• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Want new GPU that uses less power but similar preformance

how much it cost? any news abt it?
RX 590s are going for $220 and RX 5500 is similar in performance (3% slower according to TPU estimate). I'd guess price will be $200~250.
 
Last edited:
Budget for this is $250, lower power means less heat so I'd pay 20-30 extra for 30-40 less watts of waste heat.
 
Last edited:
For the OP look for a used RX Vega 56 and you will be over the moon. Don't pay more than you have to though. They can be had for about $220 US nowadays on Auction sites. If that is too much definitely a Polaris card (RX 470-590) 8 GB would do you good too. The 1660 is a good card and there is nothing wrong with them. In terms of the 5500 and new Nvidia cards waiting is an option as well (you may be able to get one of the aforementioned cards even cheaper).
 
The RX5500 makes sense, as a lot of people have already mentioned.

If price is a concern, it's really hard to ignore the RX 570. Whilst it uses more power at stock settings, Polaris is very easy to undervolt for minimum performance penalty. I have extensive experience with this (20+ unique card samples) and the community experience reinforces the expectation that you are unlucky if you can't get 1250MHz clocks at under 100W.

Tdp is garbage, dont bother using it as a metric
QFT.

Look at a TPU review of the card in question to find power draw since they always include total system draw in a whole bunch of different scenarios ('gaming average' is the most realistic load scenario) and there's always GPU-Z which provides board power draw from the on-card sensors if they are there.
 
Polaris used can be had for around $100...
 
Budget for this is $250, lower power means less heat so I'd pay 20-30 extra for 30-40 less watts of waste heat.
With that in mind, there I can't think of a reason not to go GTX 1660... It easily has the performance you want and then some for future games, etc. The power draw is lower than anything AMD has out and it fits the budget. For your uses and goals (low power/the same or better performance) feels like a no brainer to me.
 
Last edited:
With that in mind, there I can't think of a reason not to go GTX 1660... It easily has the performance you want and then some for future games, etc. The power draw is lower than anything AMD has out. For your uses and goals (low power/the same or better performance) feels like a no brainer to me.
1070 used can be had cheaper for slightly better performance, 1070 uses 150watt vs the 130ish of the 1660.
 
1070 used can be had cheaper for slightly better performance, 1070 uses 150watt vs the 130ish of the 1660.
True.. but...

1. He previously mentioned he bought a GTX 1070 and it was too much card. He did something else/got rid of it.
2. The performance metric was already met and then some by the 1660. Performance isn't the biggest concern of the OP's.
3. You cannot find a NEW GTX 1070 for $250, not even close. I don't know if the OP wants used or not... he hasn't specified even though many have mentioned it throughout the thread... this leads me to believe he wants a new card, but who knows until he actually specifies.

His key, as was stated in the OP and just recently, is power use. Suggesting a card that uses more power than 1660 + he had already and bunked, I don't think is the solution. :)
 
Last edited:
There are mini 1070's though are there not? I understand if he had a full sized one because I have 2 and they require a large case but I'm sure there are smaller versions of the card. Here in Canada I can get a used 1070 on ebay for $250 plus shipping, not new though, new is ridiculous for nvidia cards. Also you're sometimes better off buying used. I had to rma both my 1070's when I bought them new.
 
There are mini 1070's though are there not? I understand if he had a full sized one because I have 2 and they require a large case but I'm sure there are smaller versions of the card. Here in Canada I can get a used 1070 on ebay for $250 plus shipping, not new though, new is ridiculous for nvidia cards. Also you're sometimes better off buying used. I had to rma both my 1070's when I bought them new.
Sure.. but I have no idea what a smaller card has anything to do with this thread and the OP... did I miss something? Wattage to dissipate is wattage to dissipate.

Again, the OP mentioned he bought a 1070 already and didn't like it... can you drop the 1070 thing already? LOL!
 
There are mini 1070's though are there not? I understand if he had a full sized one because I have 2 and they require a large case but I'm sure there are smaller versions of the card. Here in Canada I can get a used 1070 on ebay for $250 plus shipping, not new though, new is ridiculous for nvidia cards. Also you're sometimes better off buying used. I had to rma both my 1070's when I bought them new.

OP said the 1070 was complete overkill. Meaning it was much more performance than he needed. Probably the best GPU mentioned in this thread is the GTX 1650 since he wants less heat (uses around 70 watts average gaming) and similar performance to his R9 290.
 
If it was me I’d wait for the RX 5500 on paper it should be decent and have 580 performance or better and just need a single 8 pin connection
 
The RX5500 makes sense, as a lot of people have already mentioned.

If price is a concern, it's really hard to ignore the RX 570. Whilst it uses more power at stock settings, Polaris is very easy to undervolt for minimum performance penalty. I have extensive experience with this (20+ unique card samples) and the community experience reinforces the expectation that you are unlucky if you can't get 1250MHz clocks at under 100W.


QFT.

Look at a TPU review of the card in question to find power draw since they always include total system draw in a whole bunch of different scenarios ('gaming average' is the most realistic load scenario) and there's always GPU-Z which provides board power draw from the on-card sensors if they are there.

 
If it was me I’d wait for the RX 5500 on paper it should be decent and have 580 performance or better and just need a single 8 pin connection
So, you'd wait for a card that uses more power and is slower than a GTX 1660 for the same price?

5500 has a 150W TDP versus 120W... not to mention actual power use is lower on the latter..
 
Why does it matter to you so much?
 
Why does it matter to you so much?
Just trying to help the OP... why do the same not optimal suggestions keep popping up is the real question to me.

Can you explain why we would wait for the 5500 considering TDP values... and yes, we know, eidar, TDP isn't the end all... keep posting that up like nobody saw it as well. You know damn well that it gives a general idea of the amount of power used and heat to dissipate for GPUs. Where possible, I've used actual numbers from TPU reviews to support my point. I did mention TDP though, indeed (though I could care less about how CPUs rate TDP versus GPUs, you should know that is apples to oranges, considering...).

...or the 1070 suggestions when in the OP he mentioned it was overkill? People need to pay attention and stop posting just to post.

Anyway, I'll let it go... but only because the exiled airmen stepped in to save the day... o_O :roll: :slap: :pimp:
 
Last edited:
RX 590s are going for $220 and RX 5500 is similar in performance (3% slower according to TPU estimate). I'd guess price will be $200~250.

Yeah RDNA might very well shine on smaller dies that aren't clocked to oblivion. RX 5700 isn't bad either.


This is a GPU topic, and TDP figures are definitely ballpark useful there.
 
This is a GPU topic, and TDP figures are definitely ballpark useful there.
Agreed - especially since TDP is limited by the number and type of PCIe power connectors in 75W increments. That puts an upper cap on the total power consumption of a card. Also, GPU drivers (either directly for AMD or third-party for Nvidia) provide far more control and monitoring of power limits and voltages.

When GPU-z reports power usage, that's accurate for the GPU itself. The only reason that differs from the true power consumption of the card as a whole is because it doesn't include VRM heat losses, RGBLED lighting (negligible) or power delivery for the fans, at around 6W peak draw for each fan and probably closer to 3W per fan at normal operating RPM.

VRMs on graphics cards are picked for (approximate) peak efficiency at the rated TDP of the GPU, and that efficiency is usually around 92-95% depending on their quality. That means your GPU that reports 180W power draw in the driver/utility requires an extra 9-15W for MOSFET heat losses and another 6W for fans.

So, when a card reports the GPU-only power draw at 180W, I would bet good money on the total power of the whole card being somewhere between 205 and 220W depending on build-quality and fan speed at that exact moment. The few sites that actually isolate their cards for power consumption testing (THG, of all places!) corroborate this theorycrafting which is why I say I would bet good money on it.

TL;DR
As a rule of thumb, the total card power consumption is going to be somewhere very close to 15% higher than the reported GPU-only power consumption in software.
 
Budget for this is $250, lower power means less heat so I'd pay 20-30 extra for 30-40 less watts of waste heat.

At $250ish, I like the 1660 Ti ... way more performance than you have now. and they currently running $250 - 275. It's about 28% faster outta the box than the 590 and that gaps widens when both are overclocked. Yes, more than you need but a longer future.

TDP is for sizing cooling systems ... it represents the average expected powe usage. Your GPU will spent significant time above and below that level. I measure power usage at the wall. If you want peak and average numbers, you can find them in TPU reviews.


These reviews are also useful for other things. For example, if you compare the 1060 3 GB and 6 GB versions, we see it proven conclusively, that the 6 GB adds nothing to the equation at 1080p ... the data lso shows no performance impact at 1440p. Now the 6GB version is faster than the 3 GB because it has 11% more shaders, but the advantage provided by those shaders remains constant at 1080p and 1440p showing that the VRAM is not adding anything to the equation.

I don't do use "used" cards bit I see 1060s on ebay for as low as $50 ... You didn't say whether you use the cards with "out of the box" settings or use MSI Afterburner. The 580 and 1060 deliver the same performance outta the box, but with MSI AB, the 1060 gets a decent lead.

Sticking to newer cards, the 1660 is about 20% faster than the 580 / 1060 outta the box. They start about $200 and run up to $250 for the better AIB models.

The 1660 is about 7% faster than the 590 "outta the box". The AIB 590s OC about 4% and the AIB 1660s about 15%

The AIB 590 draws about 218 watts average

The AIB 1660 draws about 113watts average

I'm going to assume a PSU efficiency of 85% at operating load and 36 hours of play time per week ... I am paying 24 cents per kwhr but the US average is now about 12 cents.

With those assumptions, the 590 will cost you an extra $27.84 per year or $83,53 over 3 years.

Cooling wise ... to maintain comparable case temps ...

w/ 140 mm fans, one (1) 1200 - 1500 rpm case fan per each 75 - 100 watts
w/ 120 mm fans, one (1) 1200 - 1500 rpm case fan per each 50 - 100 watts
 
Back
Top