lolfail9001
[H]ard|Gawd
- Joined
- May 31, 2016
- Messages
- 1,496
1525, of course. AMD gives no fucks about stereotypes you have about clocks.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
-
Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.
Vega Rumors
- Thread starter grtitan
- Start date
Magic Hate Ball
Limp Gawd
- Joined
- Jul 28, 2016
- Messages
- 366
You may be right, only time will tell.
"Frontier Edition" isn't guaranteed to replicate previous "Pro" cards. In my opinion it probably will, but new price tier/focus might alter the usual expectations.
Tom's chart shows the relationship between clocks and watts and uses best voltage they managed.
If you want to say it could be lower, then the same can be said about Nvidia.....
Anyway nothing wrong with their setup and approach as it is backed up by PCPer in quite a few way, and the scientific results from a scope and Tom's are the only ones to correlate clocks/watts/voltage matches up reasonably well with the blogs I mention and performance/efficiency, especially when one takes into compute/clocks.
You may disagree but Tom's methodology and also PCPer are pretty fine IMO for quite a few reasons; one has access to the scientist/engineers of the test gear provided support and the other employes Allyn who is a pretty good engineer (ex Nuclear technician) and worked with the Navy in electronics.
But we can both agree AMD has greater flexibility and options.
Cheers
Last edited:
Sounds great on paper.
Guess it didn't translate well in covering this usage scenario. Your Toms link was about gaming and clock speed vs watts used as clock speed climbed - no one saw that coming. I didn't actually see the part where they radically undervolted Polaris GPU AND cut its clocks, tightened the RAM timings and flashed or cut (depending on OS) the speeds to 1750 with the tighter timings ; guess I missed that piece. The guys at bitcoinist did an excellent job comparing the 1070 to the 390x and tuned the 1070 cards really well - but now you say they could have done better. I was looking for the Polaris numbers, but I missed those.
Here is a simple every day Joe https://bitcointalk.org/index.php?topic=1543701.40 mining with six cards in his rig and pulling 669 watts for the entire rig... mind you , he's not an engineer or a blogger like all those other guys and you should probably discount his and my and all other miners methods because they were based on trial and error. And you certainly should NOT look for examples and guides from the many available for bringing your Polaris under control for best mining efficiency.
Cheers
Last edited:
Don't need to be an engineer to do this stuff
. Just have to be careful on what your doing. And don't need gold samples or what not to do it either. This is my first foray into mining, and I'm beating the pants of the guy you just linkedThe rx480 is much better with efficiency than the rx580 in mining. I should have added when not dual mining and only doing eth, the rx 580 goes down to 120ish watts. But Dual mining the nV cards don't increase in power usage, AMD cards do.
Don't need to be an engineer to do this stuff
The rx480 is much better with efficiency than the rx580 in mining. I should have added when not dual mining and only doing eth, the rx 580 goes down to 130ish watts. But Dual mining the nV cards don't increase in power usage, AMD cards do.
580's are actually the same, you can flash them to 4 Gig 480's same as I do, in fact, they undervolt on average a bit better. Sometimes you give up nice to have things like oh.... display outputs but ... well... it just doesn't matter.
Dual mining, I have no doubt stresses things more, I haven't tried and I doubt I will.
In the article Tom's reduced voltage to 0.8V that ironically is what the guy your happy to link did?
It is pretty difficult to dismiss the article as it shows the actual TDP; ask yourself why it takes PCPer and Tom's hardware to get AMD to correct their 480 power distribution and lower its TDP if everyone can do this without being an engineer....
Actually the article and my context was purely the watts/clocks/voltage,
It had little to do with gaming from this context and like I keep saying you cannot break the relationship between core clocks and compute.
The blog guys i linked showed they could do 32/MHs at 140W and not using CUDA8, so what do you manage with the 580?
Another blogger can do 36 MH/s stable with multiple GPUs but involves using CUDA8 and with more overclocking.
From an engineer background myself it is not accurate to rely upon basic budget power meters to tell your usage and power traits.
Anyway that link the guy reduced to mimum stable voltage of 0.82V at 900MHz.
From Tom's article you seem to be dismissing similar voltage and frequency:
Tom's said:
And you are seriously well under 5 TFLOPs compute at 900MHz for 115W - part of this comes back to my point you cannot break the compute/clock relationship and this is still integral to crypto mining even if memory overclocking is the primary overclock.
Nvidia has greatly improved the compute capability of Pascal from the 1070 upwards, before this it was pretty mediocre and one reason they were pretty average for mining (along with lack of optimised support from algos/tools relating to crypto mining).
And this is how a power demand actually looks when measured accurately - context is how quickly it changes:
Yeah it is a game but it is still important and accurate data when considering TDP; case in point just look at the TDP of a Nvidia Tesla accelerator used purely for compute and no graphics/games and the TDP of its comparable gaming model, they end up very similar and we can use this to provide an accurate trend if applied to compared models.
Anyway if you want to measure accurately without using a scope, then you need to buy a decent calibrated PAT/ or a decent mains monitor, still not cheap but much cheaper than oscilloscope.
The reason; it comes down to data points/sampling/ticks-interval/window of the analyser.
Here is an example of the Kill-a-Watt (used by the guy you linked) being inaccurate relative to a Fluke tester someone compared with and some responses:
https://forums.anandtech.com/threads/kill-a-watt-power-meter-accuracy.137169/
Another exacerbation comes down to power regulation/efficiency design and Power Factor Correction with the PSU that skews measurements and some tools such as Kill-a-Watt.
But lets agree to disagree as we are digressing this thread.
Cheers
Last edited:
I didn't flash mine so yeah that could make a difference, that is a good idea thx, But dual mining a lot more stress on AMD cards, don't know why but even the claymore dev says the same thing
I should try that next if i can ever get my hands on the rx 580's lol. I'll have my newegg business account in a couple of weeks, lets see if I can fair better over there :/
I already requested that. Nice that you agree.
So Raja saying the gaming cards will indeed be faster then the Pro card is BS? The FE can use the gaming card drivers.Well it's a 12.5 tflop card with 4096 shaders, that's pretty self explanatory
View attachment 27505
1600MHz is the maximum boost clock detected in the compubench( I think thats what its called) leaks. Where are you pulling this number from? Cause it's not on any official AMD slides.
Gaming cards are normally clocked way faster than the pro cards? Where did you pull that up? A quick glance at the specs for the FirePro cards shows that the Polaris based SKUs have the same boost as consumer, same with Hawaii , and same with Fiji.
Based on the specs, none at all.
It's a 300W card at 12.5tflops, with a CLC. You tell me.
By what reasoning ?
Yet here Vega 10 is shown to be 40% faster than Fiji while it is 45% faster in terms of theoretical tflop/s
If it reaches those clockspeeds it will match a 1080Ti on paper, as I said earlier.
Look at current pro card generation, clocks compared to current generation gaming cards:
- AMD Radeon Pro WX 7100 1188mhz,peak 1243mhz, 2304sp, is slower then the equivalent Rx 580 at peaks of over 1.4ghz
- Radeon Pro WX 5100 typical 713mhz peak of 1086mhz, 1792sp, is slower then the equivalent Rx 570, 1792sp, with peaks over 1.3ghz
- AMD Radeon Pro WX 4100 typical 1125mhz, peak of 1201mhz 1024sp, vs clock of the Rx 560, 1024sp, up to 1.3ghz
Of course you know it will be 1525mhz, wait I thought you said it was going to be 1200mhz some time ago, maybe that was someone else.Face it, Vega will most likely be faster than a 1080Ti. Not in everything but in many things.
I love your enthusiasm when it comes to Vega and I really hope you're right. If they can just somehow match the great launch that was RyZen, then who knows. Maybe Vega can be a real winner. I'm just not quite holding out on it performing at 1080Ti levels. But if it does, then I'm in for sure.
If it doesn't then AMD better be ready to really price it well. To me the indications are it will be a great performer - yes way above a 1080 on the top model. Since there are two Vega chips Vega 10 high end and Vega 11 mid level we don't know if Vega 11 will be the 1070/1080 competitors using DDR 5(x) memory. No real word or sense of a x2 version either. We already know Polaris covers the sub $300 (if cards ever become available again), Vega 11 should be above that unless it was canned. Vega 10 speeds on the top gaming cards should naturally be faster then those on the pros since the requirements of the two are way different for different reasons. We are looking at greater then 1600mhz clocked Vega's I do believe and 1.5-1.6+ performance over Fiji unless Vega really turns out to be an over revving non-performer which does not appear to be the case.
Last edited:
Ieldra
I Promise to RTFM
- Joined
- Mar 28, 2016
- Messages
- 3,539
So Raja saying the gaming cards will indeed be faster then the Pro card is BS? The FE can use the gaming card drivers.
Look at current pro card generation, clocks compared to current generation gaming cards:
Now you saying that the next Pro cards or the highest pro card is going to be clocked faster then the gaming cards?
- AMD Radeon Pro WX 7100 1188mhz,peak 1243mhz, 2304sp, is slower then the equivalent Rx 580 at peaks of over 1.4ghz
- Radeon Pro WX 5100 typical 713mhz peak of 1086mhz, 1792sp, is slower then the equivalent Rx 570, 1792sp, with peaks over 1.3ghz
- AMD Radeon Pro WX 4100 typical 1125mhz, peak of 1201mhz 1024sp, vs clock of the Rx 560, 1024sp, up to 1.3ghz
Face it, Vega will most likely be faster than a 1080Ti. Not in everything but in many things.
The RX 580 is an inefficient power hungry RX480 that is simply overclocked well above the comfort spot in the V/F curve. The reason you don't see 1500mhz pro cards is A) they're based on the 480, B) reliability is key and C) they can't be 250W.
This is not a very well thought out comment.
12.5 tflops necessitates 1525mhz, it is already 300W according to amd at these clocks.
It needs 1770mhz to match a 1080ti.
HAve a nice day
Edit:
The boost clock of rx580 is 1340 mhz, not 1.4+ as you claim. Please do not make false claims that can be disproved with a cursory search on the amd website.
Raja sated that the FE would be the most powerful compute card, do you know that the clockrate is directly proportional to the compute throughput? I guess not, otherwise you wouldn't have asked how I got 1525...
Last edited:
Ieldra
I Promise to RTFM
- Joined
- Mar 28, 2016
- Messages
- 3,539
Gee, I wonder why this
Doesn't clock as high as this
Doesn't clock as high as this
lol, comment proved you were wrong, pro cards are clocked less then gaming cards. Gee I believe everyone else knew this except you. The FE is clocked at 1600mhz which I believe is peak. The gaming versions that don't have the blower style cooler, hint hint will be clocked faster. There is also rumored a water cooled version - will that top 1900mhz? 2000mhz?. Final top speed will be the determining factor if AMD stuffs Pascals line up. So go back to Nvidia la la land.
As for efficiency, it also depends on work loads, games etc. One simple generalization falls short of any reality. For example mining; big spew in this thread - AMD efficiency is better then Nvidia's. So even your understanding of the bigger picture maybe needs to be reviewed a little. Yes in many games Nvidia kicks some ass, some games AMD efficiency is competitive. Other stuff AMD does better.
Exactly - The gaming cards are clocked faster
lolfail9001
[H]ard|Gawd
- Joined
- May 31, 2016
- Messages
- 1,496
Actually, you proved him right because rx480 ref clocks are 13Mhz higher than Wx7100's paper clocks. A whole fucking 13Mhz! By this measure, Vega Gaming Edition will be a whole 15Mhz faster than Fool's Edition.
And please, don't dream up clocks on the spot, it's even sillier than idiots dreaming up 1600Mhz on Polaris a year ago.
Why are you using Rx480, the pro cards use the revised chips and came afterwards. So no not at all. Then again even the Rx 480 had way faster versions for the gaming cards so that would still be true - gaming cards are faster then the pro cards.
Of course not, no speeds there, water rumor versions you can find. My premise is the gaming cards will be faster then the Pro cards, this is rather normal. Raja indicated the gaming cards will be faster then the Pro FE card - we just don't know by how much. The size of the chip is way bigger then it should be indicating a less dense faster speed design - how much faster? Anyone's guess. MY GUESS is 1700mhz+
lolfail9001
[H]ard|Gawd
- Joined
- May 31, 2016
- Messages
- 1,496
Source it bruh. Pro cards use the same freaking Polaris 10 rx480 does. The only reason they were out after Rx480 is due to additional validation they had to go through as workstation cards.
You can overclock pro cards too, so you are wrong wx 5100 is faster than any rx460 on the market, if you slap a waterblock on it.
Characteristic of automated design AMD practices so much.
I am sure the FE would be the same if you OC it and slap a water block on it. So we should have some good data end of this month to next. Revised chips there, the 4100 uses 1024SP like the 560, the 5100 has 8gb of ram vice 4gb of the 570. They are all based off of the newer Polaris configurations.
lolfail9001
[H]ard|Gawd
- Joined
- May 31, 2016
- Messages
- 1,496
Just because it uses same config as Polaris rebrands does not mean it is based off of Polaris rebrands (that were released well over half a year later) lmao. By this claim Macbook Pro GPU is based off of rx 560 kek
Does it matter? The gaming cards are faster then the Pro cards in the end right now. The gaming Vega cards will be faster, particularly the AIB ones if AMD allows some lead way there over the Pro cards. The Pro cards have to perform with lower power, thermals and be 100% stable - gaming cards have some leeway to push it. So 1600mhz on the FE is very telling and does hint Vega will be a good performer unless AMD hosed up the design. I have a feeling AMD is going to get this one right, hopefully the launch will go smooth.
Ieldra
I Promise to RTFM
- Joined
- Mar 28, 2016
- Messages
- 3,539
There is no difference between rx480 and rx580. They are using the same chips. I have shown you that there are no discernible clock differences between stock P10, Hawaii XT and Fiji consumer and pro based cards.
Vega FE is watercooled, or rather, the best performing version is, which is the one they base the 12.5tflop figure on, because common sense.
So that leaves us with a Vega based pro card, which I have proven to not clock any worse than the gaming variants, with as high a TDP as any gaming variant will reasonably have (300w) sitting at 1525 mhz boost.
The Hawaii based firepro cards had a clock of 950mhz, same as original 290x. Judging by your replies you will argue it is clocked low because of the 390x... Which could draw something like 350w. Yeah no shit it's clocked lower.
Now, since the hilarious single slot cooler pro cards you linked are still running at the same clocks as an rx480, you will have to explain to us what logic exactly leads you to believe a watercooled Vega FE will clock 300mhz lower than the gaming variants without it seeming ludicrous.
Good luck!
Edit:
It appears this has not quite hit the mark... 12.5 Tera flops... 4096 shaders... 1525mhz. Not 1600.
Vega FE is watercooled, which means the 1525mhz SKU is watercooled
rx 480 and rx 580 are not same chips. that is loads of bullshit I have heard. I have tried around 3 rx 480s, none of them, absolutely none of them clocked above 1350 stable, no matter what voltage, all with 8 pin power adapters. While one rx 580 I had was rock solid at 1470 all day long with 1.17v. So no not the same chips. Rx 580s are certainly improved and revised chips.
Ieldra
I Promise to RTFM
- Joined
- Mar 28, 2016
- Messages
- 3,539
I choked on my coffee, this is a good one.
I've observed thousands of pigeons in my life time, but I've noticed three white pigeons once shat a pentagram therefore I have concluded that while pigeons in general are of this earth, the white variety are the spawn of Satan by the principle of NKD logic.
Even more relevant, I know people with RX480s that have been running them above 1400MHz since day one, by your logic this must mean that the Polaris GPU in the RX480 was also a revised and improved version of the GPU in the RX480.
This presents a philosophical problem as it is unclear what exactly they revised and improved upon at this point
Edit:
I revised and improved my post
Last edited:
I bet you $100 cash. Pull up all polaris from early days, take 100 samples and then take 100 samples of rx 580s. Clock them and pull up all rx 580s. Rx 580s will do more clock. Rx 480s simply didn't have the consistency and clock speed that 580s come with. 580s are more tweaked and tuned to run consistently at higher clocks. Don't give the bullshit logic and theories about pigeons. How many rx 480s have you come across? of my 3 or 4 non got even close to 1400. that is 100% failure rate. So don't tell me rx 480 could give you same clocks as rx 480s. yea rx 580 uses more power but no matter how much voltage I pumped in to rx 480s the clocks just werent there. So my experience with rx 480s speaks for itself. I could care less what your hate speech says.
I am no amd lover. I have nothing amd in my system. But I do know rx 480s did not clock worth shit! Shit even the reviewers had hard time getting to 1400, I can link a few. You act like all rx 480s were clocking at 1470 no problem. I am sure Kyle will say otherwise.
Ieldra
I Promise to RTFM
- Joined
- Mar 28, 2016
- Messages
- 3,539
Im not the one with the bullshit logic and crackpot theories...
Your sample size of 3 is irrelevant, and even if it were representative of all cards your conclusion is totally unfounded.
On average 580s do clock higher than 480s,but does that mean amd made changes to the chip? No.
You can find plenty of 1400mhz+ 480s,just like you can find 580s that won't clock any higher than 1400. It's a lottery.
Its the same GPU.
If Kyle were to conclude that amd made changes to a chip based on one sample he tested then I would stop valuing his conclusions, simple.
Since when is common sense hate speech...
D
Deleted member 93354
Guest
One way to know if there has been revisions is to check the dye shot which would have the build and revision number stamped on it.
Ieldra
I Promise to RTFM
- Joined
- Mar 28, 2016
- Messages
- 3,539
Been there, done that.
Am looking for the pictures but this was already debunked when they launched, all the coverage says it's just a 480 with pre overclock
Last edited:
It doesn't necessarily have to be a new revision. It could just be GF has refined the process a little where there is a less voltage fluctuation and the cards can retain little higher clocks. almost all 580s clock 1400+ guaranteed. There is something certainly they tweaked, I think it just more refinement then anything. I don't think there are many changes here. I think this was evident in later generation 480s.
you said 480s clocked just like 580s. which is not correct. So inturn are saying 580s are indeed improved and clocking higher than 480s so there is indeed a difference? See I can do it too lol
JustReason
razor1 is my Lover
- Joined
- Oct 31, 2015
- Messages
- 2,483
You have made no substantiated claims at all, just arguing for the sake of arguing. Although you did state you knew people with 1400+ 480s so they are the same as the 580. You left no room for argument that there is any reason that the 580 clocks higher, alluding that they were exactly the same even down to clocks.
Just as a reference.
The improvements by GF mean that the 5xx is more stable with higher voltages and generally this has been increased to around 1.15V for both 570 and 580; Asus Strix 570 is 1.15V where their same 470 model was for similar characteristics at 1.0V
Interestingly Sapphire were able to run higher voltages generally; their Nitro+ 480 was around 1.15V while the Limited Edition 580 is now 1.168V to 1.19V, however due to greater leakage current with the older GF process/silicon and the 480 the new Limited Edition 580 card has a pretty similar power demand/TDP when measured - that said the power demand/TDP of the Nitro+ was pretty large relative to other 480 models at lower voltage.
Although worth considering the Limited Edition are probably the best binned silicon.
But the power demand/TDP increases to support the higher voltages and clock frequencies as one would expect with it being in essence the same GPU, just greater stability due to the changes by GF.
Cheers
The improvements by GF mean that the 5xx is more stable with higher voltages and generally this has been increased to around 1.15V for both 570 and 580; Asus Strix 570 is 1.15V where their same 470 model was for similar characteristics at 1.0V
Interestingly Sapphire were able to run higher voltages generally; their Nitro+ 480 was around 1.15V while the Limited Edition 580 is now 1.168V to 1.19V, however due to greater leakage current with the older GF process/silicon and the 480 the new Limited Edition 580 card has a pretty similar power demand/TDP when measured - that said the power demand/TDP of the Nitro+ was pretty large relative to other 480 models at lower voltage.
Although worth considering the Limited Edition are probably the best binned silicon.
But the power demand/TDP increases to support the higher voltages and clock frequencies as one would expect with it being in essence the same GPU, just greater stability due to the changes by GF.
Cheers
Last edited:
lolfail9001
[H]ard|Gawd
- Joined
- May 31, 2016
- Messages
- 1,496
Substantiate this claim, because it contradicts reality.
He did, actually, just read
Ieldra
I Promise to RTFM
- Joined
- Mar 28, 2016
- Messages
- 3,539
Sorry, maybe I wasn't clear enough in my previous posts - my point was that you can't argue anything really based on one or two samples you may have had.
There is absolutely no doubt that 580 are clocked higher on average than the 480s, but claiming that the former uses a different chip is another thing entirely.
I have read 580 reviews where the highest it would go is 1430 mhz oc, some would hit 1500.
There *were* users reporting similarly high stable overclocks with 480s.
When the xfx GTR launched and it was on average clocking higher people on this forum started claiming it was a new revision of Polaris 10.
Im sure you can confirm this if you just search, and all this started because we were discussing Vega and it's clocks. Unrealistic expectations have burned amd several times, why can't we just pay attention to what they announce and not set exorbitantly high expectations then have the inevitable bitter backlash on forums/reddit etc.
If Vega 10 can run at ~1530 mhz stably for long term operation I will be impressed, that is a big step forward, does it need water for that though? Generally speaking if they announce a new product line, the tflop number the put up in massive font in the presentation will be as high as possible, and they will round up and use this squiggly line ~ I think it's called a tilde
I Expect that to be refer to the watercooled FE, and it's rated for 300w. It is not unreasonable to assume it does not have much headroom, it's not like amd have a history of overrating their boards
Edit: when the very first Vega leaks started to claim 1500mhz etc I did not believe it, in fact i think it's just as likely that they just happened to fart out a click bait article that got it right for once. I was pleasantly surprised to see clocks that high confirmed officially by AMD, it's nice for a change. Was pleasantly surprised with Ryzen performance too
There is absolutely no doubt that 580 are clocked higher on average than the 480s, but claiming that the former uses a different chip is another thing entirely.
I have read 580 reviews where the highest it would go is 1430 mhz oc, some would hit 1500.
There *were* users reporting similarly high stable overclocks with 480s.
When the xfx GTR launched and it was on average clocking higher people on this forum started claiming it was a new revision of Polaris 10.
Im sure you can confirm this if you just search, and all this started because we were discussing Vega and it's clocks. Unrealistic expectations have burned amd several times, why can't we just pay attention to what they announce and not set exorbitantly high expectations then have the inevitable bitter backlash on forums/reddit etc.
If Vega 10 can run at ~1530 mhz stably for long term operation I will be impressed, that is a big step forward, does it need water for that though? Generally speaking if they announce a new product line, the tflop number the put up in massive font in the presentation will be as high as possible, and they will round up and use this squiggly line ~ I think it's called a tilde
I Expect that to be refer to the watercooled FE, and it's rated for 300w. It is not unreasonable to assume it does not have much headroom, it's not like amd have a history of overrating their boards
Edit: when the very first Vega leaks started to claim 1500mhz etc I did not believe it, in fact i think it's just as likely that they just happened to fart out a click bait article that got it right for once. I was pleasantly surprised to see clocks that high confirmed officially by AMD, it's nice for a change. Was pleasantly surprised with Ryzen performance too
hey hey guys lets get off the polaris stuff, the rx580, rx480 same chip, both of these cards can use each others bioses lol, what does that tell us?
Revision or not, the rx580 is P10 with higher clocks along with that the crazy increase in power consumption we saw with overclocking the rx480.
And yeah gold FE with water cooling, as I stated with Fiji, there is NO reason to have water cooling on a card unless there is a NEED FOR IT.
Revision or not, the rx580 is P10 with higher clocks along with that the crazy increase in power consumption we saw with overclocking the rx480.
And yeah gold FE with water cooling, as I stated with Fiji, there is NO reason to have water cooling on a card unless there is a NEED FOR IT.