itsmydamnation
Veteran
well it has been confirmed that the resonance clock mesh didn't make it into piledriver, so maybe v2 has it and v2 is what the consoles will use?
The AMD Execution Thread [2007 - 2017]
- Thread starter overclocked_enthusiasm
- Start date
-
- Tags
- amd
- Status
well it has been confirmed that the resonance clock mesh didn't make it into piledriver, so maybe v2 has it and v2 is what the consoles will use?
AMD are in a much worse position than they were the last time they went below $2. That was just crazy back then, now it's real.
Nah, crazy is that the market as a whole is still ~40% over-valued. The only thing that has prevented another correction has been the vast printing of money, but even that is starting to lose its effect. However, that is another story...jimbo75 said:
Anyway, the value is definitely there, especially for MS, Google, or Apple.
Last edited by a moderator:
I'm not aware of any official confirmation in that direction (would be slightly embarrassing too considering they explicitly talked about it at Hotchips, and without much vagueness in terms of the part they had in mind, which was the Trinity embodiment of Piledriver). Also, the odds of any next gen console having any sort of Piledriver/BD derived thing inside are quite slim to say the least.
I believe it's in Trinity (the APU variants like the A10s) but not Vishera/Abu Dhabi (the FX and Opteron CPUs).
AMD as the source is somewhat broad. These are the same gentlemen that discovered they had ~1bln extra transistors in their counts, and who sometimes prove that not everyone is necessary very up to speed with the hardware they're currently selling, even within members of technical staff (not necessarily surprising I guess). I'd go with what they said at Hotchips, until stronger further notice comes. Not discarding the possibility, mind you, but I'm not that fond of conspiracy theories.
Or that resonant clock meshes are not without their own problems. The general feeling is that, while really interesting from an engineering and academic point of view, the benefits aren't necessarily worth the trouble.
On the other hand, if it even works a little and is present in an actual mass-market product, it's head and shoulders above a long line of half-baked initiatives or alternative technologies AMD failed to bring to fruition.
That's not to say other companies don't, but for some reason AMD deemed it fit to publicize more than its share of pointless pursuits.
Isotopically pure wafers, Z-RAM, Torrenza, probably its advanced synchronization facility, the Griffin core.
T-RAM?
Will using ASIC high-density libraries for Excavator will see the light of day?
Technically SSE5 did make it into a product, sort of to the detriment of the product it appeared in.
I'm sure I'm forgetting something.
edit:
One thing about RCM is that it seems they are supposed to be tuned to a target frequency. Can that tuning be adusted?
If not, it would seem by adding it, AMD would have admitted there is no unexplored clock speed frontier beyond it, and that BD was probably unsalvagable.
That's not to say other companies don't, but for some reason AMD deemed it fit to publicize more than its share of pointless pursuits.
Isotopically pure wafers, Z-RAM, Torrenza, probably its advanced synchronization facility, the Griffin core.
T-RAM?
Will using ASIC high-density libraries for Excavator will see the light of day?
Technically SSE5 did make it into a product, sort of to the detriment of the product it appeared in.
I'm sure I'm forgetting something.
edit:
One thing about RCM is that it seems they are supposed to be tuned to a target frequency. Can that tuning be adusted?
If not, it would seem by adding it, AMD would have admitted there is no unexplored clock speed frontier beyond it, and that BD was probably unsalvagable.
Far from being an expert on these things, the following things come to mind: relatively small power benefit even in the best case, increased design complexity, even smaller benefit for mistuned clock frequency, higher cost because an additional metal layer can't be used for regular routing/power distribution.
Judging from this diagram which was done by my colleague Marc, I'd think that RCM does not provide much benefit in the clock ranges for Trinity's desktop parts. That might be different for mobile though - maybe there's 3dilletante's point of tuning for a specific clock range/frequency coming into play which is lower for mobile parts where power consumption is much more important?
http://www.pcgameshardware.de/Core-...Test-FX-8320-FX-6300-FX-4300-Vishera-1032556/ [scrolling down quite a bit required]
http://www.pcgameshardware.de/scree...-8320-FX-6300-FX-4300-x-Power-Consumption.jpg
http://www.pcgameshardware.de/Core-...Test-FX-8320-FX-6300-FX-4300-Vishera-1032556/ [scrolling down quite a bit required]
http://www.pcgameshardware.de/scree...-8320-FX-6300-FX-4300-x-Power-Consumption.jpg
While this hardly justifies an "Intel Gloom and Doom Thread" (certainly not based on execution):
http://business.financialpost.com/2012/11/08/qualcomm-passes-intel-in-market-value/
Qualcomm just passed Intel in market cap. Intel does still have the best fabs and many interesting mobile oriented products in the pipeline, including its on-silicon analog antennas. I doubt there'd be too much of an issue w/ x86 in terms of software adoption, but I wonder how long Intel can hang onto a performance per watt advantage over ARM from just process advantages.
This has implications for AMD too since its new focus seems to be away from Intel's strengths. I hope its decision to use ARM is the right one.
http://business.financialpost.com/2012/11/08/qualcomm-passes-intel-in-market-value/
Qualcomm just passed Intel in market cap. Intel does still have the best fabs and many interesting mobile oriented products in the pipeline, including its on-silicon analog antennas. I doubt there'd be too much of an issue w/ x86 in terms of software adoption, but I wonder how long Intel can hang onto a performance per watt advantage over ARM from just process advantages.
This has implications for AMD too since its new focus seems to be away from Intel's strengths. I hope its decision to use ARM is the right one.
Last edited by a moderator:
IMHLO, ARM has been designed for power efficiency from it's very root, whereas we've yet to see a product that has been designed with power efficiency in mind from the beginning and - very importantly - that is meant to be competitive (in contrast to the rather mediocre Atom).
Does it even have one now?
Not entirely sure. Medfield seems like a competent product with comparable battery life to ARM products. It had superior performance until A15 class products hit.
c.f. http://www.anandtech.com/show/5365/intels-medfield-atom-z2460-arrive-for-smartphones
Last edited by a moderator:
Looks like it, and Intel did just fine from a performance / watt perspective using it. Among the things mentioned in that anandtech review is that most Android apps (75%) don't use native ARM code, and that the remaining native apps could run on an emulator.
Silvermont looks like a major micro-architectural improvement over the original Atom. Combined with the much better low power performance of its 22nm process, Intel looks like it will erase the gains made by A15 class products next year.
It'd be fun to jail break a phone like that to run some stripped down version of windows... hopefully there's a work around the missing PCI-E bus and drivers...
Silvermont looks like a major micro-architectural improvement over the original Atom. Combined with the much better low power performance of its 22nm process, Intel looks like it will erase the gains made by A15 class products next year.
It'd be fun to jail break a phone like that to run some stripped down version of windows... hopefully there's a work around the missing PCI-E bus and drivers...
- Status
