RX 570 vs RX 5700 XT

Given limited resourses, which combination would you rather have?

2x RX 5700 XT = ~ 110 MH/s @ 230-250W

5x RX 570 = ~ 130 MH/s @ 400W

1 Like

You need to consider mostly your initial cost of hardware and the ratio of power to hash.

For example, 2 x RX 5700 XT, smaller PSU, less risers, less heat. 110/230 = 0.4478 ratio (bigger is better)

5 x RX 570, larger PSU, 5 risers, more heat, 130/400 - 0.325

Lastly, how much money you will need to spend on getting a system with 2 x 5700 XT cards vs 5 x RX 570.

If you read the latest bios mods of 5700 XT, you will find that you can reach with some updates for a ratio of over 0.5. Some members report getting over 58MH/s on 5700 XTs, with power as low as 120W / card.

It will take you time and effort to learn to optimize and mod cards. But even as stock, my preference is to keep the rig’s density high and go with the 5700 XT.

And last, consider the resale value after a couple of years. I speculate that used 5700 XT in a couple of years will still be worth something, and with 8GB, money will be able to mine ETH and other currencies.

This is my opinion only.

AlleyCat

I ended up doing a Mh/W for the system, then comparing the performance gain vs the power consumption, the 5700XT system is better.

Thanks for guiding me in the right direction.

Would you consider buying second hand gpus? Theyre cheaper but I worry about the abuse from previous owners

Personally, I would buy new, but only on sale.

if been playing with my 5700 for weeks now and here some pointers:

  • forget 58 Mhs @ 120W claims…its only if you are lucky with getting samsung memory; watts are more realistic as 140W
  • DONT bios update 5700 to XT, there is zero mining boost to be gained but your card runs hotter, especially memory and uses ~5 - 10 watts more.
  • your realistic has for 5700 is 57.5 Mhs @ 140W, this is easy to accomplish without too much hassle and its rock stable

remember not to trust AMD drivers info about wattage. For some odd reason, this wattage is not taking into account of power usage of riser pins that deliver power to GPU (i dont mean power usage of actual riser itself, which is only few watts). You can measure true power usage of each connections with ammeter…

Thank you for your comments. I agree that accurate power loss needs to include the power draw by the risers, CPU, PSU, etc. Accurate power consumption can be measured at the wall. This question is about the comparison between the cards’ density or performance. Therefore, I expect that the same overhead of power loss would be about the same for different cards.

As far as overclocking, I am sorry that you can’t get any improvements. The question by ksukhin is about the comparison between 5700 XT and not the 5700. My answer is about the performance of RX 5700 XT and not the RX 5700.

You can find a detailed testing report on the modifications of multiple RX 5700 XT at https://www.igorslab.de/community/threads/gigabyte-5700-xt-bios-mod-fails.3304/

I encourage anyone to test and report their results. On my system, HiveOS reliably report about 59MH/s at the power of 112 - 116W.
I understand that the actual power should be measured at the wall.

I encourage you to replicate the tests and report what works and what doesn’t work with the cards you are using. Any failure is an opportunity to learn something new and to improve.

Thanks,
AlleyCat

That’s funny, that’s exactly what I planned on doing.

Where I live, a 5700 XT is around 550 cad. I’ve seen some second hand ones for around 400. They’re gonna be a mix of brands (xfx, sapphire, msi).

I was going to start with 2 due to budget constraints so I’d have the time to mess around with tweaking them.

Thanks for giving me a bit more of a realistic expectation. I just got back into mining after leaving in 2015 after the DAO hack. So it’s been fun catching up on resources.

I like this Igor website, thanks!

A 5700 was going to be my second choice, it just saved me some time by not messing with the XT bios.

I still may if I get one, I believe each card is unique and you’re going to have to mess with each one to get it just right.

I’m running 8 x 5700 XT in my rig. All with stock BIOS.
2 x Sapphire + 6 x Gigabyte
Total power consumption: 710 Watt (89 watt avg.)
Total Hashrate: 409 MH/s. ( 51 MH/s avg.)


Sapphire consume 2-3 watt less than Gigabyte and produce .30 MH more
MH/Watt ratios:
Sapphire: 51.3/87= 0.59
gigabyte: 51.03/90 =0.57

Sapphires are $70 more expensive than Gigabytes in my country.
First I bought 2 Sapphire and 2 Gigayte. Measured and compared their performance. Both are quite stable as rock. There is really very tiny performance difference. So decided to go with Gigabytes and Ordered 31 more to set up 3 full rig (My motherboard runs 13 GPU)

Damn, I wish I had the budged to order that many. How long have you been mining and have you hit ROI yet?

I used to mine back in 2015 when eth was 7usd, sold after the DAO hack and never looked back until recently. Given my electricity cost, I can hit ROI in 12 mo at current rates. I can also allow myself to keep this rig for that duration too so it’s more of a hobby/interest.

I didn’t consider the 5600s previously but I will now. The Mh/W ratio is pretty good compared to RX 570/580s.

5700 XT is 5700; they are identical products but stuffed with different bios…so why bother paying more for just bios that you can flash your self for free :thinking:

I agree with @Calathia because you spend wayyy more time doing the BIOS mods and fiddling with the settings to get it optimized and your rig is not always mining because you are fiddling with BIOS and settings.

I can use stock GPU and increase the memory speed slightly and get a 5% increase in the Mhash/sec. Then let it mine, then move on to the next rig. Less hassle and headache and keeps the ball rolling. Can you imagine BIOS modding 100 GPU or more?

So I ended up buying 3x RX 5700s but I’m having a tough time bypassing the RSA signature check.

I have then tuned so they do 161MHs @ 347W so approx 0.47Mh/W. Once I figure out the BIOS signature bypass, I’ll flash them

1 Like

software reported or out of the wall?