nvidia NVIDIA quietly announces a GT 1010 that's apparently faster than a GT 1030 DDR4

[Ash_Kechummm]

Summary

 

Tech youtuber Dapz made a youtube video (which was subsequently reported upon by notebookcheck.net) where he said someone he knew found a "GT 1010" mentioned in the NVIDIA driver download page. When he contacted NVIDIA support, they confirmed its existence, but refused to comment on specifications or a release date.

However, it was soon added to TechPowerUp's database, which had the following info:

Quotes

Quote

 The GP108 has a die area of 74 mm² and is comprised of 1,800 million transistors. Essentially, the GT 1010 is the same GPU as the GT 1030 with some CUDA cores disabled. 

 

Quote

...The GT 1010 supports 2 GB of GDDR5 VRAM on a 64-bit memory interface. The base clock of the GPU is 1,228 MHz and can boost up to 1,468 MHz. Additionally,  the card features 16 texture mapping units (TMUs), 16 raster operation pipelines (ROPs), and a TDP of 55 W...

My thoughts

With these specs, it looks to me like it might outperform the dreadful GT1030 DDR4 at a potentially lower price, which would be hilarious, to say the least (though NVIDIA hasn't said anything about pricing yet, so take my words with a mountain of salt)

Also I hadn't expected NVIDIA to release a new GT card with the Pascal architecture (I expected something like a 1630????), so it's interesting to see what else NVIDIA might do in the lower end (being a low-end enthusiast myself)

Sources

https://www.notebookcheck.net/NVIDIA-quietly-introduces-the-GeForce-GT-1010-A-Pascal-GP108-GPU-with-256-CUDA-cores-2-GB-GDDR5-VRAM-and-55-W-TDP.515158.0.html

https://www.techpowerup.com/gpu-specs/geforce-gt-1010.c3762

Edited January 17, 2021 by Ash_Kechummm
spelling and grammar lol

[.Apex.]

So this is supposedly a replacement for the GT710? if it performs the same as a GT1030 then that's not bad at all

[Ash_Kechummm]

3 minutes ago, .Apex. said:

So this is supposedly a replacement for the GT710? if it performs the same as a GT1030 then that's not bad at all

it does look like a replacement for the GT 710

since it has 256 CUDA cores and 16 TMUs instead of 384 CUDA cores and 24 TMUs of the GT 1030, it'll be slower than the real GT 1030 (the GDDR5 version), but since the DDR4 one has literally half the performance, the 1010 (with its GDDR5) could outperform the DDR4 GT 1030.

[Spotty]

I was not expecting Nvidia to release a new Pascal card in 2021.

 

Maybe if we pretend like this is a really good graphics card all the scalpers will rush out to buy it, then get stuck with a bunch they can't sell.

Wow, the GT 1010!? I've been waiting for this for years! This is going to be the best value graphics card on the market! Hopefully I will be able to buy one on launch!

[PlayStation 2]

If it's cheap enough, I could find a genuine use for a GT 1010.

[Taf the Ghost]

53 minutes ago, Spotty said:

I was not expecting Nvidia to release a new Pascal card in 2021.

 

Maybe if we pretend like this is a really good graphics card all the scalpers will rush out to buy it, then get stuck with a bunch they can't sell.

Wow, the GT 1010!? I've been waiting for this for years! This is going to be the best value graphics card on the market! Hopefully I will be able to buy one on launch!

We're at the point they'd replace the GT 710. It's every 2-3 gen both companies refresh that "makes monitors turn on" spec. It's a super small die that costs them like $3 per chip to produce. They really are likely to produce >25 million of them over the lifetime. Maybe more. They're long tail products. AMD is more notable for doing this with their APUs for ultra-cheap Chromebooks. I think the last Cat Core based one launched in early 2020, with AMD moving to a Zen based version in probably 2022. 

[Dabombinable]

My GTX 650ti is a lot faster than the DDR4 GT 1030...so that is an extremely low bar.

[Ash_Kechummm]

5 minutes ago, Dabombinable said:

My GTX 650ti is a lot faster than the DDR4 GT 1030...so that is an extremely low bar.

That is true, but if the naming is anything to go by, a 1010 is below a 1030;

If that order is shattered in real-world performance, it would be very confusing for customers going on the assumption that higher numbers mean better performance.

 

[Dabombinable]

1 minute ago, Ash_Kechummm said:

That is true, but if the naming is anything to go by, a 1010 is below a 1030;

If that order is shattered in real-world performance, it would be very confusing for customers going on the assumption that higher numbers mean better performance.

 

Nvidia is known for not following "higher number=better". I've got an FX 5600XT for example, that is actually slower than both a standard FX 5500 and FX 5200 (kind of got ripped on that one). Then they more recently haven't differentiated properly between different products (they used to use VE, LE, XT etc for segmenting low end parts). The GT 1010 being faster than a variant of the GT 1030 is quite likely.

[Drama Lama]

Would be cool to have a cheap 4x HDMI 2.1 video output but they probably won’t do that in order to sell overpriced quadro cards

[Spotty]

43 minutes ago, WikiForce said:

it's extremely weird to see it being twice as power hungry (atleast according to the TDP rating) despite being a slower cut down version of 1030, Nvidia is really doing some weird shit.

The 55W figure provided in the article is likely wrong. There's no officially released specifications for the card yet. The TechPowerUp page linked to says 30W, contradicting the article.

 

[The1Dickens]

I was reading this last night, and I am wondering if it might be useful as a dedicated stream-encoder.

 

At the launch of the 30 series, I was wondering what happens to all the "bad" chips (or, rather how bad are they) and if the NVENC part was salvageable, could be used for that. This way they can possibly make a bit of money on them, instead of total loss.

 

Anyway, would be interesting to see how these are marketed.

[SAVE-12-HK]

9 hours ago, Ash_Kechummm said:

Essentially, the GT 1010 is the same GPU as the GT 1030 with some CUDA cores disabled. 

how is it possible 1010 is faster?

[WereCatf]

24 minutes ago, The1Dickens said:

I was reading this last night, and I am wondering if it might be useful as a dedicated stream-encoder.

Nope, only GTX- and RTX-series cards have NVENC, not the GT-series.

[WereCatf]

2 minutes ago, SAVE-12-HK said:

how is it possible 1010 is faster?

It's worded poorly. What they mean is that the 1010 would be faster than the DDR4-version of 1030, not the faster, GDDR5-version of 1030.

[minibois]

Quote

a GT 1010 that's apparently faster than a GT 1030 DDR4

People who have learned "higher number = higher better":

Spoiler

 

Seeing how good GT 710 cards still sell, I am sure there is some money in it for Nvidia. If this is going to be released to see budget gaming PC channels cover it, it's fine by me.

[The1Dickens]

11 minutes ago, WereCatf said:

Nope, only GTX- and RTX-series cards have NVENC, not the GT-series.

Ah. I misinterpreted the "streaming processors" part...

 

Thanks for the clarification.

[PianoPlayer88Key]

5 hours ago, Ash_Kechummm said:

... assumption that higher numbers mean better performance.

 

If I had my way, higher numbers WOULD mean better performance.  The mobile version of a hypothetical GT 4010 (that would go into something like a tablet, phone, watch, something with extremely limited space/cooling/TDP should be faster than the RTX 3090 or the flagship Quadro/whatever-Tesla-is-now-called/workstation/server/datacenter Ampere.

 

Same goes for CPUs: slowest ULV mobile hypothetical next-gen Alder Lake or Zen 4 smartwatch CPU should be faster than a max-#-of-sockets-in-1-motherboard of the fastest previous-gen Rocket/whatever-lake-they're-called or Zen 3 flagship server CPU.

/s 

 

 

3 hours ago, WikiForce said:

it's extremely weird to see it being twice as power hungry (atleast according to the TDP rating) despite being a slower cut down version of 1030, Nvidia is really doing some weird shit.

 

2 hours ago, Spotty said:

The 55W figure provided in the article is likely wrong. There's no officially released specifications for the card yet. The TechPowerUp page linked to says 30W, contradicting the article.

 

I would have liked to see it be a lower TDP than the 1030.  I hope it's about 20 watts or so, although I would have loved to see a card with, say, 5-10 watt TDP or less that would not need a heatsink, and still be good for some 720p gaming at 30fps low/medium in modern titles. 

 

My use for one would be not gaming though, but in a NAS/backup server that has no APU and no on-board video out port (or only VGA... BTW when will server boards start having HDMI or DisplayPort? None of the monitors I have support anything else, and I'm not buying another monitor to get VGA support.)

 

[Spotty]

Just now, PianoPlayer88Key said:

I would have liked to see it be a lower TDP than the 1030.  I hope it's about 20 watts or so, although I would have loved to see a card with, say, 5-10 watt TDP or less that would not need a heatsink, and still be good for some 720p gaming at 30fps low/medium in modern titles. 

 

My use for one would be not gaming though, but in a NAS/backup server that has no APU and no on-board video out port (or only VGA... BTW when will server boards start having HDMI or DisplayPort? None of the monitors I have support anything else, and I'm not buying another monitor to get VGA support.)

I use a fanless GT 710 in my NAS for that reason. It's just using an old first gen Intel CPU that lacks integrated graphics and the motherboard refused to boot without a dGPU. The GPU does absolutely nothing other than allow the motherboard to boot since it doesn't even have a monitor plugged in to it.

[StDragon]

14 minutes ago, The1Dickens said:

Ah. I misinterpreted the "streaming processors" part...

 

Thanks for the clarification.

https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new

 

They're still good for decoding H264 and H265 FWIW.

[The1Dickens]

9 minutes ago, StDragon said:

https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new

 

They're still good for decoding H264 and H265 FWIW.

Right, but I was thinking more along the lines of, say, a fledgling Twitch streamer or something.

[PianoPlayer88Key]

47 minutes ago, Spotty said:

I use a fanless GT 710 in my NAS for that reason. It's just using an old first gen Intel CPU that lacks integrated graphics and the motherboard refused to boot without a dGPU. The GPU does absolutely nothing other than allow the motherboard to boot since it doesn't even have a monitor plugged in to it.

Yeah... A while ago when I was looking at building a NAS (haven't done it yet, it's on pause now), I had been considering the Zotac PCIe 1x variant of the 710 with the ASRock E3V5 WS board which didn't have iGPU support (I have an i3-6100, orphaned since my laptop was upgraded to an i7-6700K) to put in a 1x slot to leave the 16x slots open for HBAs like the LSI 9211-8i, 9200-16e (which I've seen pretty cheap on eBay) or if I could afford it, the Highpoint Rocket 750 (saw those fairly cheap on eBay a while back - more expensive per card than the 9211-8i but cheaper per port IIRC).

Another option was something like an LGA 775/1366/2011 board, which (except for the 775) have 7 PCIe x16 slots for the ones I was looking at (Supermicro X7DWE/X8DTH-6F/X9DRH-iF or similar).

But, I'd want to keep all the PCIe 8x/16x slots open for HBAs, I can't find any standard / non-rackmount cases that support that many HDDs. (Each Rocket 750 supports 40 SATA devices, * 7 slots = 280 drives -- FD Define 7 XL / Meshify 2 XL doesn't have that many 3.5" mounts.)

Also my budget is such that a single 12/14/16TB HDD would be more expensive than the entire rest of the setup combined. And other things going on mean that project is on pause.  (I do have a few month old backup though, via a few cloned HDDs & SSDs sitting on a shelf, but I need to find a way to keep it updated.  I have some ideas for what I want to accomplish but this is the wrong subforum for that.)

 

Edit: okay this is a bit annoying/odd/interesting/... this looks like a "wall of text" on Chrome on my Pixel 3a, but little more than a blip on the screen in Firefox (40% zoom level) on my ASUS VG289Q (Windows at 100% scaling).

Also how do I get LTT forums (& many other sites) to take the entire 3840-pixel width of my monitor by reflowing content for fewer line breaks, not zooming way in?)

Spoiler

 

 

 

OH.... back to gaming & graphics cards (while incorporating server hardware) .... I was thinking of something.... Someone (doesn't have to be Linus but I think his concept of a video doing it might be interesting) should try a project like getting one of those 11-slot server boards (or one of the Supermicro rackmount servers that, iirc, one variant supports 39 expansion slots or at least it did a few years ago),

fill all those slots with a variant of the RTX 3090 that has 8 DisplayPorts (I know GPUs exist, like the NVS 810, that have 8 MiniDP ports but I haven't heard of an RTX 3090 variant with that yet),

plug 8K monitors (like the Dell UP3218K, unless there's high refresh 8K monitors available yet? I'd heard a few years ago that someone had demoed something like an 8K 120Hz panel at some trade show or something, I don't remember details though) into every single port,

and benchmark Cyberpunk 2077 at max settings (including RTX), with DLSS off - gaming across ALL those monitors. (Kind-of like some of those walls of dozens of monitors or something like that.)

 

I wonder what GPU you'd actually NEED to run CP'77 (or Flight Simulator 2020, or whatever is the most-demanding game when such GPU comes out if it doesn't exist today) at (A) that resolution, AND better yet, (B) the max FPS supported by the game engine?

(A) If you went with the 39-slot Supermicro 7089P-TR4T server, that'd be 8K (7680x4320) * 8 monitors/card * 39 cards... 10,351,411,200 total pixels, or 168,480 (4320*39) x 61,440 (7680*8), for ~1:2.74 aspect ratio, if I calculated right.

(B) I've seen videos of DOOM reaching 1,000 fps, and I've had Fortress Forever (fan-made source engine version of Team Fortress Classic, made around the same time as TF2 but TF2 gets much lower fps for some reason) hit 900-1200 or so FPS at semi-low resolution & settings.  I feel like it could go higher if my 6700K wasn't limiting it in some way - IIRC I can get that FPS at like 800x600 medium settings or so.  (Also HL1 / TF Classic normally maxes out at 100 fps, but someone told me a way that I've since forgotten to go past that cap, but it messes things up when you go to far.)

[williamcll]

Maybe it's a successor to the NVS series of cards too?

[Kisai]

11 hours ago, Spotty said:

I was not expecting Nvidia to release a new Pascal card in 2021.

 

Maybe if we pretend like this is a really good graphics card all the scalpers will rush out to buy it, then get stuck with a bunch they can't sell.

Wow, the GT 1010!? I've been waiting for this for years! This is going to be the best value graphics card on the market! Hopefully I will be able to buy one on launch!

These cards are entirely pointless to have. The UHD 630 iGPU is nearly identical in performance to the 1030, and nvidia wants to make an even less powerful one? Who's the target for this? Potato boxes for SEA?

 

 

[dapz]

9 minutes ago, Kisai said:

 The UHD 630 iGPU is nearly identical in performance to the 1030, 

 

 

UHD 630 should pretty similar to the 1010 performance wise assuming the 256 cuda core number is correct