For high end desktop replacement workstations, technology upgrades are consistent. Every year these machines get the newest, fastest GPUs (CPUs, memory, storage, …) available.
Mid-range Mobile Upgrades
Updates in the mid-range are more challenging. For workstation vendors, performance and power requirements are important.
NVIDIA leads in mobile workstation, professional GPUs for a good reason. NVIDIA has fast, reliable GPUs. More importantly, the company has focused on squeezing performance out of a constrained power budget.
We look at this dynamic below and examine the performance and power of mid-range, professional GPUs for mobile workstations.
Always thinner, lighter, and faster – our mobile workstation dream
Mid-range mobile workstations in 2021
This dynamic became clear when analyzing the latest mobile workstation products from some of the main vendors. The 2021 GPU updates for Dell and for HP went into workstations with identical external designs and identical thermal designs as the 2020 models.
A typical transition went from a 2020 model using Quadro T1000 and Quadro T2000 to the new Quadro T1200 and RTX A2000 GPUs. Look at the GPU characteristics in the table below. The power and performance tend to support this transition as a logical upgrade.
|Quadro T1000||Quadro T1200||Quadro T2000||RTX A2000|
|Release Date||May 27, 2019||April 12, 2021||May 27, 2019||April 12, 2021|
|Process Size||12 nm||12 nm||12 nm||8 nm|
|Transistors||4.7 billion||4.7 billion||4.7 billion||13.25 billion|
|Memory Size||4 GB||4 GB||4 GB||4 GB|
|Memory Bus Width||128 bit||128 bit||128 bit||128 bit|
|Memory Bandwidth||128.1 GB/s||160 GB/s||128.1 GB/s||176 GB/s|
|System Bus Interface||PCIe 3.0||PCIe 3.0||PCIe 3.0||PCIe 4.0|
|Base Clock||1395 MHz||855 MHz||1575 MHz||893 MHz|
|Boost Clock||1455 MHz||1425 MHz||1785 MHz||1358 MHz|
|Memory Clock||2001 MHz||2001 MHz||2001 MHz||1375 MHz|
|TDP||50 W||18 W||60 W||95 W|
Mid-range Mobile Upgrades
The T1000, T1200, and T2000 are all based on NVIDIA’s Turing GPU architecture. All three come from the same GPU model, the TU117 GPU.
The differences between the T1000 and T2000 hold no surprises. The T1000 is the same GPU, but it has fewer cores, TMUs, and SMs than the T2000. The GPU’s core clock speed and boost clock speed are lower than those of the Quadro T2000. The TDP is also 10 W lower than the Quadro T2000’s power.
What might surprise some of us is that the new Quadro T1200 is just a re-branded Quadro T2000. It is the same chip. It has identical quantities of graphics processing units. It is a Quadro 2000 with a couple of notable and important changes.
The first change: a new memory controller to support GDDR6 memories. The second change: lower clock speeds.
These lower clock speeds don’t necessarily mean that the Quadro T1200 has lower performance than the Quadro T2000. Thanks to the GDDR6 memory, the memory bandwidth is higher, and the throughput is faster despite the lower memory clock speed. And while the core clock speed is almost half of the Quadro T2000 clock speed, the boost clock speed is not significantly below the Quadro T2000 boost clock speed.
For a mobile workstation manufacturer, the most important difference is that the TDP of the Quadro T1200 is less than a third of the Quadro T2000’s TDP of 60 W. Taken together, these advantages make the Quadro T1200 a great upgrade to the Quadro T1000.
Quadro T1200 pumps up entry-level mobile workstation performance
An interesting trend is seeing the Quadro T1200 show up in entry-level mobile workstations. This is the case for the Dell Precision 3561 mobile workstation.
In entry-level systems, the power budget is a critical constraint. NVIDIA created the Quadro T1200 GPU with a TDP of 18 W. This TDP makes the GPU attractive for systems which could never integrate a 60 W Quadro T2000.
What does this mean for you? A small mobile workstation will now have the performance of a 60 W Quadro T2000 in an 18 W Quadro T1200 chip.
That is good news for many users.
RTX A2000 versus Quadro T2000
For NVIDIA professional GPUs, the “2000” positioning either means upper levels of the mid-range or lower levels of the high end. Even within their respective GPU architecture generations, you can think of the Quadro T2000 as the former and the RTX A2000 as the latter.
With a TDP 63% higher than the Quadro T2000, replacing a Quadro T2000 with an NVIDIA RTX A2000 may present some challenges, but it is worth the effort. The RTX A2000 adds RT cores and tensor cores. These cores enable real-time raytracing and AI.
The important difference is the GPU architecture. Yes, the Ampere architecture more powerful with new incarnations of CUDA cores, Tensor cores, and RT cores. It is also more efficient. NVIDIA claims that the Ampere GPUs “are up to 1.9x more power efficient than Turing GPUs”.
One workstation manufacturer shows professional graphics performance 40% higher and computing performance 166% higher with the RTX A2000 versus the Quadro T2000. Several workstation models replace the Quadro T2000 with the RTX A2000 despite its 95 W TDP specification.
NVIDIA claims that the Ampere GPUs “are up to 1.9x more power efficient than Turing GPUs”
A Final Perspective
Workstation manufacturers must squeeze as much power into a constrained space and power budget. The 2021 released GPUs, Quadro T1200 and RTX A2000, support this goal in different ways.
The Quadro T1200 delivers the performance of a Quadro T2000 with a miniscule power requirement. The RTX A2000 uses more power but delivers new hardware cores and much faster performance.
• SM: Streaming Multiprocessors
• ROP: Raster Operators
• TMU: Texture Management Unit
• TDP: Total Dissipated Power
• NVIDIA Ampere Architecture Whitepaper
The Precision 5760 supports NVIDIA RTX A2000 and RTX A3000. Take three minutes to watch our video for a Quick Look at the features.
- Find the Dell Precision 5760's main specifications and features in our article: A Quick Look Inside Dell’s Precision 5760
- Check out the Dell 5760's performance results in our article: Checking Dell's Precision 5760 Performance.
- And find out Why I love the Precision 5760 mobile workstation.