AMD Zambezi news, info, fans !

[BeepBeep2]

Define stable?

What you consider stable and what AMD is diff.

Add 4 way gpu to the mix and 3d benches will crash at specced voltage.

Last but not least they're HT bus is more than ample and does not show massive gains except in IGP clocking due to the fact that the HT bus is not saturated even in heavy 4 way crossfire.

Okay? I already know there are no gains...but I was wanting to show that over 2700 is possible I run a midrange board with 2x16 slots, no way to do 4 way anyway unless I had a pair of 6990.

Did I consider 30 minutes of P95 stable? No...not at all.

[XRL8]

Ithink, we will see some diferents. UEFI bios. No, seriously, I dont know how about Turbo core 2.0 and Cross IV and how about some OC potential against 990FX+AM3+. We mus waiting at Computex or E3 (hope for launch this days)

UEFI doesnt have anything to do with socket , it doesnt give more performance either.
Turbocore is just a gimmick, just like intels turbo, you can clock cores individually manually.It doesnt change the IPC or bandwith.
OC potential of the board doesnt have anything with the socket too, its about build quality and engineering not socket revision.
I have a feeling that SB950 will have somewhat improved SSD compatibility and trim support, maybe faster regular HDD too.However northbridge looks to be a copy of 890FX.

[FlanK3r]

UEFI was my joke (smile-need better to read).
Turbocore is default value of performance CPU and turbo at Zambezi will be really impressive.

[XRL8]

UEFI was my joke (smile-need better to read).
Turbocore is default value of performance CPU and turbo at Zambezi will be really impressive.

You should used this smile
As for the turbo,it isnt really useful for enthusiasts/overclockers.But yes it has the potential to boost somewhat stock performance.As for the usability of it on zambezi ,we dont know squat as of now,so how do you know its "impressive" ?.
Turbocore 1.0 was pretty disappointing tho.

[chew*]

Okay? I already know there are no gains...but I was wanting to show that over 2700 is possible I run a midrange board with 2x16 slots, no way to do 4 way anyway unless I had a pair of 6990.

Did I consider 30 minutes of P95 stable? No...not at all.

You could run 24 hours of it or 365 days of it and it would mean nothing because you are taxing the HT link the wrong way.

In order to test something for stability you need to know what hits a nerve on it and how to hit that nerve.

As far as knowing whats possible we all know what can be done from previous IGP comps however passing 3d 01 doesn't define stability considering having no vga in a slot takes alot of stress off HT link.

[stangracin3]

Define stable?

What you consider stable and what AMD is diff.

Add 4 way gpu to the mix and 3d benches will crash at specced voltage.

Last but not least they're HT bus is more than ample and does not show massive gains except in IGP clocking due to the fact that the HT bus is not saturated even in heavy 4 way crossfire.

chew that depends.
if you have a locked processor overclocking the bus gets you a nice increase in speed, but if you have locked you're right.

[chew*]

chew that depends.
if you have a locked processor overclocking the bus gets you a nice increase in speed, but if you have locked you're right.

We are talking about HT link not bus though.

aka 3800 HT speeds not 400 HT bus.

[FlanK3r]

weknow this: server Bulldozer boost for all cores is +500 MHz boost and with half or lower cores is higher. This mean maybe +700 MHz? And think +500 MHz boost at 16 cores of INterlagos is impresive.

[crazydiamond]

the.wait.is.killing.me!

[chew*]

weknow this: server Bulldozer boost for all cores is +500 MHz boost and with half or lower cores is higher. This mean maybe +700 MHz? And think +500 MHz boost at 16 cores of INterlagos is impresive.

500 mhz boost is great but what is the nominal clock?

[XRL8]

weknow this: server Bulldozer boost for all cores is +500 MHz boost and with half or lower cores is higher. This mean maybe +700 MHz? And think +500 MHz boost at 16 cores of INterlagos is impresive.

We also know that server bulldozer doesnt need new socket nor new chipsets to utilize this new turbo.
I dont however think there is any hard data about 16 core chip having 500Mhz boost at all cores, it would just mean that it runs 500mhz faster
Youre making assumptions based on rumours.Thats not healthy.Better to expect less and get more.

[Formula350]

for me is AM3+ socket now the next future. Need max performance, not compromise...Remember, AMD had long,long, long compatibility. They can not be still on one socket. ANd still we have part compatibility (AM3 CPU to the AM3+ socket). Look at Intel. Last years again and again new LGA ...

I love AMD, and also fully think that Intel changing sockets so much is not the best, but the difference is probably from pushing tech further. While I know 'Dozer will be quite a leap, keeping the same socket I think is going to mean making some compromises (AM3+ or not) :\ Had they gone with a new socket there might have been more performance to be had thanks to addition pin counts, but only way to tell would be to do something like take an 8core C32 and 8core AM3+, then pit them against each other. I know that server sockets have a bit more in terms of capabilities (HT link is tweaked), but the server chips (SR5792 or something) is basically an 890FX

Flanker, AM3+ IS in 99% percent the same socket as AM3.They messed around with one pin to have less headache support wise.It uses the same architecture,same HT topology, same chipsets,and is compatible with am3 cpus.So its the same damn thing.Theres no new breakthrough technology behind it.
Maybe youre talking about 9xx chipsets and not am3+ socket alone.However, it looks like at least northbridge will be exactly the same, we will see about southbridge.
In essence, with 90% probability, there wont be any difference performance wise if you plug the chip to crosshair IV or a 990FX board.

Look at it a different way and AM3+ is 99% the same socket as Socket 939/940! That's a LONG time ago lol I think that is exactly the reason why ASRock was able to make a motherboard for Socket 939 that used the 785G and 790GX NB chip lol If I knew for a fact either of my S939 CPUs worked, I'd so get one of those boards

We are talking about HT link not bus though.

aka 3800 HT speeds not 400 HT bus.

So everyone knows (to clear up confusion). The HyperTransport is HT, and the reference clock bus is HTT. So stock speed is 2000MHz on the HT Link, and 200MHz on the HTT

We also know that server bulldozer doesnt need new socket nor new chipsets to utilize this new turbo.
I dont however think there is any hard data about 16 core chip having 500Mhz boost at all cores, it would just mean that it runs 500mhz faster
Youre making assumptions based on rumours.Thats not healthy.Better to expect less and get more.

I think the first round Interlagos/Valencia run at 1.8GHz (might be 1.6) for the lowest model, so 500 or even 700MHz I don't see as being a stretch of the imagination, even on all 16 cores. A lot of these server chips are meant to conserve power, (as a result) run cooler, but still be a power-house for a processor. All of that seems to be nooo problem for the 'Dozer architecture

So, as crazydiamond said: "the.wait.is.killing.me!"


While this is the 'Dozer thread, I think Bobcat shouldn't be forgotten either since it's rather similar (Maybe call it BullCat? >_> Hey... it's better than BobDozer!).
Here's the Socket FM1 for Llano.
http://www.cpuforever.com/showthread.php?tid=896
Though the person who counted pins is off by one, as there are 13 missing corner pins, not 12. That leaves us with a "Socket 905". Seems as though S1_G4 (S1G4) socket is only 638pin though o_0 (for the Danube chips, like the Phenom II mobiles). That's quite a number of addition pins. Presumably for the integration of Northbridge (PCIe lanes), IGP and anything else I'm not thinking of. Hopefully Llano has more PCIe lanes than Zacate, which I figured had 16@2.5GT/s, or 8@5GT/s available. Based on Sandra's report for this Sapphire board-- x4@5 to the PCIe slot, x1@2.5 to the NIC, x1@5 to the USB3, x4@2.5 to the FCH, and x1@2.5 to the mini PCIe slot (which isn't shown due to not being populated, but I could toss a WiFi in it).

[XRL8]

Look at it a different way and AM3+ is 99% the same socket as Socket 939/940! That's a LONG time ago lol I think that is exactly the reason why ASRock was able to make a motherboard for Socket 939 that used the 785G and 790GX NB chip lol If I knew for a fact either of my S939 CPUs worked, I'd so get one of those boards

Yup,sockets change minimally last few years on AMD, but in those cases, you were moving from DDR1 and after that DDR2, and HT was upgraded few times.AM3+ doesnt bring any notable differences, so i dont see flankers viewpoint that AM3+ will be "faster" socket.
As for the server BD`s, if a chip has a stock freq of 1.8ghz, idles at 1ghz, but at load is running ALL CORES at 2500mhz, uhm, isnt that just 2500mhz chip?
I think turbo will be alot better then on X6, thats for sure.
But you have to know that turbocore 1.0 sucks :/ .And there wasnt any specific hard info/leaks about BD so we dont know anything SOLID on TC 2.0, which is infuriating, as samples ARE OUT.Cmon peeps, give us a teaser ;-D

[Mats]

So it's about 7 weeks to launch, and no useful performance leaks yet?
That's nothing short of impressive. I mean, how is that even possible these days?

[FlanK3r]

We also know that server bulldozer doesnt need new socket nor new chipsets to utilize this new turbo.
I dont however think there is any hard data about 16 core chip having 500Mhz boost at all cores, it would just mean that it runs 500mhz faster
Youre making assumptions based on rumours.Thats not healthy.Better to expect less and get more.

heh, with socket for server u has right. What I know, with "older" desktop socket need 16MB flash ROM to update and al active socket pins (electricaly). Am3+ is ready now and supporting from AMD side. This is better choice for the future and again, with refresh at Komodo, AM3+ will be required (my speculation).
But back to the infos. About 500 MHz boost:
http://blogs.amd.com/work/2011/01/31...er-goes-to-11/

Some of the benefits of AMD Turbo CORE include:

* Up to 500MHz of additional clock speed available with all cores active. This means even with 16 cores active with server workloads, all cores can boost at the same time. For those customers that want to maximize their performance, they now have the tools to do it.
* Even higher boost states available with half of the cores active. We’re not stating exactly how high processors can boost with AMD Turbo CORE, but obviously if there is room for up to 500MHz with all cores active, fewer active cores would obviously mean less power, and more headroom to recapture with AMD Turbo CORE. At launch you will see processors marketed with a base and a maximum frequency, base will reflect the actual clock speed on the processor and max will reflect the highest AMD Turbo CORE state.
* AMD Turbo CORE is deterministic, governed by power draw, not temperature as other competing products are. This means that even in warmer climates you’ll be able to take advantage of that extra headroom if you choose. This helps ensure a max frequency is workload dependent, making it more consistent and repeatable.


I think, for desktop can be higher than for servers! Think, if we will see "conservative" desktop boost, will be the same as in servers.
So this is 500 MHz at 8x cores and with half modul or one modul will be about +700 MHz boost.
My tip about default clocks are about 3-3.5 GHz for FX (without turbo).

[Manicdan]

So it's about 7 weeks to launch, and no useful performance leaks yet?
That's nothing short of impressive. I mean, how is that even possible these days?

the only logical answer is that its not ready and will be delayed so long that skynet will take over before we see any benchmarks

[zalbard]

the only logical answer is that its not ready and will be delayed so long that skynet will take over before we see any benchmarks

Not necessarily. They could just be tight-lipped. Feel free to refer to 69x0 launch.

[ChanceCoats123]

Not necessarily. They could just be tight-lipped. Feel free to refer to 69x0 launch.

I agree with this. Amd set a pretty damn high NO-Benchmark-results bar, this time around.

[XRL8]

heh, with socket for server u has right. What I know, with "older" desktop socket need 16MB flash ROM to update and al active socket pins (electricaly). Am3+ is ready now and supporting from AMD side. This is better choice for the future and again, with refresh at Komodo, AM3+ will be required (my speculation).
But back to the infos. About 500 MHz boost:
http://blogs.amd.com/work/2011/01/31...er-goes-to-11/

Some of the benefits of AMD Turbo CORE include:

* Up to 500MHz of additional clock speed available with all cores active. This means even with 16 cores active with server workloads, all cores can boost at the same time. For those customers that want to maximize their performance, they now have the tools to do it.
* Even higher boost states available with half of the cores active. We’re not stating exactly how high processors can boost with AMD Turbo CORE, but obviously if there is room for up to 500MHz with all cores active, fewer active cores would obviously mean less power, and more headroom to recapture with AMD Turbo CORE. At launch you will see processors marketed with a base and a maximum frequency, base will reflect the actual clock speed on the processor and max will reflect the highest AMD Turbo CORE state.
* AMD Turbo CORE is deterministic, governed by power draw, not temperature as other competing products are. This means that even in warmer climates you’ll be able to take advantage of that extra headroom if you choose. This helps ensure a max frequency is workload dependent, making it more consistent and repeatable.


I think, for desktop can be higher than for servers! Think, if we will see "conservative" desktop boost, will be the same as in servers.
So this is 500 MHz at 8x cores and with half modul or one modul will be about +700 MHz boost.
My tip about default clocks are about 3-3.5 GHz for FX (without turbo).

Theres no future in AM3+ ,at least according to AMD.Komodo will have PCIE integrated and require new socket.AMD hasnt put out ANY info about "future" of the am3+.
As for the turbo, yes youre right it seems ,kind of a smart turbo i see.However if its determinded by the TDP (same technology as in cayman?) on the desktop segment i would not expect more, because desktop parts are clocked a lot faster from the get go.Unless youre thinking we will get like 5ghz on turboed chip.Which i find hard to believe, remember that BD has longer pipeline and higher base clock than the PHII.

[blindbox]

.However if its determinded by the TDP (same technology as in cayman?)

With Bulldozer, the chip tries to hit the TDP. i.e., when there's load and it's single threaded, that one, single core will clock as high as possible to hit the chip's TDP. Only time it's on stock clock is when the load is heavy enough (IBT perhaps).

Cayman strives to NOT hit the TDP - downclocks if TDP reached, but doesn't overclock if TDP isn't reached.

It's pretty much Cayman's power control but it works in the opposite direction


This is pretty awesome. I wonder how they'll pull it off with window's thread jumping that pretty much killed Phenom I's CnQ and performance.

[FlanK3r]

I think, maxboost can be 4.5 GHz with one or 2 modules active. But Im a bit spectical-I think, will be up to 4.2 GHz :-)

[Manicdan]

Not necessarily. They could just be tight-lipped. Feel free to refer to 69x0 launch.

it was just sarcasm, i have faith we will see them all launch early june

[Formula350]

it was just sarcasm, i have faith we will see them all launch early june

While I know this is probably THE worst place to read a rumor from, I was on eBay last night looking at C32/G34 boards (Valencia 8C is C32 only) and the one seller's product description for this motherboard/dual 6100 series/64GB HyperX DDR3 said the board was "Interlagos Ready" for the May 22nd release date..

Obviously take it with a grain of salt, but interesting none-the-less

[demonkevy666]

500 mhz boost is great but what is the nominal clock?

3.5ghz is said to be the top bin 8 core chip.

So it's about 7 weeks to launch, and no useful performance leaks yet?
That's nothing short of impressive. I mean, how is that even possible these days?

NDA's...

Theres no future in AM3+ ,at least according to AMD.Komodo will have PCIE integrated and require new socket.AMD hasnt put out ANY info about "future" of the am3+.
As for the turbo, yes youre right it seems ,kind of a smart turbo i see.However if its determinded by the TDP (same technology as in cayman?) on the desktop segment i would not expect more, because desktop parts are clocked a lot faster from the get go.Unless youre thinking we will get like 5ghz on turboed chip.Which i find hard to believe, remember that BD has longer pipeline and higher base clock than the PHII.

hhhmm maybe like how K10.5 is llano cores it will be new socket but won't over come the highest chips without GPU, because they'll be using cards instead of on die gpu.

I think, maxboost can be 4.5 GHz with one or 2 modules active. But Im a bit spectical-I think, will be up to 4.2 GHz :-)

I'm thinking this going to be a 8 thread vs 8 cores fight :-/

as for the 990FX better SSD not likey untill the change A link III from 1.0 pci-express to to 2.0 the link from South bridge to northbridge is a pci-express 4x link. in 1.0
they could be able change it to a hyper tranport link to each other, idrk if thats possible.

[Kobaltrock]

So it's about 7 weeks to launch, and no useful performance leaks yet?
That's nothing short of impressive. I mean, how is that even possible these days?

NDA's...

It isn't only about NDA's, since those get broken / leaked all the time.
It looks like they are only giving out info to people who they really trust to abide by the NDA.

While it might be good for AMD, you can bet that Intel already has working samples and are going to do something when BD launches--if it is good.