HWUB:Radeon RX 5700 XT Overclocked, RTX 2080-Like Performance?

S

Snowdog

[H]F Junkie
Joined
Apr 22, 2006
Messages
11,262
HWUB tests the Powerplay table hack with a liquid cooler to see if it lives up to all the hype.

Short version it doesn't.

While he gets near 20% clockspeed boost, he only gets about 7% increase in game performance, but power usage increased by nearly 40% going from 186 watts to 258 watts in Far Cry:ND.



Edit:

Text version that was not done when originally posted:
https://www.techspot.com/review/1883-overclocking-radeon-rx-5700/
 
Last edited:
Snowdog said:
HWUB tests the Powerplay table hack with a liquid cooler to see if it lives up to all the hype.

Short version it does't.

While he gets near 20% clockspeed boots, he only gets about 7% increase in game performance, but power usage increased by nearly 40% going from 186 watts to 258 watts in Far Cry:ND.

Click to expand...


Doesn't the 9900k see about the same performance/watt ratio? Everybody agrees it's a good cpu. Why not the same thing for a gpu?
 
Kind of puts a damper on the AIB cards being much faster than the blower cards.
 
2080 performance from a 5800xt???? it was a triple cross all along-
upload_2019-7-24_15-14-21.png
 
He may have been getting performance regression from pushing the chip too hard, Its a hard call but the excessive power consumption is a little telling. I didn't watch the entire thing, did he use the soft power play tables to overclock?
 
Snowdog said:
HWUB tests the Powerplay table hack with a liquid cooler to see if it lives up to all the hype.

Short version it does't.

While he gets near 20% clockspeed boots, he only gets about 7% increase in game performance, but power usage increased by nearly 40% going from 186 watts to 258 watts in Far Cry:ND.

Click to expand...



What's scary is that people are doing this on the blower. Imagine how bad the VRMs and other components are cooking on those cards. HWUB/TechSpot is quickly becoming one of my favorite tech sites. They benchmark important games like Fortnite, R6:S etc and are very even handed with respect to both NVIDIA and AMD.


Bankie said:
Kind of puts a damper on the AIB cards being much faster than the blower cards.
Click to expand...

If that Arctic Cooler mod is anything to go by, even AIB cards will likely have very hot spots in excess of 90C with the powerplay tables and clocks pushed to 2 ghz+.


Randall Stephens said:
Doesn't the 9900k see about the same performance/watt ratio? Everybody agrees it's a good cpu. Why not the same thing for a gpu?
Click to expand...

The 9900K offers world class gaming performance, can you make the same claim w/respect to 5700XT?
 
Last edited:
agreed on last poster.

GDDR5 and up are ECC built in to various levels, likely version 6 is very tight control for error checking i.e push too hard regardless of temps (sometimes extra cold HRT not help) extra voltage does little but increase energy requirement AND direct increase in temperatures (very large amount really for small gains past the "hump" )

Likely "my opinion" not took the time to see where it seemed "happy" and tweaked the fk out of it from that point vs trying to run as fast as possible (even under liquid, keeping in mind, these cores are VERY small, Ngreedia chopped a bunch of things away to allow to run as fast as 1xxx/2xxx are capable and CPU wise every very hard to push beyond 4Ghz (rough ~1.5-1.7Ghz GPU core speed)

^.^
 
5150Joker said:
The 9900K offers world class gaming performance, can you make the same claim w/respect to 5700XT?
Click to expand...

Absolutely. Out of the hundreds of gpu variants out there, it's easily within the top 10, likely in the top 5 overclocked. When we use ambiguous terms like "world class" we can make just about anything world class. I had a world class karaoke friend in college. She knew her stuff, I mean songs.
 
Randall Stephens said:
Absolutely. Out of the hundreds of gpu variants out there, it's easily within the top 10, likely in the top 5 overclocked. When we use ambiguous terms like "world class" we can make just about anything world class. I had a world class karaoke friend in college. She knew her stuff, I mean songs.
Click to expand...

9900k is the top (#1) performing gaming cpu on the market. Can you say the same about 5700XT?
 
AMD needs to get off of their asses and make a 384-bit 12 GB card from Navi. At least tease us with something!

Makes me think that Nvidia and AMD have some weird drug deal going on...
 
sabrewolf732 said:
x1950xtx was pretty beasty
Click to expand...

My last purchased ATI device was actually the X1900XT (same family I think). With Ruby on the shroud of, quite possibly the loudest blower fan to ever grace a GPU. The thing was literally a wind turbine when at 100% fan speed.
 
ZeroBarrier said:
My last purchased ATI device was actually the X1900XT (same family I think). With Ruby on the shroud of, quite possibly the loudest blower fan to ever grace a GPU. The thing was literally a wind turbine when at 100% fan speed.
Click to expand...

Nah, it has nothing on the stock blower of a 290X...I had the X1900XT, before I went to an Arctic Cooling aftermaket (this was a LAN box so I did not want water in it given this time period in history) and it was a good bit quieter then the X1900XT...The XT was loud, no doubt, but I think the 290s edge it out lol.
 
It's probably worth wc'ing the 5700's just for the thermals and reduced sound alone.
I love HWUB (my #1 view in the morning) but they're not hardcore o/c'ers. No tweaked memory tables available yet....
Also, still early launch version drivers - it's going to get better.
/a 2020 refresh with better/faster memory might be interesting
 
Lumpus said:
It's probably worth wc'ing the 5700's just for the thermals and reduced sound alone.
Click to expand...

You have to really, really want an AMD GPU at that point. As mentioned above and elsewhere, you can have more performance with less noise along with more features just by stepping up a performance bracket, without needing to watercool.

Still probably best to wait for third-party cooling unless you know that you're putting one under water and tossing the blower.
 
:/ as I suspected... the chip has already been pushed as far is it can reasonably go.
 
Last edited:
GoodBoy said:
:/ as I suspected... the chip has already been pushed as far is it can reasonably go.
Click to expand...
It sounds more like the chip isn't the issue... if you're pushing that much more mhz... and the speed isn't increasing much, it's more of a memory problem, which means they should be able to get more speed by either getting faster memory or widening the memory interface, all of which means the chip and architecture is probably not the bottle neck. What would probably help a lot is if they could figure out some better compression to alleviate this bottleneck, I think this is probably a large difference between AMD and NVidia right now.
 
I'm real confused because plenty of sites got 2100mhz or VERY close to that, in the first reviews. Before the power table mods were out.

So.....what's the point?
 
chameleoneel said:
I'm real confused because plenty of sites got 2100mhz or VERY close to that, in the first reviews. Before the power table mods were out.

So.....what's the point?
Click to expand...

Did any actually hold that when a game was running? I doubt it.
 
There appears to be an issue with the memory controller for the architecture more than the memory itself.

All Ram should have some form of overclocking headroom, but from what I have seen, there is almost none.

There are some users claiming something like a 7 % memory overclock on memory, but is this 100% stable as the results shown by gamersnexus or hardwareunboxed? Unlikely.

It's insane how quickly the positive news of igors mode without people question the results or authenticity when there were some clear red flags.

Not only was only 1 game tested, a 25% increase in frequency, added voltage and a massive power limit increase only resulted in 37 more watts or 17% more power doesn't appear physically possibly. Bumping the power limit up alone would increase the power consumption to that much. Add volts and a 25% overclock and your power consumption would climb significantly. Optimistically speaking, perhaps igor made a mistake and report the GPU only power consumption.

Other add things is the timing as Igor was able to get a full waterblock only 3 days(probably got the block before launch since testing takes time) and the close communication between him and RTG group directly.

It shows how insane AMD viral marketing department is, as the news of igors review spread like wildfire but the corrections much less so. The truth the reviews are showing is that the RTX 2070 super has the same type of overclocking headroom and can do it with the stock cooler with minimal increase in noise unlike the 5700xt which needs high fan speeds, aftermarket cooler or a waterblock. Sure might just be a 10% difference when both cards are OCed, but add the game bundle, RTX and the better noise and power characteristics when both are OCed and AMD might be coming a bit short to win over the market.
 
I don't know, I'm more prone to lower voltage to reduce power without losing much performance.. my r9 fury is still undervolted and overclocked, netting me like 5% bump in speed and I haven't measured the reduction but should be close to 10% based on simalar configs. I haven't messed much with anything newer, mostly my kids game now and pretty much all 1080... Not a huge reason to bother.
 
Ready4Dis said:
I don't know, I'm more prone to lower voltage to reduce power without losing much performance.. my r9 fury is still undervolted and overclocked, netting me like 5% bump in speed and I haven't measured the reduction but should be close to 10% based on simalar configs. I haven't messed much with anything newer, mostly my kids game now and pretty much all 1080... Not a huge reason to bother.
Click to expand...


Can't say which driver I am using, and I only have 1 50th AV ED sample, but my initial impression over multiple driver releases is that Navi does not behave like FURY/Vega did with undervolts increasing performance while lowering power usage.

Now I have not gotten my sample under water yet, EK is half a world away and just released the blocks today, but I will say memory cooling is important. Also can confirm it is not tapped out (at least my sample). I expect the 25-30C drop from water to increase this headroom, but that is just my opinion until I can test it for sure.


I will report back as soon as I'm block arrives, but I can't say a lot.
 
You must log in or register to reply here.
Back
Top