New Rev. DH-F3 and JH-F3 for K8s?

[zir_blazer]

SPOILERS: This is supposed to be a highly technical Thread, be sure that you're aware of OPN deciphering or got interesing AMD Technical Documents before posting useless or redundant data

Until now, I was aware of the existence of the following three physically different Cores with a DDR-II IMC (This PDF proves this):

DH-F2 Orleans/Manila: Single Core, 512 KB Cache L2 (Venices were DH-E3/E6)
BH-F2 Windsor: Dual Core, 512 KB * 2 Cache L2 (Manchesters were BH-E4)
JH-F2 Windsor: Dual Core, 1 MB * 2 Cache L2 (Toledo were JH-E6)

I'm partially aware of the following OPN surfixes in Socket M2 Processors:

CS: JH-F2 Windsor. 100% sure as it is in 1 MB * 2 Dual Cores and 512 KB * 2 ones where half of the Cache L2 is disabled
CU: BH-F2 Windsor. 100% sure as it is only in Dual Cores models with only 512 KB * 2 Cache L2
CN: DH-F2 Orleans/Manila. Not completely sure, it is in most A64 and Semprons models
CW: Completely unsure, it should be either a BH-F2 or/and JH-F2 with an entire Core disabled and Cache L2 too (Available on one A64 and several Semprons models)

If anyone can confirm me what is the full Revision from Socket M2 Processors with the CW OPN surfix, I will be thankful as right now I don't know and I'm curious to know what is the physical Core that they uses. Up to this point, everything should be understandable... ...Now, lets the Revisions attack begins
Anyone that checked AMD Compare should have noticed that besides the new Revision G1 (That, as now, is incomplete, but probabily belongs to the BH family as all them got 512 KB * 2 Cache L2) there is a new Rev. F3 that is used only by the new Socket F A64FXs for the 4x4 platform (The OPN surfix for them is DI, but this is useless as no Socket shares the same OPN surfixes for any of its Processors with any other). However, I can't find an AMD Technical Document that states anything about this new Revision.
Going further than this, I finded in another new AsRock BIOS upgrade some new OPNs that many would like to know...

ADA5000IAA5CZ
ADA5200IAA6CZ
ADA5400IAA5CZ
ADA5600IAA6CZ

The shining new CZ surfix that is stated to be Rev. F3, and considering than two of these models got 1 MB Cache L2 * 2, then it should probabily belong to the JH family (Call it JH-F3). Pretty interesing, thanks AsRock for OPN leaks, you did it with 65nm ones too
However, furthen than that, I tried to searching for JH-F3 Athlon in Google, and I found this extremely weird Source Code for Linux. Scrolling down to the end of it is this:

{ X86_VENDOR_AMD, 0x40f50 }, /* SH-F0 Socket F (1207): Opteron */
{ X86_VENDOR_AMD, 0x40f70 }, /* AM2: Athlon64/Athlon64 FX */
{ X86_VENDOR_AMD, 0x40f40 }, /* S1g1: Mobile Athlon64 */
{ X86_VENDOR_AMD, 0x40f11 }, /* JH-F1 Socket F (1207): Opteron Dual Core */
{ X86_VENDOR_AMD, 0x40f31 }, /* AM2: Athlon64 x2/Athlon64 FX Dual Core */
{ X86_VENDOR_AMD, 0x40f01 }, /* S1g1: Mobile Athlon64 */
{ X86_VENDOR_AMD, 0x40f12 }, /* JH-F2 Socket F (1207): Opteron Dual Core */
{ X86_VENDOR_AMD, 0x40f32 }, /* AM2 : Opteron Dual Core/Athlon64 x2/ Athlon64 FX Dual Core */
{ X86_VENDOR_AMD, 0x40fb2 }, /* BH-F2 Socket AM2:Athlon64 x2/ Mobile Athlon64 x2 */
{ X86_VENDOR_AMD, 0x40f82 }, /* S1g1:Turion64 x2 */
{ X86_VENDOR_AMD, 0x40ff2 }, /* DH-F2 Socket AM2: Athlon64 */
{ X86_VENDOR_AMD, 0x50ff2 }, /* DH-F2 Socket AM2: Athlon64 */
{ X86_VENDOR_AMD, 0x40fc2 }, /* S1g1:Turion64 */
{ X86_VENDOR_AMD, 0x40f13 }, /* JH-F3 Socket F (1207): Opteron Dual Core */
{ X86_VENDOR_AMD, 0x40f33 }, /* AM2 : Opteron Dual Core/Athlon64 x2/ Athlon64 FX Dual Core */
{ X86_VENDOR_AMD, 0xc0f13 }, /* AM2 : Athlon64 FX*/
{ X86_VENDOR_AMD, 0x50ff3 }, /* DH-F3 Socket AM2: Athlon64 */

Plus many never seen Revisions from previous K8s that were probabily only used for Engineering Samples. However, we also got the DH-F3 Core there and we still didn't saw that one, and considering that JH-F3 is just being released, then the DH-F3 Core may see the light from day in the time to come (Remember that AMD sayed that they will be fully converted to 65nm around halfway 2007, so during 6 more months there will be 90nm K8s production).
Sure, the document may be completely unoficial and coming from some obscure place in Google, but the data there seems to be pretty consistent (And done by someone with access to a lot of AMD Engineering Samples, it seems), so I believe in it. Wonderful that 65nm being officially released still don't show in stores yet 90nm is far from dead, right?

[gOJDO]

Hmmmm.....I wonder what revision is FX-7x. Could it be JH-F3?
DH-F3, smells like 3.2GHz sAM2 K8

[zir_blazer]

Hmmmm.....I wonder what revision is FX-7x. Could it be JH-F3?
DH-F3, smells like 3.2GHz sAM2 K8

It is. There is no other 1 MB * 2 Cache L2 Core around whose Silicon Revision is F3.

[Lightman]

Those F3 revisions might be ones with SiGe on 90nm. I read somewhere that AMD will add SiGe to 90nm for higher clocks. Originally SiGe was planned only for 65nm process.

[zir_blazer]

Those F3 revisions might be ones with SiGe on 90nm. I read somewhere that AMD will add SiGe to 90nm for higher clocks. Originally SiGe was planned only for 65nm process.

Yeah, I remember that too from a few AMD slides, is a solid possibility. That means that it should overclock somewhat further than current Revision XX-F2 is capable of. God knows what 65nm is capable of is 90nm is still kicking ass.

[gOJDO]

If AMD implement SiGe on the 90nm node, than it will be faster than the 65nm node for sure. The only problem here is production price and heat dissipation.

[Lightman]

Today I found one interesting thread on this forum

http://www.xtremesystems.org/forums/...38#post1881638

What is interesting that Venturi OCed those 2220SE Opties up to 3200MHz on AIR without Vcore increase! This is amazing result from 90nm K8, so I think that he might have F3 stepping.
I noticed that FX-74 CPUs used in reviews was from 35th week of this year. This means that Opterons produced after this week also might be based on F3 revision, because both are identical! Things are less optimistic for AM2 Athlons due to different core (2x512KB L2 instead of 2x1MB). Might be that later on AMD wil transfer whole 90nm production to F3 deriveration.

Anyone with Opterons or QuadFXs to check my suspicions??

[gOJDO]

@Lightman
FX-7x is different from Opteron. It has unbuffered ODMC and unlocked multiplier.
Also there are sAM2 X2 with 2x1MB L2: 4000+, 4400+, 4800+, 5200+.

[Lightman]

@Lightman
FX-7x is different from Opteron. It has unbuffered ODMC and unlocked multiplier.
Also there are sAM2 X2 with 2x1MB L2: 4000+, 4400+, 4800+, 5200+.

Right! But differences between Opteron and FX-7x are made after die is finished by cutting small bridges or connecting different pads from core to BGA.
There is also theory why FX-7x have only 2 DIMMs per socket. According to some people (I forgot who and when ) Opteron can work with unregistered 2 DIMMs, but anything more will put to much pressure on Drive Strength of memory controller. The reason why it's working on S939 and AM2 is that pins used for extra HT links in Opterons are used to add Drive Strength (more power/load can by applied). It was exactly same with old AMD760MP chipset. It was able to work with 2 unregisterd DIMMs or 4 registered. (on side note someone on this forum was claiming running unregistered memory with FX-51 on S940)

Regarding sAM2 X2 with 1MB L2 they all are discontinued, so of no importance. But we will get 1MBL2 once again with X2 6000+.

[zir_blazer]

The Unlocked Multiplier doesn't says anything, as neither the IMC that works with Unbuffered Memory Modules. Don't forget than the SH-CG Core (Sledgehammer/Clawhammer) was available on S754, S939 and S940, yet it was the same physical Core with enabled or disabled features (From an S754 A64 with only Single Channel capabilities and 512 KB Cache L2 (Half of the 1 MB Cache L2 disabled) to the S940 Opterons 8xx working with Buffered + ECC Memory Modules).
What we still don't know is what is supposed to be different in these new A64 FX-7x to work in Socket F Motherboards with regular Unbuffered Memory Modules. Its impossible than it is a Chipset based modification because the Memory Controller is in the Processor itself, probabily a Pin that controls the obligatory Buffered Memory Modules requeriment is unconnected or is a BIOS-based sort of hack. We also don't know if Opterons works on 4x4 platform Motherboards (And if they do without using Unbuffered Memory Modules) or viceversa with A64 FX-7x in regular Opterons Motherboards.
About the regular Socket F Opterons xxxx, all them uses the Rev. JH-F2 Core (Opterons 2xxx got a CQ OPN surfix and Opterons 8xxx a CR surfix, no other versions are available), being the A64 FX-7x the first ones to debut with the new JH-F3 one. However, considering than all the K8 features in that generation are there and they can be enabled or disabled, AMD can use it for any Socket M2/F Athlon 64/Opteron model that they want to.

[zir_blazer]

I can't believe that I worked a few hours to make this Thread and give away unreleased info yet almost no one payed the attention than this Thread deserves. The quality of the data was worthy of any NDA leaker or Engineering Sample tester that just post a CPU-Z Screenshot then runs away.

The Inquirer confirmed that AMD released the A64X2 5400+ and 5600+ and they OPN appeared on AMD Compare here. As expected, they are Rev. F3. Take note than the TDP for those two 2.8 GHz parts is 89W, meaning than the SiGe process that they probabily uses should decrease the power consumption needed to archieving the same Frequency than in the 90nm SOI with DSL process.

[w0mbat]

I payed mutch attention

[FghtinIrshNvrDi]

I need more layman's terms.

Ryan

[VulgarHandle]

so, is the point of the original post that the new models aren't just due to better yields, but possibly a new process?(and not to mean 90nm to 65nm)

[generics_user]

I can't believe that I worked a few hours to make this Thread and give away unreleased info yet almost no one payed the attention than this Thread deserves. The quality of the data was worthy of any NDA leaker or Engineering Sample tester that just post a CPU-Z Screenshot then runs away.

The Inquirer confirmed that AMD released the A64X2 5400+ and 5600+ and they OPN appeared on AMD Compare here. As expected, they are Rev. F3. Take note than the TDP for those two 2.8 GHz parts is 89W, meaning than the SiGe process that they probabily uses should decrease the power consumption needed to archieving the same Frequency than in the 90nm SOI with DSL process.

i payed lots of attention to this thread but it was so perfect and i couldn´t say anything about it

it is really interesting what amd does at the moment and maybe we´ll see some new WR's on AM2. (well not compared to C2D :P)

soon we´ll see what amd can do with SiGe and with (hopefully) upcoming infos about 65nm which, as far as i know, currently doesn´t make use of SiGe we can see (at least a bit) what amd can do with K8L (in terms of clock speed)

[zir_blazer]

Thanks AsRock, keep leaking OPNs...

ADX6000IAA6CZ (Rev. JH-F3? 3.0 GHz, 1 MB * 2 Cache L2)

The ADX prefix should means that it belongs to another power consumption category, presumabily a higher one due to its higher Frecuency. If I remember correctly, Opterons SE (125W TDP) prefix was OSX, so you can accurately speculate about this one. There is another weird one...

ADA4000IAA4CW (Rev. **-F2, 2.6 GHz, 1 MB Cache L2)

Remember what I was saying early about that I didn't know what physical Core the CW surfix belongs to? Now I am even more confused. On the least for this model, it should be guaranteed JH-F2 with one Core disabled but all the Cache L2 from one Core totally functional.

[oldblue]

Thanks zir_blazer. For what it's worth, I find this interesting.

[s7e9h3n]

ADA4000IAA4CW (Rev. **-F2, 2.6 GHz, 1 MB Cache L2)

Remember what I was saying early about that I didn't know what physical Core the CW surfix belongs to? Now I am even more confused. On the least for this model, it should be guaranteed JH-F2 with one Core disabled but all the Cache L2 from one Core totally functional.

This is an oddity...I'd say that it's likely a misprint or error on Asrock's part as the "4" denotes that the cpu is a 512K L2 cache cpu. If it were truly a 1mb L2 cache cpu, then the "4" would be replaced by a "5". Here's one (or two) to add to your list (as well as Asrock's) :

[zir_blazer]

Its pretty probable than you're right with the ADA4000IAA4CW. It would make sense that it was 2.6 GHz but with 512 KB Cache L2 instead of 2.4 GHz with 1 MB Cache L2 like regular A64s 4000+.

By the way, more brand new OPN info! Thanks to a recent Thread, I heared about an A64X2 with the OPN ADO3600IAA5DL that differed from the 65nm A64X2 3600+ OPN that I reported before in the fact that the surfix is DL and not DD (Though I can't identify what the difference is). Because this, I started searching around for more shining new findings, so I also found an OSA2220GAA6CX (That I suspect to be a JH-F3 part due to the lack of a better candidate to be worth the new CX surfix. s7e9h3n should know more about what it truely is that I do though).
I also found too five Retail OPN parts (I wasn't able to find the Processor only OPN that is writted in the Heatspreader) in japanese shops though all them are out of stock. The ADO3800CZBOX (TDP 65W Energy Efficient Rev. JH-F3 A64X2 3800+? Same CZ surfix than the other ones), a shop that claims than the ADX600CZBOX is 125W TDP like I sayed before, and Single Core 65nm A64s with Retail OPNs ADH3200DEBOX, ADH3500DEBOX, ADH3800DEBOX that I recall reading than these were having a 45W TDP (It may be possible because the H in the prefix wasn't used before as far that I know), though I am unsure due to the lack of info and sources.

[Lightman]

What means this additional number ('0001') on new chips?? There also is 'S3' writing on 65nm samples.



Any ideas?

[zir_blazer]

Found some further info about Single Cores. Keep saying thanks to AsRock Not to forget online retailers that list Processors before they got them in stock.

Rev. DH-F3 (90nm, 62W TDP)
ADA3500IAA4DH-ADA3500DHBOX (2.2 GHz, 512 KB Cache L2)
ADA3800IAA4DH-ADA3800DHBOX (2.4 GHz, 512 KB Cache L2)
ADA4000IAA4DH-ADA4000DHBOX (2.6 GHz, 512 KB Cache L2)

Rev. DH-G1? (65nm, 45W TDP)
ADH3200IAA4DE-ADH3200DEBOX (2.0 GHz, 512 KB Cache L2)
ADH3500IAA4DE-ADH3500DEBOX (2.2 GHz, 512 KB Cache L2)
ADH3800IAA4DE-ADH3800DEBOX (2.4 GHz, 512 KB Cache L2)

From this list, we can make some further speculations:
_Considering the specifications of the ADA4000IAA4DH, AsRock messed up with the ADA4000IAA4CW ones and give it 2.6 GHz with 1 MB Cache L2 when it is 2.6 GHz with 512 KB Cache L2 as the OPN points out.
_The ADH surfix is for 45W TDP Processors, yet a new power consumption category to fit bewthem EE SFF (35W) and regular EE (65W).
Practically, we have pretty much all the upcoming AMD Processors (Both Single Core and Dual Core) models AND Cores releases. 90nm probabily with SiGe (DH-F3/DH surfix and JH-F3/CZ surfix, maybe CX too) and 65nm (DH-G1/DE surfix and BH-G1/DD surfix, not sure about the DL one).

Hail to the king baby, I'm not an AMD insider nor have friends with good IT jobs yet I leaked away and covered the info of almost all upcoming releases in the next month or two.

[Lightman]

Congratulations on your findings zir_blazer!

Now we know for what we should wait .
I'm pretty interested in 65nm DH-G1 revision. It is single core made on purpouse or cut down dual core?

Once again GREAT JOB!

[zir_blazer]

Final list, I won't be upgrading this anymore as no one pays attention.


Rev. DH-F3 (90nm, Single Core, 512 KB Cache L2 physically)
Athlon 64 3500+-----------ADA3500IAA4DH/ADA3500DHBOX (62W TDP, 2.2 GHz, 512 KB Cache L2)
Athlon 64 3800+-----------ADA3800IAA4DH/ADA3800DHBOX (62W TDP, 2.4 GHz, 512 KB Cache L2)
Athlon 64 4000+-----------ADA4000IAA4DH/ADA4000DHBOX (62W TDP, 2.6 GHz, 512 KB Cache L2)

Same that the old list, many online stores seems to have these in stock right now. No AMD Compare official info.


Rev. DH-G1? (65nm, Single Core, 512 KB Cache L2 physically)
Athlon 64 3200+ EE?------ADH3200IAA4DE/ADH3200DEBOX (45W TDP, 2.0 GHz, 512 KB Cache L2)
Athlon 64 3500+ EE?------ADH3500IAA4DE/ADH3500DEBOX (45W TDP, 2.2 GHz, 512 KB Cache L2)
Athlon 64 3800+ EE?------ADH3800IAA4DE/ADH3800DEBOX (45W TDP, 2.4 GHz, 512 KB Cache L2)

Same that the old list, many online stores seems to have these in stock right now. No AMD Compare official info.
ADH pretty much seems to be used as a one time only stopgap bewthem xDD (35W TDP) and xDO (65W TDP)


Rev. BH-G1? (65nm, Dual Core, 512 KB * 2 Cache L2 physically)
Athlon 64 X2 3600+ EE-------ADO3600IAA5DL/ADO3600DDBOX (65W TDP, 1.9 GHz, 512 KB * 2 Cache L2)
Athlon 64 X2 3600+ EE-------ADO3600IAA5DD/ADO3600DDBOX (65W TDP, 1.9 GHz, 512 KB * 2 Cache L2)
Athlon 64 X2 4000+ EE-------ADO4000IAA5DD/ADO4000DDBOX (65W TDP, 2.1 GHz, 512 KB * 2 Cache L2)
Athlon 64 X2 4400+ EE-------ADO4400IAA5DD/ADO4400DDBOX (65W TDP, 2.3 GHz, 512 KB * 2 Cache L2)
Athlon 64 X2 4800+ EE-------ADO4800IAA5DD/ADO4800DDBOX (65W TDP, 2.5 GHz, 512 KB * 2 Cache L2)
Athlon 64 X2 5000+ EE-------ADO5000IAA5DD/ADO5000DDBOX (65W TDP, 2.6 GHz, 512 KB * 2 Cache L2)

Wasn't able to find proof of existence of a real ADO3600IAA5DD, but I did with ADO3600IAA5DL. What the different surfix means is still to be know. They're all available on online stores. All these, exept the ADO3600IAA5DL, are in AMD Compare.


Rev. JH-F3 (90nm, Dual Core, 1 MB * 2 Cache L2 physically)
Athlon 64 X2 3800+ EE-------ADO3800IAA5CZ/ADO3800CZBOX (65W TDP, 2.0 GHz, 512 KB * 2 Cache L2)
Athlon 64 X2 5000+ EE-------ADO5000IAA5CZ/ADO5000CZBOX (65W TDP, 2.6 GHz, 512 KB * 2 Cache L2)
Athlon 64 X2 5200+ EE-------ADO5200IAA6CZ/ADO5200CZBOX (65W TDP, 2.6 GHz, 1 MB * 2 Cache L2)
Athlon 64 X2 5000+----------ADA5000IAA5CZ/ADA5000CZBOX (89W TDP, 2.6 GHz, 512 KB * 2 Cache L2)
Athlon 64 X2 5200+----------ADA5200IAA6CZ/ADA5200CZBOX (89W TDP, 2.6 GHz, 1 MB * 2 Cache L2)
Athlon 64 X2 5400+----------ADA5400IAA5CZ/ADA5400CZBOX (89W TDP, 2.8 GHz, 512 KB * 2 Cache L2)
Athlon 64 X2 5600+----------ADA5600IAA6CZ/ADA5600CZBOX (89W TDP, 2.8 GHz, 1 MB * 2 Cache L2)
Athlon 64 X2 6000+----------ADX6000IAA6CZ/ADX6000CZBOX (125W TDP, 3.0 GHz, 1 MB * 2 Cache L2)
Opteron 1210 HE---------------OSO1210IAA6CZ/Strange BOX OPN (65W TDP, 1.8 GHz, 1 MB * 2 Cache L2)
Opteron 1212 HE---------------OSO1212IAA6CZ/Strange BOX OPN (65W TDP, 2.0 GHz, 1 MB * 2 Cache L2)
Opteron 1214 HE---------------OSO1214IAA6CZ/Strange BOX OPN (65W TDP, 2.2 GHz, 1 MB * 2 Cache L2)
Opteron 1216 HE---------------OSO1216IAA6CZ/Strange BOX OPN (65W TDP, 2.4 GHz, 1 MB * 2 Cache L2)
Opteron 1218 HE---------------OSO1218IAA6CZ/Strange BOX OPN (65W TDP, 2.6 GHz, 1 MB * 2 Cache L2)
Opteron 1210------------------OSA1210IAA6CZ//Strange BOX OPN (103W TDP, 1.8 GHz, 1 MB * 2 Cache L2)
Opteron 1212------------------OSA1212IAA6CZ//Strange BOX OPN (103W TDP, 2.0 GHz, 1 MB * 2 Cache L2)
Opteron 1214------------------OSA1214IAA6CZ//Strange BOX OPN (103W TDP, 2.2 GHz, 1 MB * 2 Cache L2)
Opteron 1216------------------OSA1216IAA6CZ//Strange BOX OPN (103W TDP, 2.4 GHz, 1 MB * 2 Cache L2)
Opteron 1218------------------OSA1218IAA6CZ//Strange BOX OPN (103W TDP, 2.6 GHz, 1 MB * 2 Cache L2)
Opteron 1220------------------OSA1220IAA6CZ//Strange BOX OPN (103W TDP, 2.8 GHz, 1 MB * 2 Cache L2)
Opteron 1220 SE--------------OSX1220IAA6CZ/Strange BOX OPN (125W TDP, 2.8 GHz, 1 MB * 2 Cache L2)

Athlons 64 X2 comes from a current price cut list here, Opterons 1xxx are all listed in AMD Compare (Though AMD Compare doesn't even got the officially released A64X2 6000+, mind you). A64X2 are widely available on online stores, Opterons does not. The OSO1210IAA6CZ is a very good bet if it is at the right price. However, I can't understand the difference bewthem Retail versions, some finish in BOX, and others in WOF, and online stores list both, but until I am sure about the difference, I would't speak about it.
As a sidenote, Socket F Opterons also recived the JH-F3 Revision, Opterons 2xxx got a new CX surfix like I stated before (Older one was CQ) and Opterons 8xxx got a new CY surfix (Older one was CR).

[Lightman]

Final list, I won't be upgrading this anymore as no one pays attention.

Except me ....

Thank you for this comprehensive list zir_blazer .
I noticed most of this SKUs in UK stores, but only few 65nm X2 and 90nm X2 CZ are available right now.
Most of this new processors will be available to buy in UK in 3rd week of this month.

BTW have you find any info regarding new Turion processors on 65nm node?
I was looking for some time without luck. Supposedly Turions will come out as a tweaked/rebuild Rev. G2 of A64. That would mean even lower power consumption and higher overclocks .


Once again for you zir_blazer!

[Aerou]

I also thank zir_blazer, that info is truly interesting and really makes me wonder ... Are we going to see some of those CPU early and will they have SiGe used?