Nvidia fully freshens its Quadro lineup with Pascal potency

It's been a little over six months since Nvidia launched the Quadro P6000 and P5000. Those Pascal-powered pro graphics cards were built with the GP102 and GP104 processors, and the Quadro GP100 should now satisfy pros with a need for double-precision speed. For professionals with a less-pressing need for high-precision pixel-pushing, Nvidia is also launching a full Pascal refresh of its mid-range and lower-end Quadro products. Say hello to the Quadro P4000, P2000, P1000, P600, and P400.

Starting at the top, it seems the P4000 is based on a GP104 chip like its sibling, the Quadro P5000. However, the Quadro P4000 is further shaved-down to 1792 shader cores. AnandTech says it retains the 256-bit path to its 8GB of memory from the Quadro P5000, though. The biggest change from the P5000 to the P4000 is its single-slot form factor. The Quadro P4000 also comes with a 15W-lower TDP than the card it purportedly replaces, the Quadro M4000. That's despite offering almost exactly double the performance on paper: 5.3 FP32 TFLOPS versus the 2.66 TFLOPS of the Maxwell card.

The next step down is the Quadro P2000. This card is built off a GP106 processor like the one used in the 3GB GeForce GTX 1060, although once again, it's further cut down compared to that card. This single-slot GP106 spin has 1024 shader cores enabled. The card also takes a hit to its memory configuration and ends up with 5GB of memory connected to a 160-bit bus. Likely thanks to those cuts, the card ends up with a TDP of just 75W. Its predecessor—the 1.8-TFLOP Quadro M2000—didn't, and this card has the same TDP. Performance on the P2000 is up by 66% over last generation, however.

Moving further down the range, the Quadro P1000 is built on the GP107 GPU used in the GeForce GTX 1050. It has that processor's same 640 shader cores and 128-bit memory bus, although it has twice the memory at 4GB. Its 47W TDP is actually slightly higher than the Quadro K1200's 45W, but the peak compute performance has nearly doubled from 0.98 TFLOPS to 1.8 TFLOPS. This card is a low-profile design barely longer than the PCI Express x16 slot it sits in, which means it should fit almost anywhere.

The Quadro P600 and Quadro P400 form the foundation of the Pascal-refreshed lineup. These cards use even further cut-down GP107 chips. The P600 comes with 384 shaders enabled, and the P400 with 256. That leaves the P600 with 1.1 TFLOPS of peak FP32 performance and the P400 with 0.6 FP32 TFLOPS, but peak compute performance isn't the point of these cards. Like all of its larger siblings, the P600 comes with four mini-Displayport connections. Meanwhile, the P400 offers three display outs. These cards come equipped with GDDR5 memory, a first for the bottom of the Quadro range.

Nvidia's site offers precious little information about the new Quadros, so we don't have any solid information about pricing. We would expect them to slot in roughly the same places as the Maxwell-based cards they're replacing, though. The company says these cards will be available in March.

Comments closed
    • NTMBK
    • 3 years ago

    These Quadros put most gaming cards to shame. Instead of lashing a comically oversized E-peen enhancing cooler to a low-end card, they actually provide appropriate cooling solutions, and let you fit great performance into a small form factor.

      • ImSpartacus
      • 3 years ago

      They are also cut down and underclocked compared to their gaming brethren.

      For example, that single slot P4000 does have a GP104, but it’s got fewer shaders and lower clocks than a 1070 (let alone a 1080). You don’t get to 105W by sheer will alone.

      And if you’re beyond ~100W, then single slot simply isn’t feasible for an ATX-style desktop machine.

      But even for those cards where it is thermally feasible, it rarely makes sense since effectively all motherboards and cases are built to accommodate dual slot AIBs.

        • NTMBK
        • 3 years ago

        Sure, you need to limit performance to get to those smaller card sizes. But you see ridiculous dual-fan double slot coolers even on a 75W card like the 1050ti, a much less powerful chip than the P4000.

        • Flying Fox
        • 3 years ago

        Product binning at its best. Granted it’s more environmentally friendly to recycle as much partial dies (compared to the gaming chips) as possible.

          • RAGEPRO
          • 3 years ago

          Well, “binning” would be carefully selecting fully-enabled dies that simply run with less voltage. That’s not what’s happening here — Nvidia is simply selling cut-down GPUs.

            • derFunkenstein
            • 3 years ago

            And most likely they’re selling these further-cut-down GPUs for more than the consumer chips, since these have fancy certified drivers.

      • tipoo
      • 3 years ago

      Yeah, I also love the design of the Radeon pros, while the consumer ones all look pukey.


Pin It on Pinterest

Share This