Christmas May Have Come Early for AMD...

P

Pepsiennis

[H]ard|Gawd
Joined
Feb 5, 2007
Messages
1,290
Trouble on the Core i7 horizon? If it's true, AMD may have received an early Christmas present. Now, everyone has to wait and see what Phenom II (DDR2 and then DDR3) has to offer price vs. performance-wise. It'll be interesting how everything plays out in the coming month.
 
Lol this is fudzilla we're talking about. AMD needs to concentrate on their own shit and execute a decent launch. Good overclocking procs, good mobos, and low prices. Anything else is a distraction.
 
They quote Intel's own document, if you're worried about veracity. But you're right, and as I mentioned before, we have to wait and see how AMD performs with the proc and its introduction. After that, I'm interested to see what 'fix' Intel offers and how it's received by the community.
 
Eh they will fix it in the next core revision.

It is not like I plan on getting a first generation i7 anyway, maybe when it hits a 32nm process.
 
They certainly will! And I have every belief it will have no tangible effect on anything we would do real-world. But as the TLB bug was said to have hurt the original Phenom sales, I wonder if this would have any impact on adoption if (and a big if it is) Phenom II is a marked step forward for AMD? Initial accounts of the Phenom II, though unreliable, sound cautiously positive.
 
Indeed, they say not. But the statement on Page 37 says it "may apply to the following documents... Intel® Core™ i7 Processor Datasheet" . While they are clearly intending to state the TLB in reference to IA-32/64, as Intel PR Manager (not engineer) Dan Snyder says, why include that original heading of "may apply to... Core™ i7 Processor Datasheet" if it only affected the earlier architectures?
 
If you look it's not listen on the Nehalem list of errata. And if no one noticed in Core 2 I don't think anyone is going to notice even if it is still there.
 
Well, this is what we mentioned earlier about Intel having handled the errata much better than AMD. And as I said earlier, just as the AMD errata never cropped up for the overwhelming majority of Phenom users, the Intel errata shan't for i7.
 
While I don't always trust Fudzilla, just because Intel says it doesn't have the TLB issue doesn't mean to me that Fudzilla was wrong. it could just be Intel covering itself.
 
@OP: Thread title is true to a small degree, due to the i7's being so expensive, and the tlb bug, but aren't the i7's selling great and getting even greater reviews in addition to the fact that no new cpu's will be release before xmas and they still haven't been told they won the anti-trust case?

AMD does deserve an early XMAS, imo, b/c I honestly can't understand why anyone would buy intel products, when amd beat them to the 64 bit market, had an imc long before intel who just started using it, and beat intel to dual-core. in addition to price, the fact that games are largely gpu-limited, a company that made a disastor that was the pentium d that ran insanely hot and was inferior to amd's offering that costed a little less, and the fact that intel was making clearly inferior products, they covered it up by paying oem's to not sell amd cpu's and even paid fry's to not sell amd cpu's as well as using advertising/brand name recognition to sell inferior products.

[/at least partially OnT rant]

Even AMD's latest high-end nb/sb chipsets are better than intel's have been.
 
sensationalism is the name of the game at fudzilla. i'll ignore this tlb fiasco just like i did with amd's. the vultures wanted another pound of flesh out of amd on top of what they already gobbled up from the delayed phenom launch and the poorer than expected performance. plus it didn't help that the b3 revision sans tlb overclocked better. the vultures are looking for another meal; making something out of nothing, and sqwaking at more of the same.
 
Pepsiennis said:
They quote Intel's own document, if you're worried about veracity. But you're right, and as I mentioned before, we have to wait and see how AMD performs with the proc and its introduction. After that, I'm interested to see what 'fix' Intel offers and how it's received by the community.
Click to expand...

Too bad nobody cares about the bug ...
 
The only time the bug is a problem is when web sites want it to be. Core2 had a similiar problem as well, but hardly anyone ever heard about it.
 
xdivenx said:
The only time the bug is a problem is when web sites want it to be. Core2 had a similiar problem as well, but hardly anyone ever heard about it.
Click to expand...
It was patched in a BIOS update in April 2007, with no loss in performance. The Core i7 bug was patched the same way before it was released.

Core 2 systems without the BIOS patch and an affected OS (does not apply Windows or Linux) could "rarely" encounter the bug if the TLB invalidation isn't handled as specified by the Conroe documentation. This bug was already patched for every Core i7 based system on release, so essentially it doesn't apply as the updated note in the documentation explains.
 
pxc said:
It was patched in a BIOS update in April 2007, with no loss in performance. The Core i7 bug was patched the same way before it was released.

Core 2 systems without the BIOS patch and an affected OS (does not apply Windows or Linux) could "rarely" encounter the bug if the TLB invalidation isn't handled as specified by the Conroe documentation. This bug was already patched for every Core i7 based system on release, so essentially it doesn't apply as the updated note in the documentation explains.
Click to expand...

The funny part is , the bug not being really a bug.As I've said previously , Intel went for a standard way with regards to TLB invalidation starting with Core.Basically , the TLB works as intended, only that the OS must be aware of this.To my knowledge , it doesn't pop out on Linux,Windows, Solaris,BSD and OS X.Is anything else left?

Section 10.9 INVALIDATING THE TRANSLATION LOOKASIDE BUFFERS (TLBS) of the Intel® 64 and IA-32 Architectures Software Developer's Manual, Volume 3A: System Programming Guide will be modified to include the presence of page table structure caches, such as the page directory cache, which Intel processors implement. This information is needed to aid operating systems in managing page table structure
invalidations properly.

....

In rare instances, improper TLB invalidation may result in unpredictable system behavior, such as system hangs or incorrect data. Developers of operating systems should take this documentation into account when designing TLB invalidation algorithms.
Click to expand...

As JumpingJack said on another forum :

It is like saying if you does not put on fresh underwear then you may generate unpredictable body odor. You cannot claim there is a design flaw in your undewear based on your inability to appropriately invalidate you underwear.
Click to expand...

AMD's bug was totally different in nature , being a design flaw in the TLB.As someone in the know( IIRC he works/worked for AMD as a designer) pointed out :
> 3) It _ONLY_ happens when you are using hardware virtualization.
Click to expand...

Not true.

The erratum in question (#298) is not "dependent" on SVM.

A particular virtualization setup simply was the first to make
it somewhat "reproducible at will".

I have seen it occur in non-SVM cases.

An erratum that might seem "rare" and "exceedingly hard to
trigger" to you, can quickly rear its ugly head once you have
a large enough population of machines. ;-)
Click to expand...

Full discussion here : http://aceshardware.freeforums.org/i7-bug-t700.html
 
I know it's not really a bug, but it's easier to type that than erratum or change of behavior every time. :p

The only system programmer I remember complaining about it was Theo De Raadt, and i'm not sure that he even had to make any changes since he was just ranting in general.
 
AMD have gone so far south with their CPU offerings I'm switching to Intel, AMD can't even compete with the last generation of CPU's Intel brought out, let alone the new i7's, don't know what Fudzilla is talking about?

And even if there was a problem with the i7, the fix will still out perform anything AMD have to offer.
 
Outamyhead said:
AMD have gone so far south with their CPU offerings I'm switching to Intel, AMD can't even compete with the last generation of CPU's Intel brought out, let alone the new i7's, don't know what Fudzilla is talking about?

And even if there was a problem with the i7, the fix will still out perform anything AMD have to offer.
Click to expand...

This is a total load of FUD.
If you have nothing relevant to the topic, and are just here to tell your story of "OH YEAH GUYZ I WENT OT INTEL THEY ARE BESTEST!", Leave. This is the AMD section.
 
Outamyhead said:
AMD have gone so far south with their CPU offerings I'm switching to Intel, AMD can't even compete with the last generation of CPU's Intel brought out, let alone the new i7's, don't know what Fudzilla is talking about?

And even if there was a problem with the i7, the fix will still out perform anything AMD have to offer.
Click to expand...

Well, that may be a bit of old news, true, but the newer news is that Phenom II is looking to be more competitive. And by that, I mean worth one's consideration, not head-to-head with i7, where it's obviously an iteration behind (so please no taking my words out of context, if you please). There's a CPU-Z validation of a Phenom II 940 OC'd to 4.4 GHz. Legit or not? You tell me.

http://valid.canardpc.com/show_oc.php?id=448873
 
You must log in or register to reply here.
Back
Top