Well, considering it's called the "Frontier Edition", and it's being released late June... it doesn't take much thinking to interpret that the consumer cards are coming at least a decent amount later.
Source for that blog post? It would make sense they want to wait for Computex to release info on the gaming cards.
Why hasn't the original post been updated to reflect the recent Vega news? Might not be gaming related card news, but then again neither is the Volta threads recently updated news. Not news worthy?
Well, considering it's called the "Frontier Edition", and it's being released late June... it doesn't take much thinking to interpret that the consumer cards are coming at least a decent amount later.
Well, considering it's called the "Frontier Edition", and it's being released late June... it doesn't take much thinking to interpret that the consumer cards are coming at least a decent amount later.
AMD in marketing clearly made a point for FP16 in gaming. There are uses for it, and potential massive performance gains, so I don't see them ditching packed math for a hypothetical consumer die.To go out on a limb and speculate a bit here: might AMD be copying the Nvidia playbook even more than we're seeing here? How? I'll get to that in a minute, but here's my reasoning:
Previously, AMD/ATI has been a "gaming first, everything else second, third, or whatever really, we don't care, how about some games?" type of company. Before the Radeon Pro push, the FirePro line saw little adoption compared to Quadro cards, despite being far cheaper. Nvidia has since the Kepler days been pushing "pro-ish" cards (Titans) first on new architectures, and recently relegated the truly pro-level cards to dedicated silicon - due to the use/need for DP compute and the like, which takes up silicon area with no gains for most users.
Now, we see something similar from AMD: A Radeon Pro card being the first from a new architecture, with a somewhat inexplicably large die size (despite having 2x the SPs of an RX470, the die is significantly larger than 2x the 480/470 die - and that's without accounting for the die area savings due to HBM vs. GDDR5). AFAIK, the DP compute rate of Vega hasn't been disclosed, but I'd gather the extra hardware required for packed math does represent at least some of the die size increase, and it would be logical for a pro-level card to target something better than 1:32 DP compute rate - which would explain even more of the large die.
There used to be rumors of two Vega dice, yet in recent months, all we've been hearing of is the one presented to us yesterday in the Vega FE. Would it be entirely crazy to think the other Vega die might be a consumer-focused die, ditching packed math and whatever increased DP compute rate the VFE has, and using the freed-up die space and increased power efficiency to put in more CUs? After all, packed math has near zero value in the consumer space, as does DP compute. If they could push power efficiency to a point where we could have a 5-6000 SP Vega gaming card at tolerable power consumption at <600mm2, that would trounce anything Nvidia has to offer, at least if they can push it to 1500MHz or thereabouts. It would also make sense to launch it later, as a consumer focused "response" to Volta, while in the meantime selling partially disabled Vega ... 10? whatever the VFE has cores as a lower-end (and still reasonably competitive) stopgap until then. After all, it looks like even the VFE is reasonably competitive for gaming.
Of course, again, this is entirely unfounded speculation. But some times, dreaming is fun.
So you have no idea and are just guessing. Frontier means they are going into new areas -> Machine Learning.
As for consumers, while this is as big a change for them as it is for AMD, it’s likely a hallmark of what to expect in the future for new high-end GPU launches. For AMD gamers who have been holding out for Vega, it’s clear that they’ll have to hold out a bit longer. AMD is developing traditional consumer gaming cards as well, but by asking gamers to hold off a little while longer when the Vega FE already isn’t launching until late June, AMD is signaling that we shouldn’t be expecting consumer cards until the second half of the year.
http://pro.radeon.com/en-us/vega-frontier-edition/Source for that blog post? It would make sense they want to wait for Computex to release info on the gaming cards.
The Radeon Vega Frontier Edition graphics card is going to empower the pioneers creating the next generation of gaming experiences, but it does beg one question: Can you game on a Radeon Vega Frontier Edition? The answer is yes, absolutely. But because this graphics card is optimized for professional use cases (and priced accordingly), if gaming is your primary reason for buying a GPU, I’d suggest waiting just a little while longer for the lower-priced, gaming-optimized Radeon RX Vega graphics card. You’ll be glad you did.
The Radeon Vega Frontier Edition graphics card is going to empower the pioneers creating the next generation of gaming experiences, but it does beg one question: Can you game on a Radeon Vega Frontier Edition? The answer is yes, absolutely. But because this graphics card is optimized for professional use cases (and priced accordingly), if gaming is your primary reason for buying a GPU, I’d suggest waiting just a little while longer for the lower-priced, gaming-optimized Radeon RX Vega graphics card. You’ll be glad you did.
Where do you see that I contradict 12.5 TFLOPs/1525 MHz?
There is nothing dead in 1.2 GHz RX Vega GPU rumors.
When was the last time that a professional graphics card was released with a higher clock rate than the corresponding consumer card? It never happens. Professional cards can't be clocked at the bleeding edge because they need to be 100% reliable.
It's possible that we will see a cut-down or low-power version of Vega with a 1.2 GHz clock rate. But we know from AMD's public announcement that the silicon can do >1500 MHz, so there's no way the top gaming Vega SKU will be any less than that. AMD usually pushes their gaming cards as hard as they can. Somewhere in the 1600+ MHz range for the flagship gaming card is more likely.
That's a good point.One of the keys points to keep in mind is Radeon Instinct GPUs are passively cooled by the chassis fans found in the servers. So obviously they will be clocked lower than the Vega FE which has a blower type cooler. Moreover 4-8 Radeon Instinct GPUs will be packed in a single Naples server so that makes the TDP requirements even more stringent on the Radeon Instinct. Vega FE will be sold in workstations in a single or dual GPU configuration. Given these clock speeds for Vega Instinct and Vega FE I think we are going to see RX Vega clock closer to 1700 Mhz.
Lots of good points, last several posts.
Only reminds me that yesterday was day 1 of "another 2-3 months" before we can reasonably expect to see RX Vega... eg.: day 1 of another 2-3 months of forum speculating and bickering back and forth with scant hard data to bicker about.
It's doubly tough for AMD. Not only did they have to rebuild their CPU architecture from scratch and their GPU architecture mostly from scratch but they are also rebuilding the business at the same time.RX Vega will probably launch sometime in July. so yeah we are looking of 2 more months of wait. AMD has a problem going forward because Nvidia is moving at a furious pace with their GPU generations which are now 18-24 months apart and use intermediate nodes like 12FFN. AMD do not have the resources to launch entire GPU stacks between major nodes like 14nm and 7nm . The key is going to be AMD's Navi generation at 7nm. If AMD can have a competitive GPU architecture and launch a full stack of GPUs based on 3 chips like they did with original GCN based HD 7000 series then they might be able to slowly get back to a sustainable long term market share of 35-40%. 2018 will be difficult for AMD against Volta with GDDR6. We could see a repeat of the Maxwell generation in terms of Nvidia getting to 80+% market share.
RX Vega will probably launch sometime in July. so yeah we are looking of 2 more months of wait. AMD has a problem going forward because Nvidia is moving at a furious pace with their GPU generations which are now 18-24 months apart and use intermediate nodes like 12FFN. AMD do not have the resources to launch entire GPU stacks between major nodes like 14nm and 7nm . The key is going to be AMD's Navi generation at 7nm. If AMD can have a competitive GPU architecture and launch a full stack of GPUs based on 3 chips like they did with original GCN based HD 7000 series then they might be able to slowly get back to a sustainable long term market share of 35-40%. 2018 will be difficult for AMD against Volta with GDDR6. We could see a repeat of the Maxwell generation in terms of Nvidia getting to 80+% market share.
It's doubly tough for AMD. Not only did they have to rebuild their CPU architecture from scratch and their GPU architecture mostly from scratch but they are also rebuilding the business at the same time.
Oh and using the arguably inferior GloFlo.
Sigh.Jesus more than year later than pascal and probably only with crap reference cooler and another 2 months for good aftermarket cards...
and older pics of the cards had 8+6pin power?
The card Raja held up on stage was only 8+6 as some people pointed out, I went back and got a screenshot of it:
They're still going to be using Vega in 2018. It was in the presentation yesterday (their GPU stack in 2018 did not include Navi) and on the leaked slides in the OP of this thread.but Navi is 7nm and still scheduled for mid-late 2018, no? From what it looks like, nVidia is planning to ride through 2018 with Volta, and probably into mid 2019? (consumer VOlta can only be Q1 2018, I imagine). So at least AMD will have a solid leap with 7nm while nVidia is stuck on 12nm for about a year...or is "Volta 2" or whatever scheduled to have a 7nm die out by end of 2018?