NVIDIA RTX A2000

[DeeJay1001]

Good for you mate! Rather silly attempt at rubbing my nose in it, but the reality is this: The vast majority of people won't be doing that. And BTW, 40MH/s is going to earn you about $0.95(if you're lucky) per day which means your ROI on that purchase will not break even for nearly two years. Yay for you...

it's a drop in the bucket, you had nothing to do with the purchase by the way, dont flatter yourself too much. I've been looking for one of these for a long while.

 

[Nater]

I walked into Microcenter this afternoon as any normal person and walked out with an A2000, it's currently mining away @ 40 MH/s. If these were in stock, anyone could buy them from a plethora of stores, it is a consumer card. The whole "professional use" stigma is simply to justify the cost.

I'm so disappointed my "local" (2 hr drive) MicroCenter didn't get any this time apparently. I wonder if they just didn't update the website, or somebody snagged them all day one.

 

[TheHughMan]

It's a workstation card yet TPU only benchmark games on it.

 

[wolf]

It's a workstation card yet TPU only benchmark games on it.

W1zzard was petitioned to review this from a gaming perspective, as it's the fastest low profile GPU on the market and appealing to SFF builders.

 

[LowProfileDegenBuild]

I hope to get one of these for my HP Compaq 8200 Elite SFF or the passive cooled 1650 currently on a ZOTAC International GT 1030 Fan version 2GB GDDR5

 

[Haile Selassie]

This is the same approach as MaxQ mobile graphics card. Those dies also run at about 0.7-0.8V and fit in a 75-110W power budget. Nothing new to see here.

 

[lexluthermiester]

W1zzard was petitioned to review this from a gaming perspective, as it's the fastest low profile GPU on the market and appealing to SFF builders.

This. From this perspective, the card does well.

 

[dyonoctis]

I don't know, I think its more of a case with just the cards themselves being restricted as to who can buy them. My RTX A6000 that I use for iray rendering is a Lenovo, and those cards are usually reserved for government/research/enterprise workstations, so it can be much harder(and expensive) to get those professional workstation cards. The few A6000's that I was able to find a few months ago were crappy PNY brand. So I don't think cripple currency has as much of an impact when it comes to these particular cards. Gaming cards are a different story since they don't have the same restrictions and aren't being prioritized to businesses/enterprise systems.

Isn't PNY the only authorized manufacturer for Nvidia professional GPU ?

 

[Jolly]

I've been waiting for 3 months for one of these from shop BLT. Also would love to know more about how it compares to the w6400 (sff build, no gaming, but I do want to drive 4k 144hz monitors)

 

[Selaya]

the w6400 uses the same die as the 6500xt, so you'll be ... inheriting all its bad ideas (2 outputs, 4x4, lack of hardware de- and encoding), just clocked lower to hit the 75w limit

 

[lexluthermiester]

Isn't PNY the only authorized manufacturer for Nvidia professional GPU ?

Don't think so. But anything is possible.

 

[LupintheIII]

Heck yeah! What's the rest of the system specs. Little backpack beast you got there.

Have you played with the voltage / frequency curve? Seems like there is a good chunk left to gain while still using 70w, pushing performance and efficiency even higher.

Rest of specs are:

• Asus H370 Strix
• i7 8700
• RTX A2000
• Cooler Master Masterair G200P
• Vengeance lpx 2666 CL12 16GB
• HDPLEX 160W DC-ATX
• WD Blue 250GB M.2
• Seagate Barracuda 1TB 2.5"

Added a picture to show how the back looks like.

Tried some OC following handy graph provided by @W1zzard, got second graphic score in Timespy for now (first 2 are with 2XA2000 in some kind of SLI).
To be honest I don't think there would be much benefit to mess around with voltage/freq curve since the card is very much at it's limit and I don't want to fry my PCIe slot... the little GA106 can have pretty huge spikes as per Ampere tradition.
Another strange stuff is Ampere cards like to spit out 1000+ frames during menu screens for some reason (yes it's not only The New World it's just with that game you actually sit on menu for hours). Took me some time to figure out what was causing my system to hard reset... I knew it was power related, wich is to be expected since the whole thing run on a 160W PSU, but that was happening just in menu or in some specific transition area of the benchmark, probably the CPU was also drawing a ton of power trying to keep up with 1000+ fps even tho I've limited it to 45W. Setting max framerate to 144fps in Nvidia control panel fixed the issue. Aslo had this kind of stuff with my RTX 3070 but with a 650W PSU the system didn't crash (never had any of this on my RX 6800 system tho).
Now on GPU wich are severely power limited like the A2000 or 3070 it's not the biggest of the issue, but I'm not sure how a 400W 3080 would cope with that.
In the end reached 60fps avg in Horizon Zero Dawn 1440p max settings, so mission complete

 

[AusWolf]

Why can't this be the RTX 3050?

 

[Lew Zealand]

Why can't this be the RTX 3050?

Because either people would complain about it being undervolt/clock limited to 75W in the locked/slot-power models, and/or people would OC it to match a 3060 in the unlocked/6-pin power models.

 

[lexluthermiester]

Because either people would complain about it being undervolt/clock limited to 75W in the locked/slot-power models, and/or people would OC it to match a 3060 in the unlocked/6-pin power models.

However, there are a great many people that want a card that is low profile, single slot design and runs from slot power for small form factor systems where performance better than IGP is needed. I just recently have gone hunting for such a card and have discovered a shocking lack of recent tech examples. I had to settle for a GTX-750ti because a GTX-1050 or 1650 could not be found or found for a sensible cost, to say nothing of anything from recent GPU generations. This A2000 is an excellent performer for the size but it is not going to be something most consumers will want to afford. So Auswolf's suggestion was spot on, though I think NVidia would be more likely to release something like a GT-3040...

 

[nguyen]

Rest of specs are:

• Asus H370 Strix
• i7 8700
• RTX A2000
• Cooler Master Masterair G200P
• Vengeance lpx 2666 CL12 16GB
• HDPLEX 160W DC-ATX
• WD Blue 250GB M.2
• Seagate Barracuda 1TB 2.5"

Added a picture to show how the back looks like.

Tried some OC following handy graph provided by @W1zzard, got second graphic score in Timespy for now (first 2 are with 2XA2000 in some kind of SLI).
To be honest I don't think there would be much benefit to mess around with voltage/freq curve since the card is very much at it's limit and I don't want to fry my PCIe slot... the little GA106 can have pretty huge spikes as per Ampere tradition.
Another strange stuff is Ampere cards like to spit out 1000+ frames during menu screens for some reason (yes it's not only The New World it's just with that game you actually sit on menu for hours). Took me some time to figure out what was causing my system to hard reset... I knew it was power related, wich is to be expected since the whole thing run on a 160W PSU, but that was happening just in menu or in some specific transition area of the benchmark, probably the CPU was also drawing a ton of power trying to keep up with 1000+ fps even tho I've limited it to 45W. Setting max framerate to 144fps in Nvidia control panel fixed the issue. Aslo had this kind of stuff with my RTX 3070 but with a 650W PSU the system didn't crash (never had any of this on my RX 6800 system tho).
Now on GPU wich are severely power limited like the A2000 or 3070 it's not the biggest of the issue, but I'm not sure how a 400W 3080 would cope with that.
In the end reached 60fps avg in Horizon Zero Dawn 1440p max settings, so mission complete

Try overclock + undervolt your A2000 to 700mV or less to prevent the dangerous peak power usage, capping FPS (or enabling V-Sync) work but for online games you would like the highest FPS possible (without negatively affect system stability).

Here is how to Overclock + undervolt (in your case, type in 330mhz core clocks and undervolt to 700mV)

 

[AusWolf]

However, there are a great many people that want a card that is low profile, single slot design and runs from slot power for small form factor systems where performance better than IGP is needed. I just recently have gone hunting for such a card and have discovered a shocking lack of recent tech examples. I had to settle for a GTX-750ti because a GTX-1050 or 1650 could not be found or found for a sensible cost, to say nothing of anything from recent GPU generations. This A2000 is an excellent performer for the size but it is not going to be something most consumers will want to afford. So Auswolf's suggestion was spot on, though I think NVidia would be more likely to release something like a GT-3040...

Exactly! It's sad to think about how many high-end cards nvidia has released in the Ampere range (3070, 3070 Ti, 3080, 3080 Ti, 3090, 3090 Ti), yet there isn't a single option for small form factor gaming / HTPC from either nvidia or AMD. On AMD's side, the RX 6500 XT / 6400 could have saved the segment if AMD hadn't decided on using a PCI-e x4 laptop GPU with no video decode unit whatsoever. On nvidia's side, there is absolutely no reason for the 3050 to be a full-height card with a power connector, as it is clearly demonstrated here by the A2000. -2% relative performance at 1080p with nearly half the power consumption. GeForce cards are ridiculously overpowered, which is just another sad fact. On the other hand, the 6500 XT delivers 75% of its performance while consuming 50% more, which makes it an even worse contender.

 

[Cutechri]

GeForce cards are ridiculously overpowered, which is just another sad fact.

Yep, I undervolted my 3070 and set a static core clock of 2100 MHz and it works just fine while consuming 60W less. From 240W to 180W. Insane.

 

[AusWolf]

Yep, I undervolted my 3070 and set a static core clock of 2100 MHz and it works just fine while consuming 60W less. From 240W to 180W. Insane.

Even just playing with the power target is quite interesting. My 2070 performs only 7% worse if I lower the power target by 29%.

Why can't nvidia set humanly workable voltages and power targets, and save the extra for overclockers?

 

[Selaya]

Why can't this be the RTX 3050?

because it'd make little sense from a binning perspective. The A2000 has about as many cores active as the 3060 - the point of the 3050 is to use up the dies that didn't quite make it to a 3060 or A2000, but aren't like totally broken to the point of being wholly useless.
Since the vast majority of users aren't doing SFF, the 3050 consuming 150W for higher performance's just fine, if it means more availability through binning.
After all, that's why the A2000 exists - an option for those who cannot use a 3050 instead.

 

[londiste]

Why can't nvidia set humanly workable voltages and power targets, and save the extra for overclockers?

Saving extra for overclockers is rather bad business sense.
Other than that - competitive landscape. And not only AMD vs Nvidia but also previous generation cards. It needs to be better on xx% better than the competition and to get that edge, silicon is pushed as far as reasonably possible. Improvements in manufacturing process updates have slowed down considerably, bringing even more incentive for trying to maximize gains.

 

[Lew Zealand]

However, there are a great many people that want a card that is low profile, single slot design and runs from slot power for small form factor systems where performance better than IGP is needed. I just recently have gone hunting for such a card and have discovered a shocking lack of recent tech examples. I had to settle for a GTX-750ti because a GTX-1050 or 1650 could not be found or found for a sensible cost, to say nothing of anything from recent GPU generations. This A2000 is an excellent performer for the size but it is not going to be something most consumers will want to afford. So Auswolf's suggestion was spot on, though I think NVidia would be more likely to release something like a GT-3040...

The A2000 has a wider memory bus and far fewer cores cut than the current 3050, making it in most aspects a 3055, more expensive to make than the current 3050, but downclocked to fit a specific wattage target. Which is a type of card Nvidia hasn't made for consumers in a decade or more. That's why I answered that it could never be a 3050 and instead, a cheaper to make, higher clocked, but consequently higher power card is a better match for Nvidia to call the 3050, which they did. IMO a downclocked 2560 (or lower) core 3050, called a 3040 or whatever, is a more likely 75W SFF target. I don't expect to see this product made.

No argument that there is a demand for a current-gen, well specced SFF slot-power only card, I'd love one. And while there may be a great many people interested in this product, it's not nearly a great many enough to convince Nvidia or AMD yet to release something current-gen to fill that niche. Or apparently even for the AIBs to produce enough SFF 1650s to fulfill demand during the current CF situation. I wanted one about a year ago but went a different direction instead.

Even just playing with the power target is quite interesting. My 2070 performs only 7% worse if I lower the power target by 29%.

Why can't nvidia set humanly workable voltages and power targets, and save the extra for overclockers?

That's not optimized for benchmark comparisons, throwing 50% more power at 10% more performance is. Those who care about heat/longevity of components/power usage etc. just undervolt/clock and use our PCs that way. I learned this by necessity as I started with power- and cooling-restricted PCs and GPUs and just got used to doing this optimization for every one I own.

 

[tabascosauz]

Not possible, the fan speed slider goes from 30% to 100% (not 0% to 100%)

I meant more the custom fan curve functionality in Afterburner, where you can manually draw the entire curve. Does it still limit you to 30% at the lowest point?

Though I guess it could just as easily accept a < 30% speed on the curve but just ignore it

 

[lexluthermiester]

The A2000 has a wider memory bus and far fewer cores cut than the current 3050, making it in most aspects a 3055, more expensive to make than the current 3050, but downclocked to fit a specific wattage target.

I'd be happy with a Geforce branded card like that, but it needs to be single-slot.

 

[Mussels]

Even just playing with the power target is quite interesting. My 2070 performs only 7% worse if I lower the power target by 29%.

Why can't nvidia set humanly workable voltages and power targets, and save the extra for overclockers?

Yep, I undervolted my 3070 and set a static core clock of 2100 MHz and it works just fine while consuming 60W less. From 240W to 180W. Insane.

Because AMD is genuinely competing this time around.
When theres competition, use up all the headroom.
No competition? Focus on efficiency, use that headroom for reliably known future products (for the shareholders)
(See: intel running quad cores almost the same for 10+ years, with only memory clocks really changing)