You might also like our tear-down of this card: th-cam.com/video/8OpdyP7amv8/w-d-xo.html Find Buildzoid's channel here: th-cam.com/channels/rwObTfqv8u1KO7Fgk-FXHQ.html
Pienimusta Don't sweat it. My uncle has a doctorate in electrical engineering, applied physics, software design, and masters in applied mathematics and another software related degree (I forget which as it's obscure), and he took a couple videos to get used to the explanations. He worked for NASA, various branches of the government where they'd ship him off as a private contractor to black sites for months too years at a time, so... Anyway, one of the reasons a dolt like me gets it (to an extent) is he explains things like a normal person teaching someone who does not know anything about what he is doing. If it weren't for his vids I'd probably never have attempted physical board mods beyond adding mini-heatsinks. I love how he recalls the numbers of all the capacitors and switches, so if something fries bends/breaks it's easy to fix/replace as opposed to sending the card out or needing to buy a new one. I've always been curious as to whether he was self taught or schooled in the conventional sense. I'd guess self taught, possibly with help irl from other people, as he makes what most people think is terrifyingly complex (it's probably basic to him at this point) seem, very complex? Ei climbing Mt. Everest compared to Mt. Hood. They both do excellent jobs explaining things, too the point where they could easily be incorporated into a classroom setting, especially if the teacher was hands on familiar with what they do. Thanks guys!
Actually Hardcore Overclocking That's what I figured as it rarely seems like you are talking "over people's heads", especially after watching a few videos, and possibly doing a little research as the viewer with things you (the viewer) don't understand at all. That's when the explanations seem much more efficient, as the research tends to be quite advanced in terminology. You know what they say, the most intelligent people say things the least intelligent can understand (kinda silly to make a point if no one will ever comprehend it).
So overvolting HBM2 degrades it faster than normal - what about undervolting it then? Does it make any difference? Is there a way to measure the memory degradation? Or does it ust come up as memory errors?
Hey Buildzoid, I got a question. I heard this card has pretty poor VRM cooling since the thermal pad isn't placed correctly, so it will hit ~105 degree. You say they are rated for ~125 degrees? If that for a 2000h life cycle or shorter/longer? How long would the card last on 105 degrees? It still sounds ridiculously hot and I'm not sure if I have to return the card!
Holy jezz... I didn't think any manufacturer would step up the VRM, and while it ain't by much, Asus sure did. Now Asus's RX Vega 56 will more or less be the same in terms of PCB correct?
It baffles my mind that VRM design is so solid with this card but yet asus managed to screw up the cooling. I mean hitting 100C VRM temps easily with stock cooling.
PCB should be a lot smaller with the lack of memory chips from GDDR5. This looks like a recycling from "something" with GDDR5, just with Vega silicone slapped on it. It's cheaper not to make a completely new PCB design for Vega if they don't have to I assume. :)
This isn't anything out of the ordinary, this is typical of the card industry and has been forever now. This is how they keep the cost of custom cards reasonable and within spitting distance of reference cards, in spite of the(typically) more expensive components used. The type of memory used/amount of excess space is irrelevant, this is just smart business decisions. Personally, I'll take a little "wasted" PCB and an off the shelf cooling solution if it'll result in a $50+ MSRP decrease and no meaningful affect on performance.
I was kind off surprised you didn't mention the 'overvoltage' header as being the 'VGA Hotwire'. I saw your other video where you review the ASUS Rampage IV Extreme, as being your first board. That had a header for this. also to be used with the ASUS OC Panel. Form what I can tell this is just a voltage reference (readout), and a sense pin skew. so instead of soldering to the board, one could potentially add a potmeter here to skew the reading. I have a vega strix, and was looking into voltmodding. it seems a lot easier to use this header to voltmod permanently, then using a EVC2 for example.. I do wonder where ASUS takes their ground reference though, and this card doesn't seem to have a 'enable jumper' close to the header... also there are 2 headers near the IR35217, i wonder of one is the BUS to talk to the controller (for a EVC2).
Wow, first ^^ I was waiting for that video, time to watch it! Edit: Seen it now, I am glad that the PCB was not changed to the worse, so now all I have to do is continue the waiting game with AMD. .... Still no custom cards available, I hope ASUS won't charge even more as the reference ones come close to cheap 1080 TIs already.
Oh, They'll charge more. Hopefully they keep it to something reasonable. As good as their cooler design is, it's only worth a $50 hike at worst. I'd forgive $100 though, at the absolute worst... Though I'd probably just buy reference at that point and an EK block if I were buying one. TBH, I'd probably just buy reference and an EK block anyways if I were buying one, just because I like custom loop water cooling and the PCB design isn't OMG better. Just, only marginally more efficient. Which won't really matter to the EK block in an oversized loop.
My problem is that I am not into water cooling, so no EK block for me. I hope they go with 50€ above the MSRP (< sure) max, 100 is really high for just "not throttling" cooling.
Vega owner here: The BIOS isn't the biggest problem for me; what really goes on my nerves is that the Radeon Settings program (just like Ryzen master btw) has a tendency of crashing. They both really don't like the sleep mode btw.
Can u please analyse sapphire rx 64 nitro + limited , to me it looks like it have best power delivery and cooling , it just wont go over 70c no mather what
I love that extra efficiency for free comment; dude its a ASUS ROG Strix card its going to be a premium card and they are going to fleece the f**k out of us for it, they already bent us over and shafted us on all the other VEGA cards so people better lube up if they want one of these versions with the extra bells and whistles.
I'm no genius here but I'm not bad at math wouldn't an 18 phase VRM be possible? like 3 x 6 keeping that efficiency of using all 6 actual phases while being able to increase the load split?
Silver Studio theoretically you could use any number of phases. In pracise you're limited by what contollers are avalable. It's a bit like in ICE the more Zylinders you have the smover the pover delivery is.
It IS possible though... just... not practical, and Buildzoid was only talking in practicality. I mean... they'd have to order them special made, because as Buildzoid just said, no one makes them. The reason no one makes them is also valid. They really are more difficult. Both points would contribute to an astronomic price for very little gain.
No such thing as a stupid question. I'm sure others were wondering the same thing and were just afraid to ask. You asked, we were therefore able to educate not just you, but others by proxy.
Agost40Hz my guess is amd has to sign the bios ( buildzoid mentioned in one of his videos how bios flashing is limited to signed bios) and haven't gotten around to it, so asus can't release. That seems to be the case given that some of the reviewer sample strix vega cards actually ran reference bios. We'll probably have to wait till amd specific aib partners are ready before amd green lights this whole thing as to not throw those AMD specific aib partners under the bus.
On the topic of that IR3599, why would they cut their AC current in half by sending only the top half of it? Why not time it so that the top half of your wave goes to one place and your bottom half to another? Not to mention, where does that discarded current go?
Why don't they make a separate 20 or 24 fase vcore supply , so we can use that on any (new) card we want ? that would keep ( our ) cost down , every time we buy a new GPU , And we can decide for ourselfs if we want a 12 fase , a 20 fase or a 24 fase
Someone at reddit said that you can't flash a 64 Bios onto the AIB 56 because of differing HBM brand (64 = Samsung whereas 56 = Hynix). But Buildzoid here says it's possible. Which is true?
Buildzoid did a stream on Wednesday the 18th with 2 Vega's in crossfire and OC'd and was pulling 1150W at the wall. He got like 250 FPS in Firestrike though, can't remember the settings.
Great Great Video for SURE... thanks a lot Dear.. i hope you can help me with this RADEON Vega 64... its not recognized by windows.. and running ATI flash also say NO ATI card Detected.. appreciate you effort
OK, so the VRM and PCB are a pretty good improvement but performance of the card is the exact same as the reference card... for the extra cost is the only thing we can get is marginal VRM and PCB improvements?
DoctorWho8675309 well, you will be getting a better cooler... But vega is pretty close to its limits, so you won't be getting extra performance out of it, if you discount micro throttling on the core
So is that what is holding this thing back? I'm really getting to the point of not caring what RTG's AIB's come out with because RTG doesn't deserve my money after this unfettered debacle of a release. For all the right Ryzen is... Vega is a complete 180. With the fact its still not right its safe to say AIB's won't be out until at least November if not December with the snails pace of Vega being delivered and then RTG keeping their hands clenched on the BIOS.
I'm studying electronics lately, and one thing that confuses me. If an entire computer can consume like, say, 500w from the wall (120v) that would mean that there's less than 5 amps going through the entire PC. How is it then that a GPU which is just one part of a PC be rated in hundreds of amps of current? Isn't a wall-socket only rated or like a max of 20A?
Remember the equation Watts = Volts * Amps, if the voltage is stepped down to say, 12 volts, then the amperage must go up to maintain the same wattage. In the case of 500 Watts, nearly 42 amps may be pushed over anything 12 volts. However, if you step up to 120 volts, then 500 watts only allows ~4.2 amps.
hello gamer nexus I use a very effective solution to effectively dissipate heat in audio amplifiers and voltage regulators that work with extreme temperatures such as GPUs, using common varnish and current, that if I sanded the entire surface of the heatsink and remove all clues of thermal paste, I have a hd 5870 for 7 years using this system and I use it in processors also try it and tell me the results
Did all the memory fall off in shipping? :) Edit: In case you don't understand how to spot sarcasm, a :) usually means the post is sarcasm. This was sarcasm. Lol
Your comment was my exact thought too the first time I'd seen vega uncovered :p Coming from R9-290x, this sight is quite different :p especially concerning non-fullcover/universal waterblock options this really appeals to me!
You might also like our tear-down of this card: th-cam.com/video/8OpdyP7amv8/w-d-xo.html
Find Buildzoid's channel here: th-cam.com/channels/rwObTfqv8u1KO7Fgk-FXHQ.html
I'm 4th year electrical engineering student (M.Sc) and just nodding my head "Yep, yep, don't understand a thing"
Pienimusta Don't sweat it. My uncle has a doctorate in electrical engineering, applied physics, software design, and masters in applied mathematics and another software related degree (I forget which as it's obscure), and he took a couple videos to get used to the explanations.
He worked for NASA, various branches of the government where they'd ship him off as a private contractor to black sites for months too years at a time, so...
Anyway, one of the reasons a dolt like me gets it (to an extent) is he explains things like a normal person teaching someone who does not know anything about what he is doing.
If it weren't for his vids I'd probably never have attempted physical board mods beyond adding mini-heatsinks.
I love how he recalls the numbers of all the capacitors and switches, so if something fries bends/breaks it's easy to fix/replace as opposed to sending the card out or needing to buy a new one.
I've always been curious as to whether he was self taught or schooled in the conventional sense. I'd guess self taught, possibly with help irl from other people, as he makes what most people think is terrifyingly complex (it's probably basic to him at this point) seem, very complex? Ei climbing Mt. Everest compared to Mt. Hood.
They both do excellent jobs explaining things, too the point where they could easily be incorporated into a classroom setting, especially if the teacher was hands on familiar with what they do.
Thanks guys!
what you said genuinely made me laugh. The thing is, I actually understood what he was talking about.
I'm in medical field. Tech is m6 hobby.
self taught though since I started the channel I do have some in industry contacts so if I don't know something I will ask them.
To be fair for my self, I'm focused to power grids and more "energy side of electricity".
Actually Hardcore Overclocking
That's what I figured as it rarely seems like you are talking "over people's heads", especially after watching a few videos, and possibly doing a little research as the viewer with things you (the viewer) don't understand at all. That's when the explanations seem much more efficient, as the research tends to be quite advanced in terminology.
You know what they say, the most intelligent people say things the least intelligent can understand (kinda silly to make a point if no one will ever comprehend it).
Videos like this are exactly why I am a Patron. GN is the only place online you can find out information this detailed and important.
So overvolting HBM2 degrades it faster than normal - what about undervolting it then? Does it make any difference? Is there a way to measure the memory degradation? Or does it ust come up as memory errors?
Nice I just got "like 1000", thanks so much, just got this card for my DCS/Vive Pro rig and it is rocking on water.
Hey Buildzoid, I got a question. I heard this card has pretty poor VRM cooling since the thermal pad isn't placed correctly, so it will hit ~105 degree. You say they are rated for ~125 degrees? If that for a 2000h life cycle or shorter/longer? How long would the card last on 105 degrees? It still sounds ridiculously hot and I'm not sure if I have to return the card!
There's plenty of videos showing how to fix that I think it just requires a thicker thermal pad
Holy jezz... I didn't think any manufacturer would step up the VRM, and while it ain't by much, Asus sure did. Now Asus's RX Vega 56 will more or less be the same in terms of PCB correct?
Need another review of this card besides Jays, so patiently waiting on GN's. Need to know if the Strix allows greater voltage and power limits.
We're looking forward to working on it. We are presently on hold due to the ASUS VBIOS delays.
It baffles my mind that VRM design is so solid with this card but yet asus managed to screw up the cooling. I mean hitting 100C VRM temps easily with stock cooling.
PCB should be a lot smaller with the lack of memory chips from GDDR5. This looks like a recycling from "something" with GDDR5, just with Vega silicone slapped on it. It's cheaper not to make a completely new PCB design for Vega if they don't have to I assume. :)
Филип Пешић I don't remember where I saw it but Asus used the 1080(ti?) strix board/heatsink I believe.
This isn't anything out of the ordinary, this is typical of the card industry and has been forever now. This is how they keep the cost of custom cards reasonable and within spitting distance of reference cards, in spite of the(typically) more expensive components used. The type of memory used/amount of excess space is irrelevant, this is just smart business decisions. Personally, I'll take a little "wasted" PCB and an off the shelf cooling solution if it'll result in a $50+ MSRP decrease and no meaningful affect on performance.
propably the most entertaining expert on this theme and i like it🖒🖒🖒🖒🖒🖒🖒
I was kind off surprised you didn't mention the 'overvoltage' header as being the 'VGA Hotwire'. I saw your other video where you review the ASUS Rampage IV Extreme, as being your first board. That had a header for this. also to be used with the ASUS OC Panel. Form what I can tell this is just a voltage reference (readout), and a sense pin skew. so instead of soldering to the board, one could potentially add a potmeter here to skew the reading. I have a vega strix, and was looking into voltmodding. it seems a lot easier to use this header to voltmod permanently, then using a EVC2 for example.. I do wonder where ASUS takes their ground reference though, and this card doesn't seem to have a 'enable jumper' close to the header... also there are 2 headers near the IR35217, i wonder of one is the BUS to talk to the controller (for a EVC2).
Wow, first ^^
I was waiting for that video, time to watch it!
Edit:
Seen it now, I am glad that the PCB was not changed to the worse, so now all I have to do is continue the waiting game with AMD. .... Still no custom cards available, I hope ASUS won't charge even more as the reference ones come close to cheap 1080 TIs already.
WOOOOOOOW
Oh, They'll charge more. Hopefully they keep it to something reasonable. As good as their cooler design is, it's only worth a $50 hike at worst. I'd forgive $100 though, at the absolute worst... Though I'd probably just buy reference at that point and an EK block if I were buying one. TBH, I'd probably just buy reference and an EK block anyways if I were buying one, just because I like custom loop water cooling and the PCB design isn't OMG better. Just, only marginally more efficient. Which won't really matter to the EK block in an oversized loop.
My problem is that I am not into water cooling, so no EK block for me. I hope they go with 50€ above the MSRP (< sure) max, 100 is really high for just "not throttling" cooling.
now if they can just get the BIOS situation sorted, I'll be glad to nab one myself
Vega owner here: The BIOS isn't the biggest problem for me; what really goes on my nerves is that the Radeon Settings program (just like Ryzen master btw) has a tendency of crashing. They both really don't like the sleep mode btw.
AnalKarl der vierte How does your Vega run with the new Adrenalin drivers? I just picked up the Vega 64 and was wondering if it got any better
Holy shit buildzoid!
I wonder if GN asks Buildzoid to keep these VRM reviews quick. Which for him is 45 mins or less :-P
Not really. We don't put in many requests for these! We try to target 20-30 minutes.
Not hearing Steve right at the start threw me off for a second there!
Is it just because of the cooler that the PCB is so large?
Can u please analyse sapphire rx 64 nitro + limited , to me it looks like it have best power delivery and cooling , it just wont go over 70c no mather what
This lad is amazing!
All that would fit to low profile-card? I think it would!
That would be nice to see in an ultra small form factor PC.
I love that extra efficiency for free comment; dude its a ASUS ROG Strix card its going to be a premium card and they are going to fleece the f**k out of us for it, they already bent us over and shafted us on all the other VEGA cards so people better lube up if they want one of these versions with the extra bells and whistles.
I'm no genius here but I'm not bad at math wouldn't an 18 phase VRM be possible? like 3 x 6 keeping that efficiency of using all 6 actual phases while being able to increase the load split?
Silver Studio theoretically you could use any number of phases. In pracise you're limited by what contollers are avalable. It's a bit like in ICE the more Zylinders you have the smover the pover delivery is.
no one makes triplers. You can only buy doublers and quads because doing 180 or 90 degree phase shifts is a lot simpler than 120 degree phase shifts.
It IS possible though... just... not practical, and Buildzoid was only talking in practicality. I mean... they'd have to order them special made, because as Buildzoid just said, no one makes them. The reason no one makes them is also valid. They really are more difficult. Both points would contribute to an astronomic price for very little gain.
Thanks for the explanation, I didn't know this before hand, sorry if it seemed like a stupid question.
No such thing as a stupid question. I'm sure others were wondering the same thing and were just afraid to ask. You asked, we were therefore able to educate not just you, but others by proxy.
The question is when ASUS is going to release this thing ?
killer hertz
So Asus has made a good PCB, but can it push a decent VBIOS and the actual cards now? It's october guys
Agost40Hz my guess is amd has to sign the bios ( buildzoid mentioned in one of his videos how bios flashing is limited to signed bios) and haven't gotten around to it, so asus can't release. That seems to be the case given that some of the reviewer sample strix vega cards actually ran reference bios. We'll probably have to wait till amd specific aib partners are ready before amd green lights this whole thing as to not throw those AMD specific aib partners under the bus.
On the topic of that IR3599, why would they cut their AC current in half by sending only the top half of it? Why not time it so that the top half of your wave goes to one place and your bottom half to another? Not to mention, where does that discarded current go?
So given nvme ports are in the vega spec, as manifested in vega pro ssg, is there any evidence of an nvme port on the vega 64 pcb - traces etc.?
Love these
i wish i knew enough about VRM to know if buildzoid knows his shit or is just playing madlibs
Why don't they make a separate 20 or 24 fase vcore supply , so we can use that on any (new) card we want ? that would keep ( our ) cost down , every time we buy a new GPU ,
And we can decide for ourselfs if we want a 12 fase , a 20 fase or a 24 fase
Just watching a video about Vega makes my system use more power.
0:20 Wow, how many double negatives were that?
Is there a breakdown of the Sapphire Vega Nitro limited?
Can you already release overclocking and power consumption deltas?
Someone at reddit said that you can't flash a 64 Bios onto the AIB 56 because of differing HBM brand (64 = Samsung whereas 56 = Hynix). But Buildzoid here says it's possible. Which is true?
what max HBM2 volt do you recommend? I wanna get 1125MHZ!
This video didn't appear in my notification.
Would it be feasible to liquid cool the vrms?
Soooooooo............ How many power supplies would I need to run V64 XFire x4?
Buildzoid did a stream on Wednesday the 18th with 2 Vega's in crossfire and OC'd and was pulling 1150W at the wall. He got like 250 FPS in Firestrike though, can't remember the settings.
Holy shit did it really pull that much?
Great Great Video for SURE... thanks a lot Dear.. i hope you can help me with this RADEON Vega 64... its not recognized by windows.. and running ATI flash also say NO ATI card Detected.. appreciate you effort
What about a 18 phase? 6*3
el power explosive de las placas de video
OK, so the VRM and PCB are a pretty good improvement but performance of the card is the exact same as the reference card... for the extra cost is the only thing we can get is marginal VRM and PCB improvements?
DoctorWho8675309 well, you will be getting a better cooler... But vega is pretty close to its limits, so you won't be getting extra performance out of it, if you discount micro throttling on the core
If thats the case it had better be priced the same as reference.
DoctorWho8675309 it's more expensive.
And your spot on, it offers nothing performance wise.
VBIOS doesn't presently work.
So is that what is holding this thing back? I'm really getting to the point of not caring what RTG's AIB's come out with because RTG doesn't deserve my money after this unfettered debacle of a release. For all the right Ryzen is... Vega is a complete 180. With the fact its still not right its safe to say AIB's won't be out until at least November if not December with the snails pace of Vega being delivered and then RTG keeping their hands clenched on the BIOS.
Does anybody know the hole distance of the gpu mounting holes? I want to install a watercooler on it... thanks in advance!!
But can it run ethminer? :)
Fukin miners man.
What is this, a crossover episode ?
We have been contracting BZ for about a year now.
Gamers Nexus
it's a meme :DD
i.redd.it/iruxd72qsv2z.jpg
300 Amperes lol
I'm studying electronics lately, and one thing that confuses me. If an entire computer can consume like, say, 500w from the wall (120v) that would mean that there's less than 5 amps going through the entire PC. How is it then that a GPU which is just one part of a PC be rated in hundreds of amps of current? Isn't a wall-socket only rated or like a max of 20A?
TroidHunter because 120(240)v is only to a psu, and psu then steps down to 12v, 5v, 3.3v??? And gpu uses 12v.....
and the GPU convers 12V down to like 1.2V
But if a GPU like vega needs 400A to sustain overclocks where do those amps come from? not the wall, certainly, those are capped at like 20A.
look up how DC to DC buck converters work
Remember the equation Watts = Volts * Amps, if the voltage is stepped down to say, 12 volts, then the amperage must go up to maintain the same wattage. In the case of 500 Watts, nearly 42 amps may be pushed over anything 12 volts. However, if you step up to 120 volts, then 500 watts only allows ~4.2 amps.
On fury HBM voltage wos 1.5V no 1.4 it wos lower in hynix specs ? or meaby im wrong
Fury's ran 1.3V or 1.35V. Can't rember now but defo not 1.5V because I degraded one of my cards with 1.42V.
HBM 1 vs HBM2...
Whats the standard Voltage for GDDR5??
1.55-1.65V on most GPUs. GDDR5X is 1.35V. HBM2 is 1.25v or 1.35V depending on if it's on a V56 or V64.
hello gamer nexus I use a very effective solution to effectively dissipate heat in audio amplifiers and voltage regulators that work with extreme temperatures such as GPUs, using common varnish and current, that if I sanded the entire surface of the heatsink and remove all clues of thermal paste, I have a hd 5870 for 7 years using this system and I use it in processors also try it and tell me the results
Did all the memory fall off in shipping? :)
Edit: In case you don't understand how to spot sarcasm, a :) usually means the post is sarcasm. This was sarcasm. Lol
ricky bobby lol no
See the answers to Lawrence Timmes question.
ricky bobby Trolling or really don't know how Vega is produced?
Vega uses HBM2 memory. It's part of the die, two stacks of 4GB.
Rondal Childers the sarcasm flew over your head :) it was a joke
there's no 'I' in asus
Wow, all that empty space. Is HBM2 really that much more dense than GDDR5?
Lul, nice crossover
Yeah but can it play Roblox?
why al this empty space in pcb..should they make the pcb smaller ... it's gonna be cool and compact
Cooler is slightly longer than PCB even with this conf.
BUT CAN IT RUN CRYSIS?
No, but it can handle 30fps in Pac-man on low settings.
yello
They forget to put memory round the gpu?
Cost savings
The AMD Vega cards use HBM (High Bandwidth Memory), so it is actually located on top of an interposer that also hosts the GPU die.
o ok lol that would explain that then. i never knew :P
ahahahaah :D
Your comment was my exact thought too the first time I'd seen vega uncovered :p
Coming from R9-290x, this sight is quite different :p especially concerning non-fullcover/universal waterblock options this really appeals to me!
u said asus wrong
I don’t understand anything!
Look 40% of the PCB is empty space. Feels half baked.
that because Vega uses HBM2 so the space normally taken up by memory is empty
I do actually really like that the PCB is how it is.
mike fox 40% empty with the same performance as a 1080. What's your point?
In before Vega Nano
Would you rather it came out half a year later and was smaller?