He probably had a mockup sitting around for months, and figured that was good enough for his purposes.
- Status
He probably had a mockup sitting around for months, and figured that was good enough for his purposes.
FatherMurphy
Senior member
- Mar 27, 2014
- 229
- 18
- 81
Truth. And why give competitors a glimpse into a future GPU if you don't have to. Again, the demonstration was not at all about Pascal as a Geforce GPU.
EDIT: The glimpse being the size of the die and what memory it uses.
Last edited:
nVidia might have Pascal silicon kicking around either functional or not, but the inclusion of GM204 in the Drive PX2 doesn't really shed much light on that. They might have functional silicon in smaller test dies, they could have functional GP100 dies. They may or may not have GP106, or whatever the PX2 will use. They may or may not have that silicon mounted onto GPUs and running software. They may or may not have Pascal silicon mounted onto an MXM card already.
The PX2 uses a couple MXM cards for the GPUs. You can't really just FunTack the die down into the spot where the module would go. The fact that they don't have an MXM module finished and assembled for GP106 doesn't even tell us whether GP106 is taped out or out of the foundry and brought up yet, let alone whether any Pascal silicon is finished.
In regards to Charlie's point, he's partially right but his conclusions are a stretch. The Zaruba manifest doesn't indicate anything about working silicon and it's obvious that it's test and bringup hardware, as was mentioned several times here. You definitely would want to have all that stuff in place prior to having silicon arrive in the lab, so it's not improbable that the Zaruba manifest indicates that the 37.5mmx37.5mm BGA package will arrive after the test hardware. A 37.5mmx37.5mm BGA package is almost certainly not GP100 though, and is likely either GP104 or GP106.
tl;dr, we still have very little idea about the status of Pascal, either positive or negative. The only solid information is from ISSCC where the nVidia rep said Pascal was in the lab, which is at least a bit of a positive even if it gives no indication what is in the lab or when that will translate into a commercial product.
Mr Teal hit the nail on the head so many times that if anyone argues against his points they're in serious denial of reality.
His TL;DR summed it up nicely.
His TL;DR summed it up nicely.
Last edited:
You just made that up to fit your position. There is zero evidence of that. Not only that, it's far more likely that it was made just for that presentation. Then he lied about what's on it. So we can add premeditation to it.
He may have sped on the way to the conference, tack on that charge as well. Who knows, maybe he J-walked from the parking garage to the entrance. The audacity!
xthetenth
Golden Member
- Oct 14, 2014
- 1,800
- 529
- 106
Don't be disingenuous. Whether or not he holds a Pascal up when he says he is holding up Pascal has relevance when discussing whether there is a Pascal to be held up.
If there was he would of proudly showed it off. Gut says no working Pascal silicone yet, other than what I linked already.
Nvidia would have soon shown it after AMD''s Polaris demo. Always 1 up them is the trend....JHH''s blood is likely boiling.
They probably wasted too much time trying to make a functional HBM controller that didn't impede on AMD's patent(s).
Rig needs a dGPU so money goes to 1st with a compelling offering in my price range.
There seems to be a lot of history to suggest that they could easily have had working stuff but not shown it at that show.
Seems like a quite incredibly pointless thing to speculate about though We'll know in a few months.
Seems like a quite incredibly pointless thing to speculate about though
We'll know in a few months.
RampantAndroid
Diamond Member
- Jun 27, 2004
- 6,591
- 3
- 81
If there was he would of proudly showed it off. Gut says no working Pascal silicone yet, other than what I linked already.
Nvidia would have soon shown it after AMD''s Polaris demo. Always 1 up them is the trend....JHH''s blood is likely boiling.
They probably wasted too much time trying to make a functional HBM controller that didn't impede on AMD's patent(s).
Rig needs a dGPU so money goes to 1st with a compelling offering in my price range.
I worked on a cell phone at one point. We contracted the hardware out to another company who made it. When I finally saw one, it was in an obviously unfinished state, but I was astonished to find out the value of the device - it was well into the thousands. For a phone that would sell for an order of magnitude less. The reasons why were that it was one of the first made, it had some serious issues that had to be solved after the board itself was already made (yeah, some poor person overseas was resoldering SMDs all over the place.) Additionally, it had to go through importation into the US, and then some FCC stuff I believe since it wasn't an approved device. And that was just an early prototype.
tl;dr: I don't think that what you're saying has to be true. If they only have a handful of Pascal chips on their dev boards, those hold a pretty high value. And letting someone cart one off in that situation and potentially damage it by handling it just seems...well, silly.
Are people not reading the whole thread?
NVidia has working samples of Pascal. Which GPU(s) they have and how close they are to production is something we don't know yet.
Why they weren't shown recently is also something we don't know.
We'll have to wait for answers.
This isn't an AMD vs Nvidia argument. If anything it's a debate about S/A and the truth. Since the site calls themselves "SemiAccurate" anyway, they aren't making any claims about always being right. So - meh.
NVidia has working samples of Pascal. Which GPU(s) they have and how close they are to production is something we don't know yet.
Why they weren't shown recently is also something we don't know.
We'll have to wait for answers.
This isn't an AMD vs Nvidia argument. If anything it's a debate about S/A and the truth. Since the site calls themselves "SemiAccurate" anyway, they aren't making any claims about always being right. So - meh.
Too much speculation. When does Nvidia disclose quarterly financials? Keep in mind what's said during that time to get a better idea on how far along Pascal is from being released.
ShintaiDK
Lifer
- Apr 22, 2012
- 20,378
- 146
- 106
They won't be (never have been) specific on release dates. They will be vague (2H 2016, or later this year) or even more vague and will say "is on schedule." JHH likes to say XXX is ramping up, which in itself is also vague.
Last edited:
Neither of us has proof of either possibility, but that doesn't matter. He could have shown a doorknob, and it would matter just as much as showing a real video card. It was simply a marketing stunt. I do happen to know that many products we buy are fakes in their commercials. It doesn't matter unless he demonstrates it's use.
Last edited:
Which is worse? The defender? The accuser?
The accuser has proof.
The defender has excuses.
Leaning towards the defenders....Or is it deflectors.
We know the product he held in his hand was fake. I just don't really care, as I see it simply as a marketing stunt that we accept in many other examples all the time. Even if he could have had a real one available, he still may choose to use a mockup, as there simply is no reason he has to have the real one on hand if he believes no one would see a difference.
I also find it funny that someone would call the person a liar for doing so.
Last edited:
You find it funny that someone who lied in the past and got caught out, then lies again now and get caught out is called.... a liar?
Strange sense of humor.
ps. How about we test you with this, is Joe Macri @ AMD a liar for saying "Fury X is an overclocker's dream!"?
I'm sure he did, but he still lied.
Fury X is a very gimped overclocker, it didn't even have +vcore access for months after release. So everything is contrary to his statement, therefore, he lied.
Just as JHH lied when he claims what he's holding has 6X the compute deep learning throughput compared to Titan X!
Maybe Nvidia has Pascal silicon back in the labs as their scientist mentions at ISC back in Nov 2015. Maybe Nvidia is keeping the lid tight on Pascal development and does not want to give AMD any information. I do believe Nvidia will surely talk more about Pascal at GPU Technology Conference. http://www.gputechconf.com/ in Apr 2016. Typically thats their biggest event and when they talk about new architectures and upcoming products. I think Nvidia will have working demos by that time.
Yeah Joe Macri's statement was really silly given the fact that Nvidia was selling Maxwell GPUs which overclocked 200+ Mhz on stock voltage and did even better (300+ Mhz) with custom OC cards with voltage control.
Right from the HD 7970 to R9 290X to Fury X, AMD's transistor density kept on increasing. But that came at the cost of max clocks. AMD's architectural efficiency was not good and they ended up needing more transistors to compete but still lost. Thats something which never happened in the past. AMD GPUs were much smaller, used lesser transistors and would trail by 10-15%. For the first time AMD lost on all counts. Perf/transistor, Perf/sq mm and Perf/watt. To make it worse the GPUs had very less OC headroom. Fury X was 8.9 billion transistors and GTX Titan was 8 billion transistors and still lost decisively.
I am keen to see if AMD has corrected their design philosophy to focus on efficiency - perf/transistor, perf/sq mm, perf/watt. AMD also needs to do have a design which has good OC headroom.
From what I gather, leakage on 14nm finfet is much less, so the density is a bonus on all fronts.
This generation there are going to be interesting comparisons - Polaris 10 on 14 LPP vs Pascal on 16FF+. Polaris 11 with HBM2 might be on 16FF+ (due to better yields for larger dies on 16FF+ and TSMC's vast experience making large GPUs 300+ sq mm for more than a decade) and will go up against GP104.
- Status
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter