Hyperthreaded i5-3570k? There is one, but can it be replicated?

[BulletSpongeRTR]

(I am re-posting this here in Tech News as another member suggested. I apologize for the duplicate threads.)
 
I know, I know, you're jumping to the same conclusion I did...........FAKE! But, 60+ pages of discussion/study of the CPU (live streamed validations, benches, BIOS dumps, etc, etc) have proven that the CPU in question is actually Hyperthreading. He has no idea how it happened but comparison of the microcode between his 3570k and a 3770k show that there is only a difference of a small number of characters (encrypted of course) between the two. A few OCN members knowledgeable on the subject are investigating to see if the hyperthreading can be reproduced in another 3570k. If they succeed it may, in theory, mean that HT can be unlocked in all 1155 i5's. Let that sink in a moment. Goodbye i7 sales. I'll not bore you with anymore details I barely understand but will close with the link below. Oh, be sure to check out his screenshots.
 
http://www.overclock...d-4c-8t/0_100

[Enderman]

I doubt it...probably fake. Hyperthreading is hardware reliant no?

[Ryan Leech]

I doubt it...probably fake. Hyperthreading is hardware reliant no?

No, just microcode. The Celeron, Pentium, i3, i5, and i7 are all the same chip (in the same gen). They just have different microcodes

[IPCharger]

Well, getting an i7 for gaming has always been about bragging rights (e-penis), not actual performance benefits. Reminds me of the batch of 290 that could be flashed to 290x.

[Ren]

Possibly the best, most original content posted here. Nice job.

 

 

I'm not a frequent user of OCN but I do check it occasionally, didn't see this at all (although it was posted less than a week ago). Looking forward to developments because of how game-changing this may be. I don't think it'll be that big of a deal but I guess it'd be neat if you wanted a budget performance chip.

 

 

That thread is so huge, though. I'll need to actually read thru all of it, maybe even post there and ask to find out if this is something a bios update will fix. That is unless someone here knows or read whether or not that's possible already..

 

 

Well, getting an i7 for gaming has always been about bragging rights (e-penis), not actual performance benefits. Reminds me of the batch of 290 that could be flashed to 290x.

 

I'm not sure if this is a small number of people, ones that may not know what they're talking about but I've seen the statements uncontested, so I'll give it the benefit of doubt. On OCN, a bunch of people are supporting the idea that hyperthreading will be useful in games soon. There are supposedly some performance benefits from using an i7 over an i5 in Battlefield. I haven't seen proof nor believe it but I haven't bothered looking around and this was just from me reading a few random posts about why people were choosing haswell refresh i7's over i5's.

 

Take it with a grain of salt, but don't eliminate the possiblity just yet .

[Enderman]

No, just microcode. The Celeron, Pentium, i3, i5, and i7 are all the same chip. They just have different microcodes

um no

I would like to see a celeron that can match a 4790k with just a change of code.

[BulletSpongeRTR]

He has live streamed benches, validations, etc to prove it. It would appear that Hyperthreading is enabled/disabled within the CPU's microcode. Comparison of the microcode between a 3570k and 3770k shows a difference of only a few characters (encrypted of course). I believe that it is a 3570k and it is hyperthreaded but I am very skeptical that they will ever be able to reproduce it. 

[Ryan Leech]

um no

I would like to see a celeron that can match a 4790k with just a change of code.

That's what intel does at the factory...

[Vitalius]

um no

I would like to see a celeron that can match a 4790k with just a change of code.

Um, yes. 

They are all the same chip, but not all of them have the same level of imperfections. 

The more stable chips are sold as higher tiers. i.e. 4790k. The crappier chips are sold as lower tiers, i.e. celeron.

It's not unreasonable to jump up a single tier (i.e. 3570k to 3770k), if you can figure out how. You'd probably just have to underclock a little to keep it stable. 

This is something most manufacturers of silicon chips have done for a long while to maximize yields. That way they don't have to throw away any of the chips. They just sell the crappier ones as lower performance. 

[Enderman]

That's what intel does at the factory...

then why do they have different numbers of cores?

and different frequencies?

and different amount of cache?

and different TDP?

and different manufacturing processes?

 

Is a celeron really just an i7 with 90% of stuff disabled?  :blink:

@Vitalius

[Ryan Leech]

then why do they have different numbers of cores?

and different frequencies?

and different amount of cache?

and different TDP?

and different manufacturing processes?

 

Is a celeron really just an i7 with 90% of stuff disabled?

They actually have the same manufacturing process, as long as they are the same gen on the same socket. But yes, a Celeron is just an i7 with 90% disabled

[Vitalius]

then why do they have different numbers of cores?

and different frequencies?

and different amount of cache?

and different TDP?

and different manufacturing processes?

 

Is a celeron really just an i7 with 90% of stuff disabled?  :blink:

@Vitalius

Yeah, pretty much. 

If you enabled it and tried to run it, your system would probably constantly BSOD though. 

It's not like they just do it so they have a range of performance chips. 

[BulletSpongeRTR]

Um, yes. 

They are all the same chip, but not all of them have the same level of imperfections. 

The more stable chips are sold as higher tiers. i.e. 4790k. The crappier chips are sold as lower tiers, i.e. celeron.

It's not unreasonable to jump up a single tier (i.e. 3570k to 3770k), if you can figure out how. You'd probably just have to underclock a little to keep it stable. 

You should have a look at his overclocks since the "unlock". 4.5/1.08v and I believe 5/1.18v, you'll have to suspend disbelief long enough to check out his screenshots and validations.

[Vitalius]

You should have a look at his overclocks since the "unlock". 4.5/1.08v and I believe 5/1.18v, you'll have to suspend disbelief long enough to check out his screenshots and validations.

I'm not saying, "there aren't golden chips", I'm just saying, more than likely, you'll need to underclock to keep the chip stable if it's borderline.

[brob]

Just because one can change the microcode to enable hyperthreading does not mean that the chip can run error free. It seems likely that at least some i5 are chips with die imperfections that affect hyperthreading. Just as some i3 have imperfections in one or two cores.

[BulletSpongeRTR]

I'm not saying, "there aren't golden chips", I'm just saying, more than likely, you'll need to underclock to keep the chip stable if it's borderline.

In hindsight, I agree completely. I forgot that the purported 3570k was already "golden" before it was unlocked.

[EChondo]

Honestly, I just thought about this the other day.

 

"What would happen if someone figured out how to enable HT for i5 CPU's?"

 

And now this thread pops up...

[mvitkun]

Honestly, I just thought about this the other day.

 

"What would happen if someone figured out how to enable HT for i5 CPU's?"

 

And now this thread pops up...

Read through the entire thread, took a while though, and it seems somewhere between page 20 and page 30 they were more or less content with the proof.

Then on page 36 somebody tried comparing the bios dump he gave to a stock bios.

 

I have completed the comparison between a normal 1.60 ROM and the dumped one from aerotracks.

 

The first section from 00000000 until 0015DCF1 is still unknown as it is garbled, but it may very well be a CPU microcode block or some kind of other encrypted verification block. Since it is missing from aerotracks's dump, I am more inclined to believe that it is the former rather than the latter. In this case, the dump is quite worthless as that part obviously won't ever get dumped along with the BIOS image in such a case.

 

Between here and the next differential section there is a section containing default BIOS fallback settings, in the same format as the section following this on aero's ROM. May be useful for changing defaults in the ROM before flashing, but I'd advise against doing so.

 

From position 00200659 until 002341C5 there is a lot of data present in the dump but not in the OE ROM, which I identify as being NVRAM contents.

Then there is only one more deviation in the 0023F310 row, which is three bytes long and shouldn't be of any notable importance.

 

That is literally it, there are no other differences between the ROMs other than these two long contiguous sections, one of which is unidentified and only found in the first part of the OE ROM, the other being settings data and as such only found in aerotracks' ROM.

 

I hereby form my opinion that indeed this chip has developed a defect that has proven to be of a rather beneficial nature. It should work exactly the same if you put it into another Z77 Formula board, provided there haven't been secret updates in other BIOS versions to circumvent misidentifications like these.

Then on page 49 OP disassembles the bios and starts comparing his 3570K (3770) to regular 3770K, 3770K w/ HT disabled, and 3570K bios dumps provided by the OC.net community.

   8c ff ff ff             callq  0x344     3b8:       45 33 db                xor    %r11d,%r11d     3bb:       4c 8b d0                mov    %rax,%r10     3be:       48 85 db                test   %rbx,%rbx     3c1:       74 4e                   je     0x411     3c3:       4d 85 c9                test   %r9,%r9     3c6:       74 49                   je     0x411     3c8:       48 83 f8 ff             cmp    $0xffffffffffffffff,%rax     3cc:       74 43                   je     0x411     3ce:       8b 10                   mov    (%rax),%edx     3d0:       49 83 c2 04             add    $0x4,%r10     3d4:       85 d2                   test   %edx,%edx     3d6:       74 16                   je     0x3ee     3d8:       4c 8b c2                mov    %rdx,%r8     3db:       41 8b 02                mov    (%r10),%eax     3de:       44 03 d8                add    %eax,%r11d     3e1:       03 c0                   add    %eax,%eax     3e3:       49 83 e8 01             sub    $0x1,%r8     3e7:       4d 8d 54 82 04          lea    0x4(%r10,%rax,4),%r10     3ec:       75 ed                   jne    0x3db     3ee:       48 8b 44 24 50          mov    0x50(%rsp),%rax     3f3:       89 13                   mov    %edx,(%rbx)     3f5:       45 89 19                mov    %r11d,(%r9)     3f8:       48 85 c0                test   %rax,%rax     3fb:       74 03                   je     0x400     3fd:       83 20 00                andl   $0x0,(%rax)     400:       48 8b 44 24 58          mov    0x58(%rsp),%rax     405:       48 85 c0                test   %rax,%rax     408:       74 03                   je     0x40d     40a:       83 20 00                andl   $0x0,(%rax)     40d:       33 c0                   xor    %eax,%eax     40f:       eb 0a                   jmp    0x41b     411:       48 b8 02 00 00 00 00    movabs $0x8000000000000002,%rax     418:       00 00 80      41b:       48 83 c4 20             add    $0x20,%rsp     41f:       5b                      pop    %rbx     420:       c3                      retq        421:       cc                      int3        422:       cc                      int3        423:       cc                      int3        424:       48 83 ec 28             sub    $0x28,%rsp     428:       8b ca                   mov    %edx,%ecx     42a:       45 8b d0                mov    %r8d,%r10d     42d:       e8 12 ff ff ff          callq  0x344     432:       4d 85 c9                test   %r9,%r9     435:       74 42                   je     0x479     437:       48 83 f8 ff             cmp    $0xffffffffffffffff,%rax     43b:       74 3c                   je     0x479     43d:       48 8d 48 04             lea    0x4(%rax),%rcx     441:       44 3b 10                cmp    (%rax),%r10d     444:       73 33                   jae

At this point he's trying to figure out which parts of it affect HT so that he can enable it and disable it with other processors.

[bcredeur97]

If someone makes a tool or even just a how-to on how to do this. I will love them forever. suddenly video rendering times will get cut. lol

[jezza39]

Just because one can change the microcode to enable hyperthreading does not mean that the chip can run error free. It seems likely that at least some i5 are chips with die imperfections that affect hyperthreading. Just as some i3 have imperfections in one or two cores.

 

Of course, it may also need a voltage bump to keep stable

[Kamina]

I wouldn't compare a celeron to an i7. The die shots are different but they are based on the same architecture. 

 

They are grouped, like i5 low tier to high tier and then up to K, and the imperfections made are binned or thrown away.

[BulletSpongeRTR]

This is a link to a 3-year old story someone posted on OCN. Relevant? 

 

http://www.engadget.com/2010/09/18/intel-wants-to-charge-50-to-unlock-stuff-your-cpu-can-already-d/

[Vellocity]

For those who don't feel like going through all 50+ pages of that thread, here's the video he posted to provide more proof.

 

[Akolyte]

I think hyper threading is hardware based.  I would think the most you could do is trick the OS into thinking you have more cores therefore pushing your cores further than what they were supposed to be. 

[BulletSpongeRTR]

I think hyper threading is hardware based.  I would think the most you could do is trick the OS into thinking you have more cores therefore pushing your cores further than what they were supposed to be. 

I'm siding with those who think that L3 cache is lasered off but other features (HT,unlocked multi, and who knows what else) are disabled or enabled within the microcode. Here is a quote from the article I linked above, "Hold onto your hyperthreaded horses, because this is liable to whip up an angry mob -- Intel's asking customers to pay extra if they want the full power of their store-bought silicon. An eagle-eyed Engadget reader was surfing the Best Buy shelves when he noticed this $50 card -- and sure enough, Intel websites confirm -- that lets you download softwareto unlock extra threads and cache on the new Pentium G6951 processor." 

 

That was an experiment Intel conducted on an 1156 CPU. Could this feature have been integrated into 1155 CPU's but never implemented by Intel in the market?