[SemiAccurate] Tesla K20 specs: 13 SMX, GeForce probably 12-13 SMX

[ocre]

I think the reason AMD doesn't get the margins is because they are often a follower, not a leader. Nvidia was getting booted out of the chipset business, all that R&D money being flushed down the toilet. So what do they do? They build a GPGPU market.

The computing world is going mobile with ultra lower power parts, not all out performance. Nvidia creates Tegra. AMD says, oh yea, we can make something too... only after ousting Dirk do they head that way.

Usually the smaller company should be more nimble than the giants, but it seems like Intel was able to turn on a dime with the Pentium 4 while AMD kept trucking along with the very mediocore Phenom I and Bulldozer.

To me it seems like AMD doesn't have the margins because they never have the foresight to see an upcoming opportunity. They're always following after a competitor creates the market.

As you said, you have to look towards management for that.

this is how i feel. I think AMD is wasting away chasing behind instead of blasting ahead. Keysplayer, AMD dumping tons of cash trying to catch up with nvidia supercomputers is exactly what they should not do!! Nvidia maybe having mediocre success, sure but its been a long road thus far. Intel is already showing their interest and they got billions to burn away. I think AMD investing in this direction is a terrible idea.

I think AMD has many valuable technologies and no direction. The days have having great hardware and it selling itself look to be behind us. We are in a new era. AMD needs to be creative and carve their own way. To become the next big thing they cannot follow. They cannot depend on others to do something great with their hardware. If AMD wants to be great, it is up to them. I believe that these times are different and that AMD has a unique position. Instead of designing chips, become the application.

 

[3DVagabond]

this is how i feel. I think AMD is wasting away chasing behind instead of blasting ahead. Keysplayer, AMD dumping tons of cash trying to catch up with nvidia supercomputers is exactly what they should not do!! Nvidia maybe having mediocre success, sure but its been a long road thus far. Intel is already showing their interest and they got billions to burn away. I think AMD investing in this direction is a terrible idea.

I think AMD has many valuable technologies and no direction. The days have having great hardware and it selling itself look to be behind us. We are in a new era. AMD needs to be creative and carve their own way. To become the next big thing they cannot follow. They cannot depend on others to do something great with their hardware. If AMD wants to be great, it is up to them. I believe that these times are different and that AMD has a unique position. Instead of designing chips, become the application.

Yes, AMD needs to sell the "AMD experience". For example, people wanting to buy an nVidia card ask, "I need a new card. What's the best nVidia card for $XXX." Like Toyota (The automobile, not our illustrious forum member), people buy a Toyota first and decide which model they want second. That never happens with AMD. AMD needs to package themselves and sell themselves.

 

[Ajay]

Charlie purposefully left out that K20 at ORNL is 14 SMX, his other stuff is pure speculation to make Nvidia look bad. Nothing new in the west.

Yes, and, being in a super-computer, those 14 SMXs must perform flawlessly 24x7 for years! These are the pick of the litter and probably stress tested b/4 delivery. The consumer cards need not support the same duty cycle and level of reliability, so it will be easier for NV to use more SMXs or higher clocks. Duh.... (to Charlie, not you)

 

[boxleitnerb]

He is pointing out that those K20 may be specially binned parts only for ORNL and normal K20 get only 13 SMX. But if yields were so bad, how would Nvidia have been able to produce so many of them in such a short time?
Afaik production started in September, tapeout was March.

 

[Ajay]

He is pointing out that those K20 may be specially binned parts only for ORNL and normal K20 get only 13 SMX. But if yields were so bad, how would Nvidia have been able to produce so many of them in such a short time?
Afaik production started in September, tapeout was March.

They may have been binned, we don't know how many wafers TSMC ran for NV. Still, Supercomputer gfx > Workstation gfx > Desktop gfx in from the standpoint of duty cycle and absolute reliability. Which means a 14 or even 15 SMX GF 7xx @ 800 MHz or whatever is much easier to hit for a consumer AIB than for the other two categories.

 

[Keysplayr]

this is how i feel. I think AMD is wasting away chasing behind instead of blasting ahead. Keysplayer, AMD dumping tons of cash trying to catch up with nvidia supercomputers is exactly what they should not do!! Nvidia maybe having mediocre success, sure but its been a long road thus far. Intel is already showing their interest and they got billions to burn away. I think AMD investing in this direction is a terrible idea.

I think AMD has many valuable technologies and no direction. The days have having great hardware and it selling itself look to be behind us. We are in a new era. AMD needs to be creative and carve their own way. To become the next big thing they cannot follow. They cannot depend on others to do something great with their hardware. If AMD wants to be great, it is up to them. I believe that these times are different and that AMD has a unique position. Instead of designing chips, become the application.

Ocre, I was talking in the past, when AMD had a chance in hell of doing something worthy. That time has passed. I really didn't mean now.

 

[BoFox]

this is how i feel. I think AMD is wasting away chasing behind instead of blasting ahead. Keysplayer, AMD dumping tons of cash trying to catch up with nvidia supercomputers is exactly what they should not do!! Nvidia maybe having mediocre success, sure but its been a long road thus far. Intel is already showing their interest and they got billions to burn away. I think AMD investing in this direction is a terrible idea.

I think AMD has many valuable technologies and no direction. The days have having great hardware and it selling itself look to be behind us. We are in a new era. AMD needs to be creative and carve their own way. To become the next big thing they cannot follow. They cannot depend on others to do something great with their hardware. If AMD wants to be great, it is up to them. I believe that these times are different and that AMD has a unique position. Instead of designing chips, become the application.

No - no, AMD definitely needs to compete!!!

They do have an impressive architecture that can churn out so many GFLOPs. Look at Bitcoin mining, look at password decrypting, etc.. If their engineers can keep up the up-and-up work like with what they just did with the Never Settle drivers, and with the expanding developer-relations project (Gaming Evolved, like NV's TWIMTBP) - they do have some potential especially if the new Sea Islands GCN2 arch is impressively better with further optimizations (say, 5-15% due to arch optimization). Each supercomputer has different needs, and AMD might just fit a specific need other than Bitcoin minting or pwd decrypting.

Cloud gaming is a big future, so AMD does have big potential with the cloud projects at least.

Yes, AMD needs to sell the "AMD experience". For example, people wanting to buy an nVidia card ask, "I need a new card. What's the best nVidia card for $XXX." Like Toyota (The automobile, not our illustrious forum member), people buy a Toyota first and decide which model they want second. That never happens with AMD. AMD needs to package themselves and sell themselves.

It's not actually that bad - some gamers do put AMD before NV with their buying decisions. Yet, I agree with you that AMD needs to work harder on that else they go bankrupt - which is definitely not what you guys want.

 

[nforce4max]

Personally I think AMD is finished, not that it doesn't make decent and competitive products but the VPs with in the company are leaches slowly sucking the company dry. The lack of leadership and the fact that many in management seam to lack a spine so the company sinks further debt while sales plummet.

 

[ShintaiDK]

No - no, AMD definitely needs to compete!!!

They do have an impressive architecture that can churn out so many GFLOPs. Look at Bitcoin mining, look at password decrypting, etc.. If their engineers can keep up the up-and-up work like with what they just did with the Never Settle drivers, and with the expanding developer-relations project (Gaming Evolved, like NV's TWIMTBP) - they do have some potential especially if the new Sea Islands GCN2 arch is impressively better with further optimizations (say, 5-15% due to arch optimization). Each supercomputer has different needs, and AMD might just fit a specific need other than Bitcoin minting or pwd decrypting.

Cloud gaming is a big future, so AMD does have big potential with the cloud projects at least.



It's not actually that bad - some gamers do put AMD before NV with their buying decisions. Yet, I agree with you that AMD needs to work harder on that else they go bankrupt - which is definitely not what you guys want.

Cloud gaming is dead. Just look at OnLive.

And for bitcoins and passwords, its mainly all due to a single instruction BIT_ALIGN_INT. Nothing hinders for example nVidia to implement this. The commercial value is just very low.

Also bitcoin mining only works when your power is very cheap. Else its just another expense.

The HPC race is already over for AMD.

AMD being busy now with evolved gaming is the path to go. Unfortunately they start behind so to say, and thats something that needs to be worked on. And in such a case, spreading out the already very lacking resources to HPC or anywhere else just means all will fail.

 

[tviceman]

Cloud gaming is dead. Just look at OnLive.

Cloud gaming isn't dead. It just isn't viable yet. As google rolls out it's gigabit internet in KC, and other telecom companies follow suit in the future, it will expand opportunities. Sony is rumored to be moving to cloud-based emulation to provides backwards support for PS3 titles with the PS4. If Valve ever tries to crack the console market, they'll likely need some kind of cloud-based compatibility for current steam users to port all their existing titles over.

Of course, we could be looking at a case where by the time the bandwidth exists in the U.S. and across the rest of the world to have a good cloud gaming experience, mobile devices will be powerful enough to act as consoles and will also have commonplace necessary capabilities for television set broadcasts.

AMD being busy now with evolved gaming is the path to go.

It does seem like AMD has refocused it's efforts on gaming, to a point much more noticeable than Nvidia lately. There only seemed to be a small handful of AAA TWIMTBP games in the past year, whereas GE titles are coming out every month now (it seems). Is it a sign that AMD is committing more than Nvidia and/or that Nvidia is shifting resources away from it's TWIMTBP program? 2013 should be an interesting year!

 

[ShintaiDK]

Cloud gaming isn't dead. It just isn't viable yet. As google rolls out it's gigabit internet in KC, and other telecom companies follow suit in the future, it will expand opportunities. Sony is rumored to be moving to cloud-based emulation to provides backwards support for PS3 titles with the PS4. If Valve ever tries to crack the console market, they'll likely need some kind of cloud-based compatibility for current steam users to port all their existing titles over.

Of course, we could be looking at a case where by the time the bandwidth exists in the U.S. and across the rest of the world to have a good cloud gaming experience, mobile devices will be powerful enough to act as consoles and will also have commonplace necessary capabilities for television set broadcasts.



It does seem like AMD has refocused it's efforts on gaming, to a point much more noticeable than Nvidia lately. There only seemed to be a small handful of AAA TWIMTBP games in the past year, whereas GE titles are coming out every month now (it seems). Is it a sign that AMD is committing more than Nvidia and/or that Nvidia is shifting resources away from it's TWIMTBP program? 2013 should be an interesting year!

Main reason is latency. And latency you cant fix.

nVidia said it was on purpose. Due to PhysX 3.0 coming.

 

[Dribble]

Cloud gaming isn't done yet - nvidia is investing a lot in this technology - although admittedly that's more then just gaming but cloud gpu compute (which includes gaming as one thing you could do).

Whether amd should be getting into this market I don't know - clearly if they wait it'll be too late (nvidia already has a big headstart) but who knows if it'll take off.

 

[Elfear]

Yet, I agree with you that AMD needs to work harder on that else they go bankrupt - which is definitely not what you guys want.

Wait? You do?

IMO we need both companies so that we don't get $600 graphics cards and 15% increases each gen.

 

[Ajay]

Cloud gaming is dead. Just look at OnLive.

And for bitcoins and passwords, its mainly all due to a single instruction BIT_ALIGN_INT. Nothing hinders for example nVidia to implement this. The commercial value is just very low.

Also bitcoin mining only works when your power is very cheap. Else its just another expense.

The HPC race is already over for AMD.

AMD being busy now with evolved gaming is the path to go. Unfortunately they start behind so to say, and thats something that needs to be worked on. And in such a case, spreading out the already very lacking resources to HPC or anywhere else just means all will fail.

This is still a large and profitable workstation market that AMD/ATI needs to grab more of. They had a larger share at one point, but have been losing out lately. Gaining back share there would help enrich the graphic unit's profits a fair bit.

 

[Stoneburner]

If Charlie said the sun was going to come up tomorrow, I would be getting my wife and kids together to say goodbye.

But your wife and kids would be very confused and probably have you institutionalized. ANd then they would walk out of the asylum crying while you are screaming, "CHARLIE!" at the top of your lungs.

Don't let him do this to you.

 

[tviceman]

Nvidia's quarterly earnings announcement and Q&A is tomorrow. Hopefully, if we're lucky, one of the investors will ask some questions regarding GK110 and if/when it is coming to PC consumer space.

 

[BoFox]

Wait? You do?

No.

Main reason is latency. And latency you cant fix.

nVidia said it was on purpose. Due to PhysX 3.0 coming.

Nvidia talked about the latency issues hardly being any worse than what Xbox360 users experience in general:

http://www.nvidia.com/object/cloud-gaming-benefits.html

 

[tviceman]

I have a hard time believing latency on consoles is that bad.

 

[BoFox]

Yeah, probably not always 100ms for the console itself (although 65ms definitely being true for many HDTV's), but with KINECT, it's like a whopping 500-1000ms input lag!!! And yet the Kinect was like the best selling thing ever in the history of consoles.. I couldn't tolerate it at all myself!

EDIT- Perhaps memory served me wrong, but I thought NV was referencing to an Xbox360 when I read it months ago? Perhaps it's the Wii that they're referencing to, with the Wiimote having at the very least 100ms lag.

 

[ShintaiDK]

Nvidia talked about the latency issues hardly being any worse than what Xbox360 users experience in general:

http://www.nvidia.com/object/cloud-gaming-benefits.html

What had you expected nVidia to say? nVidia want to sell you cloud gaming just as much as AMD.

 

[BoFox]

What had you expected nVidia to say? nVidia want to sell you cloud gaming just as much as AMD.

From the article:
"That's why NVIDIA has focused on reducing game server latency throughout the pipeline, achieving an amazing 30ms reduction with the GeForce GRID Platform. The end result is that cloud gaming now feels like playing on a local console.

These improvements are the result of NVIDIA's fast and concurrent game capture APIs, on-chip H.264 hardware encoding, and working with cloud game infrastructure providers to place servers next to major cities."

And with a lag-free HDTV/monitor (like my 65" DLP), it'd at least be more responsive than my Wii that has the usual 150+ms input lag.

When I play a fast-paced FPS game through a public server hosting 32-64 players (with my 24" CRT with zero lag either), it's usually more responsive than a single-player game on any console.

If people spent billions of dollars on Kinects with 500+ms of lag, it's basically already a promising testament to the potential prospective market of cloud gaming with *some* lag. Simply upgrading the HDTV to a really fast SmartTV one would probably make it even "faster" for anybody with a <50ms ping connection.

 

[sontin]

I think it's now official:
Titan is using K20 with 14 SMX. nVidia's Rob Csongor (Investor Relations) said in the CC that Titan has over 50.000.000 Cuda Cores. With 18.688 K20 cards it makes 14 SMX.

 

[boxleitnerb]

Any news on GK110@Geforce in the CC?

 

[Ajay]

Any news on GK110@Geforce in the CC?

Only that they are still supply constrained on TSMC's 28nm node. Margins are much higher on the workstation cards, and NV is expecting growth there with the introduction of the K10 and K20. Probably not good news for a consumer card coming from GK110 dice.

 

[tviceman]

Any news on GK110@Geforce in the CC?

Nope.