AMD Commercial?

[JerseyDevil]

I understand AMD is going for the Jugular of Intel with a Commercial yes a Commercial during the superbowl. Its not one of those crappy gaming commercials they did a while ago about 64 bit computing. Its will only run during the superbowl. They cant afford to do commercials all the time like Intel so they strategically are doing a one Shot do or die attack to get consumer awareness the one time a year when everyone focuses on the commercials.

They will post server and gaming benchmarks claiming the worlds fastest home and buisness computers use AMD. If its not AMD processors its Obsolete and you paid too much. Sun may finance a portion of the commercials with performance ratios of SUN based Opterons against intel Xeons.

This would be huge.

 

[SonicIce]

It's not like AMD doesn't advertise at all. They have ads in magazines and stuff. BTW, how do you know all this stuff?

 

[DrMrLordX]

I remember AMD had an ad during the super bowl a few years ago. It was some silly ad in which two computers were being used to route a train on its tracks or something? Anyway, the AMD-based machine was fast enough to respond to the out-of-control train and get it onto a safe track, while the slower computer was too slow, and it crashed into the really annoying announcer.

Or, at least, I think that's how it went.

 

[effee]

Lol..Making a AD during the superbowl will cost millions

 

[imported_michaelpatrick33]

Originally posted by: effee
Lol..Making a AD during the superbowl will cost millions

Yet it will reach 50 million people (or some such incredibly large amount of people). One shot for a lot of awareness. Of course if they show it after half-time most of us will be too blitzed to notice

 

[JerseyDevil]

We had an AMD rep at the office plugging away AMD. IBM tagged along trying to get us to use IBM/AMD servers. Were mostly a HP Cough Cough Crap Service shop. Lousy presentation IBM should update their charts they quickly passed by many outdated and old Opteron graphics. IBM really needs to do more homework and quit relying on their name to get them in the door. One thing is for sure HP should be kicked out the door. They are so incompetent and have such lousy support now I could get better answers from akid working at CompUSA or even worse BestBuy than these guys.

So one of the questions was how is AMD getting the word around and its mainly word of mouth and presentations however superbowl sunday they may plunk down a few million to spread awareness of AMD processors. Also since the Superbowl ads sit on websites after the superbowl for more people to check out. When the AMD rep said the SUN word IBM eyes quickly perked up and nothing more was said.

As for the 250FSB comment I have flying around one tech said something about intel going to a 1066fsb which he felt would increase Intel performance. The AMD rep made a comment that they arent the only ones raising the FSB. Its been done in the past with AMD too. So we are indeed assuming a 250FSB is comming. This also makes sense that AMD will prevent overclockers in a way because they could lock own the multiplier lower and this would allow for less FSB overclocking room.

Now for the 110nm. Someone commented on speed of the CPU not being able to clock very high on 90nm having the same problems as intel and how they thought they could stay competitive in a speed war. The AMD rep made the comment that sometimes you need to take smaller steps instead of taking big steps in order to find the perfect performance ratio. Moving a process thats 40% 90nm or even 65nm is the future but sometimes going with a small die shrink is a better solition that taking such a big step. Obviously no one is ready for such a big step. When one tech said 110nm the AMD rep just smiled and winked. So he was either gay or hinting that Opterons might get a run on 110nm.

 

[BW86]

awesome

 

[ALIEN3001]

Now for the 110nm. Someone commented on speed of the CPU not being able to clock very high on 90nm having the same problems as intel and how they thought they could stay competitive in a speed war. The AMD rep made the comment that sometimes you need to take smaller steps instead of taking big steps in order to find the perfect performance ratio. Moving a process thats 40% 90nm or even 65nm is the future but sometimes going with a small die shrink is a better solition that taking such a big step. Obviously no one is ready for such a big step. When one tech said 110nm the AMD rep just smiled and winked. So he was either gay or hinting that Opterons might get a run on 110nm.

AMD spent months/years on preaparing 90nm process for their CPUs, now they'll just throw it away, and concentrate for a few more months on producing 110nm, PLUS you'd have to have THREE separate production lines - one for 130nm, one for 110nm, one for 90nm. That costs a lot of money.
Secondly - there is no reason at all to increase the "FSB" (FSB = HTT with K8), since the HTT link is running at 1GHZ (200x5), and is capable of 8GB/s, while dual channel DDR400 (PC3200) memory can only put out 6.4GB/s max.
Maybe there would be some difference with Opterons, since they also communicate with each-other via HTT links, but going from 800MHz to 1GHz didn't increase speed, why would going to 1250 change something in the near future?
DDR2 and faster HTT -> that would help with DDR2 800 (PC6400), it wouldn't make much of a difference, but anyway. But DDR2 + AMD is luckily far away.

 

[imported_michaelpatrick33]

Originally posted by: ALIEN3001

Now for the 110nm. Someone commented on speed of the CPU not being able to clock very high on 90nm having the same problems as intel and how they thought they could stay competitive in a speed war. The AMD rep made the comment that sometimes you need to take smaller steps instead of taking big steps in order to find the perfect performance ratio. Moving a process thats 40% 90nm or even 65nm is the future but sometimes going with a small die shrink is a better solition that taking such a big step. Obviously no one is ready for such a big step. When one tech said 110nm the AMD rep just smiled and winked. So he was either gay or hinting that Opterons might get a run on 110nm.

AMD spent months/years on preaparing 90nm process for their CPUs, now they'll just throw it away, and concentrate for a few more months on producing 110nm, PLUS you'd have to have THREE separate production lines - one for 130nm, one for 110nm, one for 90nm. That costs a lot of money.
Secondly - there is no reason at all to increase the "FSB" (FSB = HTT with K8), since the HTT link is running at 1GHZ (200x5), and is capable of 8GB/s, while dual channel DDR400 (PC3200) memory can only put out 6.4GB/s max.
Maybe there would be some difference with Opterons, since they also communicate with each-other via HTT links, but going from 800MHz to 1GHz didn't increase speed, why would going to 1250 change something in the near future?
DDR2 and faster HTT -> that would help with DDR2 800 (PC6400), it wouldn't make much of a difference, but anyway. But DDR2 + AMD is luckily far away.

The HTT would obviously change to a 250 x 4 multiplier (still 1000) at 1:1 with 250DDR1 memory. Remember the on-die memory controller runs off the HTT so increasing the frequency and lowering the multiplier will give you increased memory performance. This is what overclockers do to get increased performance. It is finding the balance of latency and frequency on the AMD64's for they are most sensitive to latency. The probem is that the JDEC standard for DDR1 memory stops at DDR200 and I doubt AMD would officially move its memory to a non-standard frequency especially for Servers and Workstations. The reason there was no (little) improved performance on the 1000HT was because the memory frequency is still the same. On multi-cpu setups the extrea HTT will boost performance because of the increased frequency of the cpus talking to each other but the memory is still sitting at the 200 divider. Increasing the the HTT to 250 would increase the memory but lowering the multplier would still kepp it in the 1000HTT spec. Of course the increased frequency raises a host of other potential problems i.e. voltage control, heat issues, trace noise etc.

 

[JerseyDevil]

Maybe the FSB increase is for when AMD will move to DDR2? Since the chips beyond 4000+ are I believe 1st or 2nd quarter next year this might be the plan when DDR2 comes about.

Anyone do a speed comparison of an AMD 2Ghz at 200FSB and 250FSB? I will have to look and see if there is any benefit but it makes sense and AMD has increased the FSB before with Athlon XP.

PC4400 ram is readily available isnt that 550/2 = 275mhz. With memory scaling to smaller micron processes this still seems possible.

Im having a tough time rling this out becuase it is probable. Anand reached 290mhz of FSB but I will have to look at what speed his ram reached.

P4 is quad pumped but Opteron is dual correct?

 

[ALIEN3001]

Remember the on-die memory controller runs off the HTT

The memory controller runs on the same speed as the entire core (eg 2GHz for a 2GHz Athlon64).

 

[ALIEN3001]

Anyone do a speed comparison of an AMD 2Ghz at 200FSB and 250FSB? I will have to look and see if there is any benefit but it makes sense and AMD has increased the FSB before with Athlon XP.

AthlonXP's FSB is different than Athlon64's "FSB". Read about that, then post comments.
Higher HTT frequency wouldn't change anything, since an increase from 800MHz to 1GHz changed nothing, a further increase would not help much.
250x4 is stupid. AMD made a decision to go the 200x5 route, why change that?

 

[JerseyDevil]

Remember I was dealing with a rep not a engineer so he may mistake FSB with HTT. Ok so let me get this right.

If were talking HTT increases and we increase it from 200-250 since RAM is currently not fast enough there would be no benefit to the processor overall? Correct or no?

If were still talking HTT and DDR2 could there be a possible benefit?

A buddy of mine asked me if AMD could be planning to use RDRAM? But I think there might be some legal aspect to why they cant use RDRAM.

I dont think the AMD rep was talking out his but but that would be lame of them to suggest things even subtle things like this without having some merit?

Also think Server chips more than desktop chips.

Thoughts Comments?

 

[imported_michaelpatrick33]

Originally posted by: ALIEN3001

Remember the on-die memory controller runs off the HTT

The memory controller runs on the same speed as the entire core (eg 2GHz for a 2GHz Athlon64).

Sigh. Yes, the frequency regulation of the on-die memory controller runs at full processor speed but the actual dimms are obviously not running at 2Ghrtz. They are running at DDR1 200MGHRTZ. The on-die controller timing is controlled by the HTT clock which runs at 200 x 4 or 200 x 5 depending on chipset and cpu.

The on-die controller would still run at 2000Mghtrz (using your example) but the HTT frequency would then become 250 x 4 (HTT) instead of 200 x 5, with 250DDR1 on dual channel equaling DDR 8000 (DRR1 250) [instead of DDR1 200 equaling 6400 in dual channel] memory speed. Of course this would throw off the core frequency speed because 200 x 10 to get a 2000 core would have to change to 250 x 8 to get the same 2000 core. Of course a 3500+ at 200 x 10 would change to a 250 x 9 for 2250 total core. You see the problem. Lots more PR speeds on top of the billion we already have.

In other words the on-die memory controller runs at full proc speed to lower latencies and for scaling but it is limited still by the DDR1 200 memory dimms that it is controlling. HTT is 200 because DDR1 200 memory is used. If AMD raising the memory to 250DDR1 then they have to raise the HTT to 250 at 1:1 to match the timings of the memory/cpu interface. This would change the current HTT of (200 x 5) to (250 x 4) to keep the same HTT ceiling of 1000DDR (2000 in market speak). This then changes obviously the HTT determining the actual core frequency for the core frequency is generated off the HTT multiplier. So the 4200+ if operating at 250 HTT would have a frequency of 2625 if they use 1/2 multipliers and use 1meg L2 cache. See the problem? 1/2 divisers and strange frequencies would occur.

 

[SimsFreak]

I always thought more FSB, less multilier = good?

 

[JerseyDevil]

Great Explanation.

Would AMD change the internal multipliers to work at 250? I know thats a core change but would they?

1/2 multipliers would just be a bios update correct? For some reason I recall 1/2 multipliers before but again I have fallen out of this stuff thats why Im hoping to get my answers in here.

Thanks MichaelPatrick33 great info.

But looking at Anands article he was running fine at 290 so maybe nothing internal needs to change just updating motherboards and memory with the new?

I guess I am still a little lost to what AMD was hinting at then?

 

[imported_michaelpatrick33]

Originally posted by: SimsFreak
I always thought more FSB, less multilier = good?

AMD64's have no FSB but people still use that term to refer to the HTT frequency. With AMD64's it seems to work better with lower latencies than higher bandwidth so overclockers find the balance between the lowest timings and the fastest bandwidth.

For example. a HTT of 290 x 9 = 2610 at memory timings of 3-4-4-10 could be slower than a
HTT of 260 x 10 = 2600 at memory timings of 2-3-3-10

Does that help?

 

[imported_michaelpatrick33]

Originally posted by: JerseyDevil
Great Explanation.

Would AMD change the internal multipliers to work at 250? I know thats a core change but would they?

1/2 multipliers would just be a bios update correct? For some reason I recall 1/2 multipliers before but again I have fallen out of this stuff thats why Im hoping to get my answers in here.

Thanks MichaelPatrick33 great info.

But looking at Anands article he was running fine at 290 so maybe nothing internal needs to change just updating motherboards and memory with the new?

I guess I am still a little lost to what AMD was hinting at then?

The problem isn't with the overclockers it is with JDEC and memory standards. The official JDEC standard for DDR1 stopped at DDR200 so AMD would not be running with memory recognized by the industry standard governing body. I don't think AMD wants non-industry standard memory running in servers and workstations even though it would be better performing. This is where Intel's market dominance comes into play. Since their Prescott's need more bandwidth and timing isn't as important DDR2 memory became the standard with higher latencies which penalize the AMD more efficient lower latency cpu/memory interface

 

[JerseyDevil]

That does so 250 could be a sweet spot then?

What about 110nm being the sweet spot for the die shrink instead of 90nm having electron/current leakage problems? Could that be a sweet spot for AMD SOI? The way the AMD rep put it that 90nm will require something more than current SOI technology to fully utilize moors law but 110nm opterons might be the ticket. Its taken out of context but thats the way everyone felt he was implying.

 

[JerseyDevil]

But wasnt everyone using PC3200 and the term before JDEC finalized it anyhow? 802.11 standards were out before finalizing them as well. IF AMD sets the stage for the standard I dont think anyone will care what JDEC has to say. Its been overlooked before. I remember MODEM 56k fights too where the technology outpaced the standards but it still went to market.

 

[imported_michaelpatrick33]

Originally posted by: JerseyDevil
That does so 250 could be a sweet spot then?

What about 110nm being the sweet spot for the die shrink instead of 90nm having electron/current leakage problems? Could that be a sweet spot for AMD SOI? The way the AMD rep put it that 90nm will require something more than current SOI technology to fully utilize moors law but 110nm opterons might be the ticket. Its taken out of context but thats the way everyone felt he was implying.

He is pulling your leg. There will never be 110nm processor fabrication. It would cost millions upon millions and their roadmaps simply don't match what he is saying. The 110nm fab is for flash memory not processors. Amd is now converting to 90nm and will have both 90nm and 130nm chips coming out of their foundries until early next year and the 90nm will eventually replace all the 130nm chips. They get more chips per wafer on the 90nm and would have to be retarded to introduce a new 110nm process when they already have the 90nm processor running and expected to completely take over on the A64's by the middle of next year

 

[imported_michaelpatrick33]

I am not saying AMD won't do it but it would be surprising. DDR1 250 will never, I repeat never become a JDEC standard because DDR2 at 266 is already a standard. AMD has also stated they will move to DDR2 eventually so the JDEC doesn't see the need to give another DDR1 validation.

Off to work great conversation!

 

[JerseyDevil]

Thanks Michael I believe you and that the rep may have been just trying to stir up more optimism. Too much ambiguity behind the pitch letting us build up our own ideas to what they might mean.

In AMD's favor.
In a game of High Stakes Poker I think AMD may show Optimism but is not revealing their full hand so that Intel may be caught a bit off guard to what they may have up thier sleeve. They wont be Miracles by any means but I did get the impression AMD has some WOW factor they want to see if they can pull off before putting thier foot in their mouth.

Thanks everyone for the input I will share it with my team.

 

[imported_goku]

Originally posted by: JerseyDevil
the AMD rep just smiled and winked. So he was either gay or hinting that Opterons might get a run on 110nm.

Great stuff man! LOL

 

[Frew]

Maybe they will have "the green men group!"