For the lowest-end video cards, a common practice by manufacturers is to substitute in slower VRAM – recently, this meant swapping out GDDR5 for DDR3 – saving costs in a product that typically would not be able to utilize the extra memory bandwidth. For quite some time now, we have been used to seeing DDR3 in those ultra low-end cards – until now. Sometime last month, NVIDIA quietly released the GeForce GT 1030 DDR4, with several models filtering out through add-in board partners.

In addition to slower-clocked DDR4 VRAM, the GT 1030 DDR4’s core clocks have also been reduced, though the exact boost clocks differ between custom boards. The end result of 2100Mbps DDR4 from 6008Mbps GDDR5 does reduce the bandwidth from around 48 GB/s to 16.8 GB/s, though with no change to memory size and bus width. It appears that the DDR4 variants are also lower power, with a 20W TDP as opposed to 30W.

Specifications of Selected NVIDIA GeForce GT 1030 Cards
  Palit GT 1030 MSI GT 1030 Gigabyte Low Profile GT 1030
Base Clock 1227 MHz 1151 MHz 1265 MHz 1189 MHz 1227 MHz (Gaming) 1151 MHz (Gaming)
1252 MHz (OC) 1177 MHz (OC)
Boost Clock 1468 MHz 1379 MHz 1518 MHz 1430 MHz 1468 MHz 
1379 MHz 
1506 MHz (OC) 1417 MHz (OC)
VRAM Clocks 6000 Mbps 2100 Mbps 6008 Mbps 2100 Mbps 6008 Mbps 2100 Mbps
Capacity 2 GB
Bus Width 64-bit
Power Consumption 30W 20W 30W 20W unspecified

Interestingly, some of the listed documentation shows two different core names: “GP108-300” and “GP108-310.” But while “GP108” is affiliated with the lower-clocked, lower TDP DDR4 variant, certain GDDR5 models (AERO ITX 2G OC, AERO ITX 2G OCV1, 2G LP OC, 2G LP OCV1, 2G LP OCV2) list both core names in their specification sheets but with everything else unchanged, leaving it unclear what the differences are between these two GP108 bins.

The official NVIDIA GeForce GT 1030 product page mentions nothing of the change, though it does note that specifications of partner board may vary. In any case, each of the model names bar one differentiate the peculiarity with a “D4” moniker. Though the consequence of the bandwidth difference is much less with such low-end cards, particularly with ones under a tighter power budget, it is unclear how much of a performance hit this entails.

Nevertheless, given the overall shift away from DDR3 production, we should soon see future lower-end cards equipped with DDR4 over DDR3.

Source: MSI (via SH SOTN)

Comments Locked


View All Comments

  • eastcoast_pete - Wednesday, April 4, 2018 - link

    Thanks, that's helpful! Fingers crossed that these DDR4 economy versions will do so too.

    I also agree with other posts here that they could have used at least mainstream-fast DDR4 in these new cards, not the bottom-of-the-barrel slow chips.
  • ToTTenTranz - Wednesday, April 4, 2018 - link

    That link isn't from nvidia, it's from MSI..
  • Zero11s - Wednesday, April 4, 2018 - link

    good for a cheaper HEVC upgrade for PC without iGPU
  • Ej24 - Wednesday, April 4, 2018 - link

    Good god... It's like the 27 flavors of Gt710 all over again. That old thing ranges from 96 Cuda core to what 296? Could be Fermi or Kepler. 64bit to 128bit bus. Ddr3 to gddr5 memory. No clear delineation between the models.
  • MrCommunistGen - Wednesday, April 4, 2018 - link

    Without doing any searching to back up my suspicions... I don't think Mbps is the correct unit for memory speeds in the article or table. Maybe MT/s?

    This doesn't really change the meaning behind the article (YAY! There's new slower variants of the GT 1030 available to confuse the ill-informed with now!) but if the units on the specification in question are wrong it'd be nice if they could be corrected.
  • Ryan Smith - Wednesday, April 4, 2018 - link

    This matter comes up in GPU-related articles now and then. The GPU and memory manufacturers have told us that they'd prefer we use bps instead of MHz or MT/s. It's somewhat arbitrary, I agree, but it does avoid some unnecessary confusion about frequency versus bandwidth.
  • spaceship9876 - Wednesday, April 4, 2018 - link

    I want Nvidia to roll out a $40 GT1010 gpu for htpc usage as their GT710 uses the ancient fermi architecture. Both Nvidia and AMD have abandoned this market segment. Some people buy ryzen cpus and just want a gpu that can hardware decode the latest video codecs and accelerate applications using latest opencl. Instead they are forced to pay ~$80 for the latest gen lowest cost gpu.
  • mode_13h - Wednesday, April 4, 2018 - link

    The market for new GPUs that are no faster than an iGPU is too small. I really wouldn't expect to see anything slower than a GTX 1030 or a RX 550.

    And for HTPC, I see 3 different passively-cooled GTX 1030's on newegg.
  • IBM760XL - Thursday, April 5, 2018 - link

    That would be nice. Reminds me of how a couple weeks ago I saw someone buying a Radeon HD 6450 at Micro Center - a card that will be 7 years old next month. Don't know if they were replacing an old part or just needed a display on CPU, but it's the most powerful CPU in stock for $50 (though there's a $40 710 that's sometimes in stock).

    At least they weren't forced to go with the 8-year-old HD 5450, or *shudder* the almost-9-year-old 1 GB 8400 GS.
  • bennyg - Friday, April 6, 2018 - link

    Years ago I tried a passive 5450 for htpc duties. Quickly ditched it as it couldn't handle even mild bitrate 1080p content without dropping frames.

Log in

Don't have an account? Sign up now