Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Done it! - Nehalem W5590 Octad 3,33 GHz
- Thread starter gugucom
- Start date
- Sort by reaction score
Errata is frequent in engineering samples, only to be corrected in the final release. It is the worst chip to consider if your computer is used professionally. Gamers love them since all they do is game. Professionals, however, consider them "throw aways" and will never touch them --- they are consistently overloaded with errors. I am not talking a few, but many. I have not read the whole thread, but I seriously hope the most you do with your rig is gaming. For you will inevitably find yourself tracking down abstract errors and attempting to troubleshoot them to no avail. Usually when you least expect it. Unless you have nothing better to do with your time, those chips are nothing but a lesson in futility. Sounds harsh, but anyone who knows what ES stands for also knows they are simply not worth the discount you receive. You could not even pay me to run those chips, the amount of headache is not worth it IMO.
Huh, interesting opinion. Any testing that shows these problems? Seems like most benchmarks and testing you see done is with ES chips, I'd be interested to see some proof of this. I've never personally had any problems with ES chips or other hardware (I'm still using an ES X-25M) from Intel and I'm not a gamer. Maybe I'm just lucky.
Personally, I think it's gotten better overall, but it can still be an issue. Even a single bug can be a problem, as the solution may only be in hardware (most are attempted to be repaired via firmware or the OS).
Do a search of "Intel CPU errata information" (or similar), and see what comes up.
http://www.intel.com/assets/pdf/specupdate/321324.pdf
Errata seem to be given by a complete line of processors which would make sense to me. The above document outlines the current errata for the D0 stepping. They do not seem to be different to the W5590 to any of the other CPUs of that range. I believe that the higher grades are just singled out from testing and that they are not different in silicone design. They just reflect the yield advances in the manufacturing process. Since these samples showed up at the end of the 5500 series release I believe that there is no more risk for errata than it would occur with a regular Apple supplied chip. Now, if I was using Gulftown samples I would perhaps be concerned, but in the specific case I'm not. In fact testers seem to be concerned with the opposite. The samples they are given may be better than what is actually initially sold to the public in order to get better reviews. Due to this reason some testers actually source their chips on the market to avoid being seen as giving out favours.
Errata seem to be given by a complete line of processors which would make sense to me. The above document outlines the current errata for the D0 stepping. They do not seem to be different to the W5590 to any of the other CPUs of that range. I believe that the higher grades are just singled out from testing and that they are not different in silicone design. They just reflect the yield advances in the manufacturing process. Since these samples showed up at the end of the 5500 series release I believe that there is no more risk for errata than it would occur with a regular Apple supplied chip. Now, if I was using Gulftown samples I would perhaps be concerned, but in the specific case I'm not. In fact testers seem to be concerned with the opposite. The samples they are given may be better than what is actually initially sold to the public in order to get better reviews. Due to this reason some testers actually source their chips on the market to avoid being seen as giving out favours.
The processors errata is a big problem with replacing processors in a Mac Pro and using processors such as ES processors or any processor that wasn't originally sold in the Mac Pro model you're using.
Traditionally, system makers/motherboard makers will issue bios updates in order to support new processors. It doesn't mean new processors won't work without an updated bios usually, but it's not officially supported. The reason is, when a new processor comes out, the motherboard makers have to incorporate the microcode updates into the bios to patch the system to support the errata's in each processor.
The problem is, Apple doesn't need to and probably doesn't patch their firmware to take into account for any processor errata's that may come out for future processors that may be compatible with that specific Mac Pro. Lets say Apple knows they only used a specific model and stepping of a processor in a Mac Pro. Then, a newer stepping of the same processor comes out. Apple knows what processors they used, so they know they never need to account for microcode updates for that processors errata. If you use that stepping of the processor, even though it may be an exact same model that Apple used (but a different stepping), you may have issues.
Here is some info from Microsoft on this. They explain it pretty simply and it may help people understand how this works. Basically, they are explaining how it can be updated from within Windows, but they speak to the fact that it could be updated in the bios (or in Apple's case, the EFI).
http://support.microsoft.com/kb/288302
And here is the wikipedia article.
http://en.wikipedia.org/wiki/Microcode
Now, have there been any documented cases of someone actually finding a problem with a upgraded Mac Pro that has failed because of a failure related to an errata of a processor? Probably not. But it's not something I am willing to try with any production Mac Pro.
So truly, the issue is not just ES processors, but any processor that wasn't officially used on the model of Mac Pro that you're trying to upgrade the processors. And not just the model of processor, but it would have to be the identical stepping.
Good luck with your processor upgrades!
Traditionally, system makers/motherboard makers will issue bios updates in order to support new processors. It doesn't mean new processors won't work without an updated bios usually, but it's not officially supported. The reason is, when a new processor comes out, the motherboard makers have to incorporate the microcode updates into the bios to patch the system to support the errata's in each processor.
The problem is, Apple doesn't need to and probably doesn't patch their firmware to take into account for any processor errata's that may come out for future processors that may be compatible with that specific Mac Pro. Lets say Apple knows they only used a specific model and stepping of a processor in a Mac Pro. Then, a newer stepping of the same processor comes out. Apple knows what processors they used, so they know they never need to account for microcode updates for that processors errata. If you use that stepping of the processor, even though it may be an exact same model that Apple used (but a different stepping), you may have issues.
Here is some info from Microsoft on this. They explain it pretty simply and it may help people understand how this works. Basically, they are explaining how it can be updated from within Windows, but they speak to the fact that it could be updated in the bios (or in Apple's case, the EFI).
http://support.microsoft.com/kb/288302
And here is the wikipedia article.
http://en.wikipedia.org/wiki/Microcode
Now, have there been any documented cases of someone actually finding a problem with a upgraded Mac Pro that has failed because of a failure related to an errata of a processor? Probably not. But it's not something I am willing to try with any production Mac Pro.
So truly, the issue is not just ES processors, but any processor that wasn't officially used on the model of Mac Pro that you're trying to upgrade the processors. And not just the model of processor, but it would have to be the identical stepping.
Good luck with your processor upgrades!
Nothing that I read in those linked documents tells me about possible additional errata of speed steps within one stepping (in this case D0). So my conclusion is that a different socket compatible CPU model with different stepping not anticipated by Apple might be a big risk, but staying inside one design but with a faster clock should not be a problem. I'm by no means an expert so additional information on that issue by more knowledgeable people would be appreciated.
This is completely false, Intel never has and never will release engineering samples for review. The only chips that are sent for legitimate review are product samples. Engineering samples are used internally by Intel engineers only. They are used to stress test their new design and to make improvements on them before release. Outside of this context they are illegal, and Intel will never release them for the purpose you state. Even showing them off on this forum is showing off stolen merchandise. The chips are defective, all of them, due to the abnormal testing done to them. Further, if you worked for Intel you would know that they are considered beta, are equivalent to prototypes, and not a single one of them will function as a final release CPU. The bottom line is that these CPUs are damaged, and are always selected from poorer silicon examples. There are no ifs, ands, or buts about this. If you knew someone at Intel, or you yourself worked there, you would know this. Almost all of the false information about ES samples are invented by gamers with vivid imaginations. By gamers I mean enthusiasts who share the same mentality whether they game or not. Only this type of individual promotes the myth that ES samples are handpicked, something to be desired, etc. Nothing could be further from the truth. The truth of the matter is you would have been better off if your project failed and you returned them. You just don't know this yet. Please do not promote false information to unsuspecting readers on this forum...
Woah woah woah, slow down there. Lots of review sites use ES chips for testing and reviews.
Here are two, from some fairly well known sites.
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3634
http://www.tomshardware.com/gallery/core_i7_core_2,0101-164571-0----jpg-.html
Yes, please don't. And this is directed at you, tome viewer.
So if I were to bullet point this for you it would be
- Anand never gets or tests ES chips, in fact no one outside of Paul Otellini's grasp uses ES chips
- They're crap
- Gamers really like them, use them and rave about them but they don't know what they're talking about
- They're going to fail
- They're crap
Would that about sum up your opinion?
The ES chips are not meant to be reviewed no matter what you may think or see. The chips they are using are product releases stamped ES, and there are so few of these legitimate samples around it is impossible that a consumer could obtain one. They are on loan only. The only ES chips that can be obtained by the consumer are the ones I described.
Here is an article that has most of its information correct:
http://www.techarp.com/showarticle.aspx?artno=407
Here is an article that has most of its information correct:
http://www.techarp.com/showarticle.aspx?artno=407
That article contradicts itself. Initially they talk about ES samples being given out for reviews...
Then they say they are only used internally and by partners...
I'd like to see the more details that are supposed to follow the conclusion of that report, but they seem to be missing.
They're just trying to scare people into not buying ES samples. I agree with some of their points, but this is just a scare tactic post, trying to make sure people who buy these things understand what they would be buying. I do believe ES is not to be sold, even though they are, but I do know they are used for reviews and testing. This is easily seen and found by reading many of the review sites.
This is similar to how the first Mac Intel based development systems were 'sold' to end users. They bought them with the understanding that at the end of the program they had to return them, and would lose their $1k they spent on them. Apple, in the end, realized they probably wouldn't get a lot of them back so they changed their mind and as an incentive for returning the product, they were given a new iMac that was actually worth more than the initial $1k they spent. I think their motive, at the time, was to keep the TPM chip out of the general public's hands. There were still a few boxes that made it to eBay (not sure if they ever got sold). Granted, they no longer use the TPM chip to protect OSX from being installed on only Mac hardware, the idea was the same. You didn't own this box, just as you don't own the ES samples that you may obtain legitimately being a review site or partner. You can bet that if Intel promised to give you a retail copy of a processor if you returned the ES after the retail versions were available, there would be far fewer ES processors out there, as almost everyone would rather have the real thing than a sample processor.
The ES samples are cheaper for a reason. I wouldn't buy them for any machine that I used in production.
Some bad news for now.
The 2,26 Nehalem Mac Pro arrived and I managed to get thermal pads although it was an odyssee. The Octoberfest is on full song and getting into the red light district (where all the electronics shops are) is a PITA. Traffic is a killer. None of the shops had the material and after 2h of chaos I visited a trusted Apple AASP who came up with some blue stuff of 1 mm thickness after searching all his junk drawers and card board boxes.
I carefully disassembled the CPU and RAM daughter board and took off the two heat sinks. Unscrewing works much like working on an engine cylinder block. You best approach it with the diagonal technique. I started with sink A and released all four 3 mm hex screws. I then lifted the heat sink with the cpu attached out of the socket. It is essential not to drop it back in when you have started to pull it from the socket because that is probably how Anand managed to screw his socket. Remember the sockets have no clamps!!!
CPU B looked slightly out of angle when I took it of the heat sink. The removal must have been not quite ideal. Apple was perfect with thermal grease application I must say. A very nice pattern.
I fitted the W5590s and increased the thickness of the thermal padding at the rim of the heat sink where the little voltage regulators are contacted by the padding. I cleaned the heat sinks to mirror finish and oh so carefully mounted them back on the sockets using again diagonal tightening technique.
After carefully re assembling the CPU/RAM board I switched power back on and got a white light but no gong sound came to tell me the CPUs had posted and were booting OS X.
I have taken the CPU A out again and re seated it but it looked perfect as the socket looked perfectly ok.
I tested again without joy and I will take out the CPU B to check for damage. If this CPU is also ok I may put the X5520s back in. For now I'm badly frustrated because this project got me very excited.
For the moment I see only two possible options. I got bad CPU from my seller or Apple have enabled the Nehalems only up to W5580 as Tutor used those with success.
If someone has a good idea what could have went wrong please let me know!!!
I will come back and make further reports as the action continues.
Good luck!
With the Oktoberfest, did you hear about them shutting off the entire outside road for a suspicion of a terrorist attack?
Common sense would indicate that tome viewer is following his own agenda here. He has not addressed any of the technical points at all. Instead he builds up a scare scenario which isn't entirely convincing. How about answering those questions tome viewer?
1. Is W5590 silicone any different to W5570 silicone which would indicate that there can be errata particular to that hardware version?
2. What are the differences in design?
3. Where can documentation about those errata - if they exist at all - be found?
On a different note we seem to have no forum contributors here who had negative experience with ES. To the contrary we had Tutor running regular chips outside Apples marketing scope with good success for some time.
1. Is W5590 silicone any different to W5570 silicone which would indicate that there can be errata particular to that hardware version?
2. What are the differences in design?
3. Where can documentation about those errata - if they exist at all - be found?
On a different note we seem to have no forum contributors here who had negative experience with ES. To the contrary we had Tutor running regular chips outside Apples marketing scope with good success for some time.
No, that escaped my attention. But then I never listen much to radio or to TV anyway. So I may have missed that one.
tome viewer is talking about ES chips, as I was also.
The other point that I was trying to make was that chips with newer/different steppings may have a different errata, and microcode patches that normally would be introduced into the bios on boards that officially support newer processors, may not be fixed on a Mac Pro that never supported that processor. A W5590 with the same stepping and errata as a W5570 that was officially used in the Mac Pro should work fine. The only issues I can think of would possibly be thermal issues.
As far as the different chips and steppings, for instance, running a 5300 series processor in a 2006 Mac Pro may not work as well because the 2006 Mac Pro never came with a 5300 processor, so they don't have to add in any microcode patches into the firmware of a 2006 Mac Pro. A 2007 Mac Pro did come with 5300 chips, so as long as you stick with chips using the same steppings and same errata as the ones that came with the 2007 Mac Pro, you should be fine. Then again, nothing is 100% of course.
Afaik the peeps who updated the 1.1s with 5365s got the same benchmarks as the 2.1s?
Cant find the goddam charts now though.
I'm not talking about benchmarks at all, I'm talking about bug fixes for errata's for processors.
But you were not addressing my point there. This is an engineering sample for the fastest version of that chip. It was preceded by several versions with identical D0 stepping with slower speed grades. As such it may not even have any different hardware at all but only been build to check the thermal situation or other possible issues. So what I'm trying to get to is the question of any difference between W5570 and W5590 hardware wise. If there isn't any there ought to be no difference in errata and this whole discussion is pointless.
Here is an Intel Specification Update document which discusses errata specific to the W3500 series Xeon. You will note that the vast majority of errata is the same for all steppings, but some of the errata has been eliminated in the later steppings. In the case of errata AAM69 (CPUID Instruction), it was not present in the C-0 or C-1 steppings but is present in the D-0 stepping. This is the only example I could find where a later stepping introduced an errata not present in a previous stepping. Here is the Specification Update for the W5500 Series (including W5590). This references errata specific to the W5500 series and may be of interest to those who are contemplating upgrading their 8-core Mac Pros.
From what I see in this document, some of what tome viewer has said is correct. From my obviously non-technical reading, errata appears specific to stepping but not necessarily clock speed. Example - W3520 @ 2.66 GHz used in the 2009 Mac Pro Quad is the same D0 stepping as the new W3580 @ 3.33 GHz. Same goes for the Mac Pro 8-core - E5520 (2.26 GHz) has same D0 stepping as W5590 @ 3.3 GHz. It would appear that all 2009 Mac Pros in production are using D0 steppings. Since errata is specific to steppings, replacing the stock CPU with a higher clocked version of the same stepping should not introduce any errata not already addressed in the original firmware.
Of course I know nothing about this topic, so I could be completely wrong.
From what I see in this document, some of what tome viewer has said is correct. From my obviously non-technical reading, errata appears specific to stepping but not necessarily clock speed. Example - W3520 @ 2.66 GHz used in the 2009 Mac Pro Quad is the same D0 stepping as the new W3580 @ 3.33 GHz. Same goes for the Mac Pro 8-core - E5520 (2.26 GHz) has same D0 stepping as W5590 @ 3.3 GHz. It would appear that all 2009 Mac Pros in production are using D0 steppings. Since errata is specific to steppings, replacing the stock CPU with a higher clocked version of the same stepping should not introduce any errata not already addressed in the original firmware.
Of course I know nothing about this topic, so I could be completely wrong.
Total cost was 3900 €. Prices being generally higher here you can take that for a $ figure if you translate it. My x5365 octad is on ebay now and I expect it to fetch 2600 € next Sunday. It currently has 8 bids topping over 1700€ after 36 h out there. I had over 370 hits and have over 80 observers of that auction. I expect the pair of X5520s to get me another 800€ if it goes well. So effectively the upgrade should cost me about 500€ if all goes well. If Apple would offer a 3,33 octad the extrapolated price (linear from the 2,26 - 2,93) here would be 6500€.
Hat's off to you Gugucom. Well done! I'm very impressed, you now own my dream machine. With any luck, I'll be attempting the very same build in the next week or so. This thread is a phenomenal resource. The details you provided in your posts have been excellent.
THIS MESSAGE IS FOR TUTOR:
Apologies for the unorthodox method of contact. Although, I've gathered much from the amazingly informative forums here, I have an urgent and specific question to ask you. I can send you my email if need be. Please PM me at your earliest convenience. Thank you!
THIS MESSAGE IS FOR TUTOR:
Apologies for the unorthodox method of contact. Although, I've gathered much from the amazingly informative forums here, I have an urgent and specific question to ask you. I can send you my email if need be. Please PM me at your earliest convenience. Thank you!
My old X5365 MP1,1 with 12 GB RAM went for 2.810 on Ebay last night. So it did exceed my expectations. It got me more than the basic 8 core Nehalem which I got refurbished at 2.599. I will be selling the original E5520s soon.
I had a contact with the supplier of the W5590 ES at the weekend but I don't know if any of the samples remain available. You can give me your email address and I will pass it on.
I had a contact with the supplier of the W5590 ES at the weekend but I don't know if any of the samples remain available. You can give me your email address and I will pass it on.