It's debatable whether or not the plant was an asset.
The AMD Execution Thread [2007 - 2017]
- Thread starter overclocked_enthusiasm
- Start date
-
- Tags
- amd
- Status
I object to your axes not starting at 0. That's right, axes.
silent_guy, reading down to the Q&A shows a lot of Qs trying to suss out how much console wins will affect that embedded figure. AMD's As were basically no comment until the consoles are released.
I object to your axes not starting at 0. That's right, axes.
silent_guy, reading down to the Q&A shows a lot of Qs trying to suss out how much console wins will affect that embedded figure. AMD's As were basically no comment until the consoles are released.
Will it help if the graph eventually ends at zero...?
Snark aside, the console revenue could push forward 5% to 20% share that the embedded products are meant to take up.
The exact arrangements would be interesting. If the two high-end consoles use x86 cores, the relationship between the console manufacturer and AMD would not be the same as what happened with the current gen, since AMD can't grant the same rights to the CPU cores.
That does pose the possibility of something higher than the money earned on Xenos, but with the responsibilities and costs invovled in having to make the chips themselves. The penalties for failure could be ruinous on top of the usual destructive effect of struggling projects at AMD, and if it gets into a Llano or Bulldozer scenario, the efforts to overcome systemic problems that are now on another powerful corporation's dime could make it a net loss.
Any console manufacturer using AMD's chips in this scenario may be uncomfortable with the latest turn of events, with an AMD that may not be around in a capacity to work on die shrink.
If there were an argument for not having an APU, it would be that at least a GPU could be licensed and shrunk separately, but an APU leaves two big swaths of silicon bound together with little opportunity to optimize costs.
I'm wondering at this point where some of the cuts may be, and if the depth of the cuts may have to do with maintaining the company or matching any covenants on its debt deals that could break the bank prior to actually running out of money--at least until the hopefully on-time console rollout.
Vishera reviews are out, and it seems to be finally at the same level roughly as the Phenom II X6 1100T or the X4 980 models. In cases where the code is highly multi-threaded and integer heavy, the top end model beats quadcore Core i5s and i7s much more often, though if it's due to the mild IPC enhancements, the 400 MHZ higher base clock or both isn't known yet.
For most applications, especially if they're lightly threaded, single threaded, or FPU heavy, you'll still want a quad core i5 or a quad/six core i7, of course.
Well, the six core i7s are still better than AMD's new stuff.
So, why am I bringing this up here? The internet in general seems to be reacting with a bit of hope as they didn't expect the new Visheras to be as "good" as they. This could lead to increased sales on top of the higher sales due to the lower price of the highest end version.
For most applications, especially if they're lightly threaded, single threaded, or FPU heavy, you'll still want a quad core i5 or a quad/six core i7, of course.
Well, the six core i7s are still better than AMD's new stuff.
So, why am I bringing this up here? The internet in general seems to be reacting with a bit of hope as they didn't expect the new Visheras to be as "good" as they. This could lead to increased sales on top of the higher sales due to the lower price of the highest end version.
According to Hardware.fr's tests, there's a ~10% IPC increase (at 4GHz). A little less in general applications, a little more in games, but that's more or less the average.
It's a really good CPU for a cheap workstation, provided that power isn't much of a concern. Still not really worth considering for a gaming rig, though.
hardware.fr themselves say 7.7% average below the table, and I'm going to assume they did the math correctly. Of course, the gains are all over the place. Pretty nice overall, but I was right to expect that the 17% numbers THG were promoting several months ago for Trinity wouldn't be an average or even typical.
Of course, despite being clock normalized and utilizing the same number of cores this doesn't strictly represent an IPC increase. It's possible that a higher IPC gain is offset by a worse module sharing penalty - that is, since the big front-end bottlenecks haven't really been improved (L1 icache aliasing, shared decode too narrow for two threads, etc) increasing utilization on the individual cores will make the sharing penalty more apparent. So it's possible that the average improvement for scenarios w/o module sharing is higher, assuming that the clock improvement isn't lower (someone will have to remind me if BD running 4M/4C can still only hit 3.6GHz)
The 11% base clock increase really helps sell things for the multithreaded loads they already did well on. I wonder how much base clock Intel could eke out if they increased TDP to 125W. My impression is that it's not that much, not for Ivy Bridge anyway :/
They say the average is 7.7% for "applications" (i.e. the red bars), but 13.5% in games (green bars); which is why I said 10%.
Bulldozer would be pretty decent if it was on an Intel process, wouldn't it?
Anyway I think the recent problems AMD has had are due to the fact that Intel has realised they don't need to keep AMD around any longer due to competition from the ARM sphere and the hypothetical gloves have indeed 'come off'.
Anyway I think the recent problems AMD has had are due to the fact that Intel has realised they don't need to keep AMD around any longer due to competition from the ARM sphere and the hypothetical gloves have indeed 'come off'.
Guess I missed that, thanks.. Should have been more thorough with the translation >_>
This is based on what? All intel has done is be 4-5 months late with Ivy Bridge, and even at that it's basically a 10% speed bump. AMD has closed the gap with Piledriver, which considering how much of a mess the company is in is pretty bizarre.
AMD's problems are all of their own making, no need to look elsewhere.
Nah, on an Intel process it would just suck less. Piledriver would be okay-ish.
In technical terms, Intel has never had gloves on, they just screwed up with NetBurst, but their execution has been pretty much flawless ever since.
However, the real gloves are actually still on, in the form of high prices. Intel could, if they wanted to, slash the prices of all products close to AMD's price range by 30% and basically smother AMD to death. But they would make less money, so anti-trust problems aside, they prefer not to.
Well speaking of gloves, isn't Intel selling "updates" that allow overclocking of some of their CPUs?
I may be wrong but I also believe that on core i3 the turbo is disable for no reason. Then there is the clock speed, conservative.
Overall, the competition being what it is, I wonder if Intel is not purposefully crippling its CPUs, for the sake of generating demand for the higher end products. For example Intel sells more core i5 than core i7, I would bet a lot more, does the fact that hyper threading is disable reflect yields or is in fact an artificial constrain to allow them to sell their higher margin core i7?
I would bet on the later.
I think it is the same with Pentium, they are artificially cripple core i3 (and yields have to be really good on that one...), it is better for Intel to cripple CPU than to fight with them-selves (by removing intensive for consumer to go with higher end parts).
They can do that because the competition is so weak it almost no longer put any pressure on them. If we were in the K8 era, core i3 would be clocked faster, with turbo and they could be overclocked from scratch. I think that there would be more difference between core i5 and core i7.
Intel is not selling "salvage parts" (say a tri core) I suspect htey yields are really high, the removal of SMT is artificial for me and the reason is to create demand for higher margin chips (in the context of a too weak competition).
AMD is no longer pressuring Intel, sadly, If they were to go down in the GPU realm I would expect Nividia to adopt the same practices as Intel. Not that they would no longer fight for our money but they would create artificial segmentation to rip money out of our pockets.
Last edited by a moderator:
Product differentiation like this happens all the time, not just at Intel. Companies choose to offer fewer features or lower capability for lower-rung products to justify the pricing differential in the eyes of buyers.
For processors, it's also much cheaper to disable parts of a chip than it is to manufacture a wholly separate niche variant at every SKU.
I don't see much reason to merely suspect that Intel is doing what everybody pretty much always does.
No need to wonder, that's exactly what they've been doing for years.
caveman-jim
Regular
This is based on what?
In other news, AMD has announced they are having a news day to announce something about their ambidextrous stategy, with a unannounced guest visitor.
http://ir.amd.com/phoenix.zhtml?c=74093&p=irol-newsArticle&ID=1749646
Merger announcement? Obvious attempt to placate investors and get some value back in the stock? Long term planned announcement?
I was invited to the news event... if I happen to be in the SF area. PM me if you want to pay for flight and hotel.
While I agree it was a bad move you noticed that was in 2009 right?
I guess that they'd still need to increase the IPC... Yup.
- Status
