process, including my 24x7 settings and performance results.