AMD Vega Hardware Reviews

Once again, this further proves how limited DSRB is, with 30 % memory savings only 10% higher fps was achieved, which is a best case scenario here, the driver dont really need to have it active for all applications, as most of them will not benefit from it anyway. as mentioned here before only turly bandwidth starved chips will have a tangible benefits from DSRB.
Excluding the only confirmed test with the Energy benchmark with a 100% increase, this could be accurate.

I'd also have to imagine reducing the bandwidth, TMU, ROP, ALU, cache usage, etc with DSBR and culling would affect more than just "bandwidth starved" chips. Not as if reducing energy spent on memory accesses would allow higher clocks either.

now manufacturers will resort to the ridiculous tactic of giving false hopes to potential buyers with the concept of magical drivers down the road!
If by false hopes you mean features guaranteed to increase performance. It's not magic to toggle features on and off. If you examine your control panel there are plenty of settings to force AA and other features. They're hardly magic. Calling something magic is just straight ignorance.
 
Writing the software relatively easy. Standardizing an API for integration with graphics APIs across vendors a bit more problematic. Intel and Nvidia should be able to do something similar, so just a matter of "quickly" pushing it through a committee.

Not forgetting the application compatibility issue.
 

That's lolworthy as hell. They (ASUS) promised higher current limits for their card, and now nothing? With all the VRM power phases they've stacked onto that giant PCB that's nothing short of utter bullshit.

Looking at what that finnish guy is reporting in that Reddit thread when it comes to overclocking and undervolting, the ASUS board is actually performing worse than AMD reference designs. Man, what a shitty product, I'm feeling so let down.
 
The reviewer says that the Hotspot temp is 98C,no wonder the card barely nudges ahead of reference.

I'm facing the issue where the core temperature can be below 60C yet the Hotspot runs away to 100C once the power limit is raised. I don't think it's the vrms now, many people have separate vrm temperatures. But if it's on-die, I'm flabbergasted as to how it can have this big of a delta from the core.

Very frustrating to see the card capable of 15-20% extra and not being able to exploit it. :(
 
New The reviewer says that the Hotspot temp is 98C,no wonder the card barely nudges ahead of reference.
It appears other AIB are facing the same issue as well, it's delaying their custom cards, among other issues.

So what gives? Sources tell us that there is too much variance in the quality of the chips AMD is providing. AIB partners are unable to figure out a stable overclocked GPU frequency that works for all cards, and therefore cannot provide any sort of warranty on factory-tuned cards. Further, there continues to be discrepancies between the temperatures the GPU is reporting and what AIB partners are finding in actual measurements. This is true of the actual GPU and the capacitors below the GPU. We have some follow-up testing that will reveal more about these issues.

MSI isn't even planning to release custom Vega cards as a result.
MSI’s response surprised us. The company traditionally offers re-engineered graphics cards with custom PCB designs for all high-end GPU platforms, but it appears to be skipping the Vega lineup. A company representative told us that MSI “won’t be making a custom card anytime soon,” but could offer no additional information.
http://www.tomshardware.com/news/amd-vega-custom-graphics-cards-problems,35514.html
 
Bad core temp. sensor would explain why some people are getting lower temps than the cooling liquid with liquid cooling setups and why the exhaust air is hotter than what the core temp. would suggest.

I'm not sure if the reviewers got better chips or not, their firestrike/other benchmark scores are similar to what I get and so I don't think they're boosting any further.
 
It appears other AIB are facing the same issue as well, it's delaying their custom cards, among other issues.
One of those issues undoubtedly being HBM chip stacks being shorter than the GPU die on some assemblies. This was the case with Hexus' original ASUS Strix review board, and as a result the HBM ran quite hot, since it had no direct contact with the cooler. It's an unbelievable shitshow that AMD is producing such units and seemingly mixing them randomly into customers' orders. You'd need to custom-machine the hotplate on the cooler for the mismatched die heights to get proper contact, but if your chip orders may vary you can't prepare and place custom hotplate orders accordingly.

As a result it seems 3rd party manufacturers aren't bothering to account for the height mismatch (that Strix cooling plate is flat as a mirror), leading to poor RAM cooling and limited performance on what is supposed to be an expensive, high-end, premium product. Very very bad!
 
Hasn't made much difference in my case, maybe a bit worse besides the voided warranty, but not sure if it's a nocebo now that I've started monitoring the hotspot temperature. I have no problem with the fan noise of reference cards but it's ridiculous that this temperature isn't exposed in wattman if it's the main cause of throttling and I was running the chip ragged while trying to figure out the throttling.

Mine has samsung memory, if gpuz is correct, and is also molded which apparently happens with samsung chips.
 
One of those issues undoubtedly being HBM chip stacks being shorter than the GPU die on some assemblies. This was the case with Hexus' original ASUS Strix review board, and as a result the HBM ran quite hot, since it had no direct contact with the cooler. It's an unbelievable shitshow that AMD is producing such units and seemingly mixing them randomly into customers' orders. You'd need to custom-machine the hotplate on the cooler for the mismatched die heights to get proper contact, but if your chip orders may vary you can't prepare and place custom hotplate orders accordingly.

As a result it seems 3rd party manufacturers aren't bothering to account for the height mismatch (that Strix cooling plate is flat as a mirror), leading to poor RAM cooling and limited performance on what is supposed to be an expensive, high-end, premium product. Very very bad!

I dont think your right at all.

I dont know anyone who is having a hard time clocking the HBM, ( my vega 56 hits 1100 and i haven't bothered to try higher). hotspot temp is on SOC and this is the one that people appear to hit first. My card doesn't seem to have this problem(i haven't check but my card is always at temp target when it throttles ).

Based off my playing with an actual Vega, i think AIB problem is Vega is already really amazing at pushing right up to the boundary of stable clock ( just like Zen). Its dynamic adjustments to keep stable while setting a higher dynamic target beats out setting a higher static clock rate. Vega also has a massive clock wall around 1650-1700 mhz, there is nothing i can do to break that on any test or game ( this is just like Zen). The only people breaking 1700 mhz are people putting it under water and even then they aren't getting much above it ( this is just like Zen).

So put all that together and what exactly can AIB's do? All they can do is really offer cards with AIO or cards with better fans, but there isn't much room from improvement in performance over a ref Vega with a vega 64 bios, setting power limit to +50 upping dynamic clock + 100 and taking 50-100 mv off voltage.

I really think the fundamental problem is butting up against the boundary of 14LPP, hopefully 12nm improves max clock of the process, both Vega and Zen in the enthusiast space will benefit from that.

I really think people are getting carried away shitting on Vega..............
 
I dont know anyone who is having a hard time clocking the HBM
Again, Hexus' Strix Vega wouldn't clock even up to 1000 before they got garbled graphics on-screen. As I recall, they were some ways from 1000 as well actually. Unfortunately that review was unpublished, due to the "wrong" BIOS being used, which turned out to be the only BIOS we'll get... :rolleyes:

Seen other reports of people whose HBM can't do 1000, much less 1100.

I really think people are getting carried away shitting on Vega..............
You're way off base there mate. Just...way off.
 
Again, Hexus' Strix Vega wouldn't clock even up to 1000 before they got garbled graphics on-screen. As I recall, they were some ways from 1000 as well actually. Unfortunately that review was unpublished, due to the "wrong" BIOS being used, which turned out to be the only BIOS we'll get... :rolleyes:

Seen other reports of people whose HBM can't do 1000, much less 1100.
Maybe they actually had a fault card?
I've seen the vast majority or "reports" from people who have no problem, you want to floop it out and measure it i will smash you on that point. All i have to do is go to any forum go to the vega OC thread and search the words HBM/memory.

here is the first one:
http://forums.overclockers.com.au/search.php?searchid=33229385
dont see any one with a vega 64 or moded 56 below 1000 there.

http://www.overclock.net/newsearch/...=recency&sdate=0&search=HBM&type=all&start=25
Or have a look here almost everyone in the 1000-1100 range.
https://forums.overclockers.co.uk/search/5314988/?q=HBM&t=post&o=relevance&c[thread]=18789713
oh look lots more 1000-1100 HBM clocks.



You're way off base there mate. Just...way off.
Orly. jpg
Maybe its your own bias playing out, Vega already performs where it performs it doesn't need the invention of mythical problems.
 
Again, Hexus' Strix Vega wouldn't clock even up to 1000 before they got garbled graphics on-screen. As I recall, they were some ways from 1000 as well actually. Unfortunately that review was unpublished, due to the "wrong" BIOS being used, which turned out to be the only BIOS we'll get... :rolleyes:

Seen other reports of people whose HBM can't do 1000, much less 1100.


You're way off base there mate. Just...way off.

I could do only 930MHz HBM2 with stock Vega 56 BIOS, but after flashing Vega 64 BIOS I can hit 1100MHz. My problem is keeping HBM2 temps in check. They loose performance when going above 85C HBM temp. and it's quite easy to get there when mining. GPU will stay at 67C but HBM will go up to 87C if I want to keep fan below 3000RPM. I need around 3300RPM to keep it just under 85C ... That's why I'm taking my card apart tomorrow, wish me luck as I hope to get HBM2 temp. under control without changing sink just yet.
 
I could do only 930MHz HBM2 with stock Vega 56 BIOS, but after flashing Vega 64 BIOS I can hit 1100MHz. My problem is keeping HBM2 temps in check. They loose performance when going above 85C HBM temp. and it's quite easy to get there when mining. GPU will stay at 67C but HBM will go up to 87C if I want to keep fan below 3000RPM. I need around 3300RPM to keep it just under 85C ... That's why I'm taking my card apart tomorrow, wish me luck as I hope to get HBM2 temp. under control without changing sink just yet.
I run @ 3200 rpm, which is above the sound level i would like, but its a low frequency its not that loud so it doesn't bother me, I do hope good value coolers do appear for vega, i would love to replace the blower, i dont even need more cooling capacity just not a blower....lol
 
I dont know anyone who is having a hard time clocking the HBM, ( my vega 56 hits 1100 and i haven't bothered to try higher). hotspot temp is on SOC and this is the one that people appear to hit first. My card doesn't seem to have this problem(i haven't check but my card is always at temp target when it throttles ).
Are you using official stock bios?
 
Back
Top