Der8auer Tries to Overclock Radeon VII

uberzero

n00b
Joined
Mar 13, 2018
Messages
13
Its a rebrand of the Vega Instinct card, its a compute monster, stock though it loses to a 1080Ti and the 2080 from almost every benchmark I have seen,
Only on benchmarks heavily favouring Nvidia code wise.

On the plus side Vega.....ehhh...7nm 25% performance increase is great, Navi will in all likelihood be a lot better of an arch. so the benefits should be exceedingly better, I think if AMD can bring 60fps 1440p max settings to the base budget model GPU on the new arch they will have a clear winner, the unknown is NVidia and what they will do, 7nm double/triple Pattern or wait for the EUV/UVL in 2020, they are starting the 5nm EUV/UVL process thus dropping the prices on the 7nm to affordable terms that we *MAY be able to afford as consumers, if they drop die size with the already proven arch AMD might have some troubles ahead, either way 2020 will be great Navi 10, Navi 20 from what I understood was supposed to be this card but they scrapped the idea and went with another idea from a guy who retired which was the Vii(stopgap until Navi), so depending that may or won't show its face(keep in mind that info is from leaks so it could be BS about Navi20).
Key issue is software related. When an application is not properly optimized, a powerful and well designed hardware will not make a difference. That is currently the problem in PC from developers to publishers. Note how AMD sponsored application does not penalize competition unlike vice versa.
 

NKD

[H]F Junkie
Joined
Aug 26, 2007
Messages
8,274
Only on benchmarks heavily favouring Nvidia code wise.


Key issue is software related. When an application is not properly optimized, a powerful and well designed hardware will not make a difference. That is currently the problem in PC from developers to publishers. Note how AMD sponsored application does not penalize competition unlike vice versa.


I agree. you can see that AMD optimized and sponsored games to seem to do just fine on their hardware (AC odyssey being exception lol). Then in other games its like you scratch your head lol. Hopefully one day AMD can have more developer mindshare but its been the case forever. Looks as if AMD has to optimize quite a few games on their end and some titles get ignored if they are not demanding titles. Their DX 11 drivers don't seem to be an issue with popular games but older ones still seem to hurt quiet a bit.
 

Qthulu

Weaksauce
Joined
Apr 19, 2018
Messages
103
When you blah blah blah just so you can have videos over 10 minutes long, of course viewers will watch something else. I had to skip 3/4ths of Der8auer's video just for him to get to the point. Digital Jesus is the same way. At least he marks topics with time stamps. Digital Jesus has a really bad case of the diareah of the mouth. Jay is more mainstream, and he doesn't repeat himself as often. Just my two cents...
Obviously you've never had to lick your way to the center of a tootsie pop
 

GoldenTiger

Fully [H]
Joined
Dec 2, 2004
Messages
19,828
No not every benchmark. next time when you make a point lets stick to facts shall we? There were games where it beat the 2080 stock. When you started off with that, I lost interest in what you had to say next. lol.
He said almost every benchmark, not every one. Let's stick to the facts shall we?
 

NKD

[H]F Junkie
Joined
Aug 26, 2007
Messages
8,274
He said almost every benchmark, not every one. Let's stick to the facts shall we?
Almost every benchmark is pretty close to every benchmark. There were more then a few games it did just fine against the 2080. Not really going to care much about older games or AC odyssey.
 

horrorshow

Supreme [H]ardness
Joined
Dec 14, 2007
Messages
7,797
When I see this guys name and the word "tries", I immediately think: not. good.
 

filip

2[H]4U
Joined
Aug 15, 2012
Messages
2,288
So AMD does it again, I was really hoping that AMD was hitting the market hard competition wise so we can get some reasonable prices or better performance for the cost. I can see the future now 10% bump in performance: RTX 3060 $399, RTX 3080ti $1499, Titain 2.0 $2999.
 

PhaseNoise

2[H]4U
Joined
May 11, 2005
Messages
3,297
So AMD does it again, I was really hoping that AMD was hitting the market hard competition wise so we can get some reasonable prices or better performance for the cost. I can see the future now 10% bump in performance: RTX 3060 $399, RTX 3080ti $1499, Titain 2.0 $2999.

Not if folks don’t pay it.

If they keep selling at higher prices, expect higher prices.
 

psyclist

Gawd
Joined
Jan 25, 2005
Messages
844
Cheesus Crust AMD! you had one job...Well seems to be the trend these days, all releases are rushed from all teams. I do like that were are seeing 2000+ clocks using the auto overclock. Give it a months time, hopefully OverdriveN, AB and GPU-Z will get there programs updated for the new design. And in that time AMD can un-fuck the driver situation. I really do think that once opened up, this chip is going to fly.
 

FlawleZ

[H]ard|Gawd
Joined
Oct 20, 2010
Messages
1,149
When I see this guys name and the word "tries", I immediately think: not. good.
Yeah, it's more than just fishy. I'm just a regular dude and the first thing I thought of was "its Vega arch, it likely responds to undervolt."
This dude is a professional overclocker. A legend in some eyes. To assume he is incompetent would be naive. The agenda was clear from the get go on this.
 

PhaseNoise

2[H]4U
Joined
May 11, 2005
Messages
3,297
Yeah, it's more than just fishy. I'm just a regular dude and the first thing I thought of was "its Vega arch, it likely responds to undervolt."
This dude is a professional overclocker. A legend in some eyes. To assume he is incompetent would be naive. The agenda was clear from the get go on this.

But generally when undervolting is the answer, that shows that the manufacturer felt for their yields they had to use a higher voltage to ensure stability. Thus, success stories with undervolting should be assumed to be no more than ~80% of the parts (gaussian distribution of quality).

While that sounds pretty good for the single consumer, they are after those remaining parts which would otherwise fail (I'm using 20% as a reasonable handwaving number) to increase yields. This becomes more urgent when the yields are already low - which seems to be the case.

So basically, I wouldn't fault the guy. He appears to be working with settings which are more applicable to The Everyday Buyer.
 

odditory

Supreme [H]ardness
Joined
Dec 23, 2007
Messages
6,482
But generally when undervolting is the answer, that shows that the manufacturer felt for their yields they had to use a higher voltage to ensure stability. Thus, success stories with undervolting should be assumed to be no more than ~80% of the parts (gaussian distribution of quality).

While that sounds pretty good for the single consumer, they are after those remaining parts which would otherwise fail (I'm using 20% as a reasonable handwaving number) to increase yields. This becomes more urgent when the yields are already low - which seems to be the case.

So basically, I wouldn't fault the guy. He appears to be working with settings which are more applicable to The Everyday Buyer.
It's very simple what's going on here, and the review sites are going to twist themselves into pretzels trying to be delicate about it: like the Fury days, AMD has resorted to maxing the chip out from the factory just to be able to keep up, leaving no OC headroom. This is the wall der8auer ran into. The undervolting nonsense can't abstract this truth.

So what we're really seeing in the benchmarks is max OC "brand A" versus brand N with room to spare -- most will recognize theyre getting 20-25% more card with the latter.

A $699 MSRP only pushes it further afield from non-fanboy/novelty purchase consideration.
 
Last edited:

PhaseNoise

2[H]4U
Joined
May 11, 2005
Messages
3,297
It's very simple what's going on here, and the review sites are going to twist themselves into pretzels trying to be delicate about it: like the Fury days, AMD has resorted to maxing the chip out from the factory just to be able to keep up, leaving no OC headroom. This is the wall der8auer ran into. The undervolting nonsense can't abstract this truth.

So what we're really seeing in the benchmarks is max OC "brand A" versus brand N with room to spare -- most will recognize theyre getting 20-25% more card with the latter.

A $699 MSRP only pushes it further afield from non-fanboy/novelty buying considerations.

Exactly. This cost and performance level was available about 2 years ago (but with better power efficiency and headroom0, so it is unclear to me why this could be considered a success.

Not bagging on AMD, I wish them the best, but this is another stopgap.
 

5150Joker

Supreme [H]ardness
Joined
Aug 1, 2005
Messages
4,533
Funny that this guy figured it out so quickly when others didn't. The key is undervolting...

Well the question is what temperatures are people paying attention to? Junction or edge temperature? Obviously the OC'ing of this card is dependent on junction temperature so undervolting would probably help bring it down and thus be able to add a meager overclock. However, high junction temperature would be worrysome to me as an owner of this card because it means a point on the die is running like an oven. I've heard reviewers try to compare this with CPUs but those don't have interconnects directly tied to the hbm ram which might be sensitive to high temperature spots--it would explain why AMD has added in junction temp monitoring. So then I wonder when people do reviews, do they take an avg of junction + edge or what? How can you compare the AMD temps vs NVIDIA if AMD is reporting two different types of temperatures?
 

jaggafeen

n00b
Joined
Feb 5, 2019
Messages
12
been playing games at 5k ultrawide 5120x2160 with radeon 7 i get well over 60 fps on all latest games including bfv at ultra settings and the card is quiet and undervolted to 940mv
 

jrobdog

Limp Gawd
Joined
Dec 4, 2006
Messages
300
been playing games at 5k ultrawide 5120x2160 with radeon 7 i get well over 60 fps on all latest games including bfv at ultra settings and the card is quiet and undervolted to 940mv
Nice, any way to measure the power draw at that mv? Does the undervolt increase the boost clock?
 

jaggafeen

n00b
Joined
Feb 5, 2019
Messages
12
Nice, any way to measure the power draw at that mv? Does the undervolt increase the boost clock?

i didnt try to adjust the clocks so boost was the same but saw a massive drop in temps and fans were quiet before the undervolt and even quieter now.

only proper way to measure would be at the wall but i dont have the equipment , but a 108mv drop in voltage should see a massive drop in power draw. In the graph below you can see r9 fury undervolt and power draw.

im so happy with my purchase, the press release drivers are the reason why it had bad reviews i have had no stability issues while using it under stock settings or playing with wattman.

i just punched in 940mv and it was stable will try later and see how far i can go but i doubt it will be much lower judging by other attempts at undervolt. I think have a good card that is a keeper. at least im not scared of memory degradation and my card failing.

I bet performance per watt is far better than rtx cards with this undervolt.


R9 FURY NITRO UNDERVOLT VS POWER DRAW.
HMucG5n.png
 
Last edited:

jrobdog

Limp Gawd
Joined
Dec 4, 2006
Messages
300
i didnt adjust the clocks so boost was the same but saw a massive drop in temps and fans were quiet before the undervolt and even quieter now.

only proper way to measure would be at the wall but i dont have the equipment , but a 108mv drop in voltage should see a massive drop in power draw. here is r9 fury undervolt and power draw.View attachment 140732

Thanks, 940mv is quite low. You're making me regret not getting one, although I have a 1080ti. I'm thinking R7 will be quite a performer once the drivers are ironed out. I'm wondering if a brave soul will try to flash the MI50 vbios.
 

jaggafeen

n00b
Joined
Feb 5, 2019
Messages
12
Thanks, 940mv is quite low. You're making me regret not getting one, although I have a 1080ti. I'm thinking R7 will be quite a performer once the drivers are ironed out. I'm wondering if a brave soul will try to flash the MI50 vbios.

looking forward to unlocking more power with a bios mod
 

HeadRusch

[H]ard|Gawd
Joined
Jun 8, 2007
Messages
1,365
It always makes me sad to see how Der8auer and GamersNexus combined have less subs than Jayztwocents

GamersNexus's delivery is just too dry.....love the metrics he provides but he's kind of a humorless guy, his videos feel more like briefings than entertainment, which perhaps is the point. But I get why he might not have as many subs, he's not as entertaining to watch as the Linus's and Pauls and Jays of the world tend to be.....IMHO of course.....
 

jaggafeen

n00b
Joined
Feb 5, 2019
Messages
12
cant stand that jayz guy. they may be entertaining for some but guys like that lack serious content.
 

ffd

n00b
Joined
Feb 14, 2019
Messages
3
For the price there is no reason to buy this over a 1080ti or RTX 2070. AMD must be smoking crack if they think people will stump up over $600 for a card that barely beats an already overpriced RTX 2070.

People are blaming Nvidia for overpricing and I guess so especially the competition is not there.

However, the main reason Radeon 7 is priced as a 2 years-old 1080ti at 699 - is simply that Moore's law stopped working, even it seems the industry is going towards 7nm, 5nm. The pricing is simply keep up with the inflation rate.
 

ffd

n00b
Joined
Feb 14, 2019
Messages
3
I didn’t know that? They’ll never crossfire or the drivers just don’t support it yet?

AMD was puttting the coals to Nvidia at the release of the RX580 talking how two RX580 in crossfire would beat Nvidia’s top card for hundreds $ less and how they aimed to support Crossfire long into the future because they supported their products and the technology better than their competition. This publicized about the time Nvidia was publicly starting to to depreciate SLI’s future and I took it as AMD poking a bit of corporate finger at Nvidia for dropping inferior SLI while they were going to push the superiority of crossfire. Anyone else remember that? Crossfire did typically work better than SLI, and so I was excited they were going to continue to support it. My crossfire experience with my Fury X was great and I had no notable issues. My brothers SLI experience with his Nvidia 970 cards was so poor he said he’d never SLI again.

Then Crossfire didn’t work on Vega for months after launch —- unannounced to launch day buyers of course — me being one of the people burnt on that lack of advance information with the purchase of my pair of Vega 56 which got 10 FPS in Battlefront while Crossfired. Vega was the first card past RX580. So that promise to continue to support Crossfire didn’t even make it to the launch of the very next card!!! They got Crossfire working I heard, but I no longer owned the cards when they did.

Now Radeon VII is the first card past Vega and they did it again??? Hopefully it’s only temporary...

AMDs marketing fell flat in a hurry if Crossfire is truly not to be supported into the future? That kind of stuff makes me mad. It’s opportunist and anti-consumer to trumpet support of something, and then drop it hardcore the very next product cycle.


I had same hope, since I got 2 RX480, then 2 Vega 64. It would be super sad/stupid decision Radeon 7 did not support CF, hence AMD do not have 2080ti-grade offering in games support CF :)

BTW: anyone knows if Instinct MI50 card support CF-like technologies?
 

ffd

n00b
Joined
Feb 14, 2019
Messages
3
I had same hope, since I got 2 RX480, then 2 Vega 64. It would be super sad/stupid decision Radeon 7 did not support CF, hence AMD do not have 2080ti-grade offering in games support CF :)

BTW: anyone knows if Instinct MI50 card support CF-like technologies?


This is CF or multi-GPU???

 
Top