AMD limits Clawhammer to 800MHz

[Adul]

Link

Heye provided some guidance on the performance of the Clawhammer. Although the final speed in high performance systems which will be sold in December of this year is top secret, he said that sample CPUs in the hands of the mainboard makers are all locked at 800MHz. The reason for this, he said, was that he didn't want his partners overclocking chips and so providing "great expectations" which AMD probably couldn't meet.

 

[MrGrim]

That title is going to scare a lot of ppl.

 

[dullard]

That article had conflicting statements:
(1) "'great expectations' which AMD probably couldn't meet."
(2) "will be the performance leader and Intel had nothing to match it"

From 800 MHz, it would take a 150% overclock to meet the common 2.0 GHz rumor. Lets suppose that the overclocking gave a 100% speed boost (quite a large overclock in my opinion) then that would put it at 1.6 GHz which is scary if AMD probably can't meet it given 4-6 more months.

A PR3400+ Hammer should beat a 3.06 GHz P4 with 533 MHz fsb - but just barely. Thus the statement that Intel has nothing to match the performance is odd. It would mean that the Hammer is performing really well - but that conflicts with the comment that AMD probably couldn't meet a chip overclocked much beyond 800 MHz.

Oh well, I'm more confused than ever.

The pricing information did help though. A $500 processor barely can fit in the sub $1000 system category so the Hammer will probably be just about $500 (maybe a bit less).

 

[ISAslot]

Whoooooo hooooo!!!

This things gonna kick the Pentium II's a$$!!!!

 

[ElFenix]

still faster than itanium

 

[dullard]

Originally posted by: ElFenix
still faster than itanium

Do you have benchmarks that shows it beating the 1 GHz Itanium's score of 1350 in SPECfp2000? I'd love to see them.

 

[ElFenix]

Originally posted by: dullard

Originally posted by: ElFenix
still faster than itanium

Do you have benchmarks that shows it beating the 1 GHz Itanium's score of 1350 in SPECfp2000? I'd love to see them.

itanium is up to 1GHz now?!? i guess it isn't faster.... used to be 733... nevermind then.

 

[hudster]

Originally posted by: dullard

It would mean that the Hammer is performing really well - but that conflicts with the comment that AMD probably couldn't meet a chip overclocked much beyond 800 MHz.

I don't think they're saying that the chip will have a problem overclocking and/or gettting much beyond 800MHz. I think most likely that AMD purposely underclocked these processors to the extreme...so that no one would even bother trying to benchmark them, etc, at this point.

I think that's the whole point...as the article states "the final speed in high performance systems which will be sold in December of this year is top secret"...and that's exactly the point, it's top secret and AMD would like to keep it that way...for now.

just my $0.02


-hudster

 

[Soccerman]

From 800 MHz, it would take a 150% overclock to meet the common 2.0 GHz rumor. Lets suppose that the overclocking gave a 100% speed boost (quite a large overclock in my opinion) then that would put it at 1.6 GHz which is scary if AMD probably can't meet it given 4-6 more months.

umm.. if you don't want to be called a dumbass for responding like that, I would suggest you rethink things...

first of all, we haven't a CLUE how fast this thing will actually be, just talking about mhz here.

second, we don't know the actual POWER of the chip; ie, the amount of things that it can do per clock cycle (just look at what it was compared to: Itanium).

heck, we don't even know the amount of power this thing is sucking through that motherboard, the cooling it's using (most likely something fairly decent considering AMD still doesn't have any sort of proper heat spreader), the transistor count (should be below 80 million considering Parhelia can only hit around 220mhz with that many transistors at .15 micron) or even the die process it was manufactured on (most likely .13 seeing as T-breds are probably already being mass produced getting ready for their launch).

there is no way in hell you can tell anything from the 800mhz number other than what the guy said, which could be true (we all know AMD tries at least a little bit to keep things secret).

remember, AMD relies on XP numbers now, not mhz. performance, not marketing. I'm just wondering how sweet that onboard memory controller will be, as it nearly eliminates the whole FSB bandwidth (unless you have an 8 MB AGP card and some sort of sweet SCSI setup + Gigabit Ethernet) and allows the former northbridge to be placed differently (nearer to the AGP slot) while the CPU can move closer to the RAM, making latencies due to trace lengths shorter.

this is going to be one interesting architecture, you have to think a little bit different with it.. doesn't it reduce the pin count of the 'northbridge' (no 64 bit bus for the DDR SDRAM)? I think SiS might have had something going with their single chip solution, they just placed it on the wrong architecture! this way the devices connected to the southbridge don't have to go so far (through the northbridge AND the CPU) to get to the RAM.

 

[Nemesis77]

Originally posted by: Soccerman
heck, we don't even know the amount of power this thing is sucking through that motherboard, the cooling it's using (most likely something fairly decent considering AMD still doesn't have any sort of proper heat spreader)

Hammer has heat-spreader. All of them do.

the transistor count (should be below 80 million considering Parhelia can only hit around 220mhz with that many transistors at .15 micron)

You can't compare CPU's to GPU's. Todays fastest (in raw MHz) run at over 2000 MHz. todays fastest GPU's run at about 300Mhz.

 

[First]

Hey Adul, you want Brandon's job?

Btw, anyone who's interested in Clawhammer's cooling solution can read Kyle Bennett's coverage here.

 

[Athlon4all]

Hey Evan Lieb, you don't fool me with that fancy new Name!!! I know you're AGodSpeed!!!!

Anyway, I was just thinking that same thing. It seems its Adul who's getting all the news today. What is Brandon sleeping or something?

 

[Mem]

heck, we don't even know the amount of power this thing is sucking through that motherboard, the cooling it's using (most likely something fairly decent considering AMD still doesn't have any sort of proper heat spreader),

There`s an interesting article on Hammer cooling,here at

HardOCP ,very interesting read with pics.

 

[Adul]

Originally posted by: Evan Lieb
Hey Adul, you want Brandon's job?

Btw, anyone who's interested in Clawhammer's cooling solution can read Kyle Bennett's coverage here.

I think I would loose to him sooner or later

 

[BrunoPuntzJones]

Originally posted by: Mem

heck, we don't even know the amount of power this thing is sucking through that motherboard, the cooling it's using (most likely something fairly decent considering AMD still doesn't have any sort of proper heat spreader),

There`s an interesting article on Hammer cooling,here at

HardOCP ,very interesting read with pics.

We used to think that TBird and Palomino CPUs required a lot pressure when mounting a heatsink, but it will be nothing compared to what's coming. The Sample 2 shown here will utilize an incredible 40 pounds of force per square inch. What's even more surprising is the the Sample 1 shown below that's the "true" Thermaltake reference design calls for 70 PSI to be exerted.

Sure hope the finals won't use the clamp design. Suck move.

 

[dullard]

Originally posted by: hudster
I don't think they're saying that the chip will have a problem overclocking and/or gettting much beyond 800MHz. I think most likely that AMD purposely underclocked these processors to the extreme...so that no one would even bother trying to benchmark them, etc, at this point.

I think that's the whole point...as the article states "the final speed in high performance systems which will be sold in December of this year is top secret"...and that's exactly the point, it's top secret and AMD would like to keep it that way...for now.

Neither you nor Soccerman were willing to comment on the quote: "great expectations which AMD probably couldn't meet". What does that mean? Suppose they were given 800 MHz unlocked chips, instead of 800 MHz locked chips. These could be tested to find their maximum clockspeed with reasonable air cooling (lets pretend it is 1.7 GHz which with 4 more months to develop it will easilly reach 2 GHz). I realize that AMD doesn't want this information known quite yet so that is why they locked the chips. However that doesn't explain the AMD couldn't meet expectations part. If these factory samples can reach my pretend 1.7 GHz, why couldn't AMD produce chips by the end of the year to match this speed? That is the point of my post, what expectation will AMD not meet?

800 MHz was chosen for a reason. Of course they underclocked them - either a little or a lot. Underclocking helps guarantee it will not overheat (that would be bad publicity) and it keeps the true speed unknown - both are marketing tactics. The only reason AMD will hide the speed at this late date is for marketing reasons - 4 months isn't enough time for Intel or someone else to suddenly develop a chip that will beat AMD. So since 800 MHz was chosen for marketing reasons, why was 800 MHz chosen? Both 500 MHz and 1000 MHz would accomplish the same goals (preventing overheating and preventing the true value from being known). However doesn't a 1 GHz sample sound much better? First there is the psychological 1 GHz level, plus it is the same speed as the Itanium 2 to be released any day now. Since 800 MHz was chosen for marketing reasons, wouldn't 1 GHz have been a better option? Unless that is where the failure to meet expectations comes into play.

It is perfectly possible that the Hammer will top out at a low MHz and yet provide great performance. You'd have to be quite computer ignorant to think otherwise. MHz doesn't mean performance - but it doesn't help marketing either. The PR numbers are for marketing only - they don't affect the performance of the chip.

 

[majewski9]

Exactly! I dont think the hammer will have any problem crashing the 2 ghz barrier which it is suppose to debut at. Their first model should be the 3400+ which we all know means it will perform better than a p4 clocked at 3.4 ghz. I know it is suppose to be compared to the Athlon at that speed but come on!

AMD Should have no problem ramping up clock speed!

I am alittle bit wary about the onboard mem controller. A lot of people are criticizing it for limiting memory bandwidth. I know it has a DDR controller intergrated and this could limit mem bandwidth to 2700 with DDR333. will different chipset allow for DDR 400 or even DDR II or can they work with hammer's mem controller? Is memory bandwidth all that important? I mean RDRAM has tons of memory bandwidth, but we know that DDR SDRAM and the right chipset can perform better even though it has less memory bandwidth. Another point is onboard graphics are going to suffer from the onchip mem controller.

Also why does he say late december? I mean Hammers launch is suppose to be october and every press release and roadmap concurs with that. Does this mean AMD is hinting for a delayed launch?

 

[Elcs]

Apologies to anyone who read my post. Im having a really really bad day. Just forget Im here.

 

[majewski9]

yeah I forgot we were talking about clawhammer which is suppose to debut alittle later, but is AMD even going to be able to release hammer in october?

 

[dullard]

Also why does he say late december? I mean Hammers launch is suppose to be october and every press release and roadmap concurs with that. Does this mean AMD is hinting for a delayed launch?

That is 6 months away, an eternity in the computer world. Any release date this early needs to be taken with a grain of salt. One minor chip glitch, driver problem, marketing problem, etc could delay a release significantly. Late last year many rumors said throughbred will be out 1st quarter 2002 (hinting at March). It is now June. Wasn't the Athlon also delayed - as well as some other AMD chips? My point is that the rumors are never that accurate. I'm a pessimist and I always add 2-3 months to the rumored release date (which is the very reason I posted 4-6 months above). I wouldn't be at all surprized with an October launch or even a January launch.

 

[joohang]

Originally posted by: Adul

Originally posted by: Evan Lieb
Hey Adul, you want Brandon's job?

Btw, anyone who's interested in Clawhammer's cooling solution can read Kyle Bennett's coverage here.

I think I would loose to him sooner or later

Yes I know. You need to work on your spelling, woman.

 

[BD231]

Originally posted by: dullard

Also why does he say late december? I mean Hammers launch is suppose to be october and every press release and roadmap concurs with that. Does this mean AMD is hinting for a delayed launch?

That is 6 months away, an eternity in the computer world. Any release date this early needs to be taken with a grain of salt. One minor chip glitch, driver problem, marketing problem, etc could delay a release significantly. Late last year many rumors said throughbred will be out 1st quarter 2002 (hinting at March). It is now June. Wasn't the Athlon also delayed - as well as some other AMD chips? My point is that the rumors are never that accurate. I'm a pessimist and I always add 2-3 months to the rumored release date (which is the very reason I posted 4-6 months above). I wouldn't be at all surprized with an October launch or even a January launch.

Well judging from the past, it's been clear that AMD makes chips for their main buyers like Compaq/HP before they sell to the public. Since they can only make so much at a time, it's obvious why we dont see chips for a good while after their release. Once AMD satisfies their main buyers chip needs, they start shipping chip's to everyone else. I'm sure there will be a huge demand on the Hammer, so you can bet that we will be waiting for a while unless AMD is already pumping out Hammer's as we speak.

 

[Rand]

I still remember initial previews by Anand, and FiringSquad of the classic K7 Athlon at 500MHz getting it's ass kicked by Katmai PIII's at 500MHz... when it finally introes the 500MHz Athlon competes extremely well with the 600MHz Katmai.

After that little fiasco, I wouldnt be at all surprised if the suddenly told us the Hammer was running at 100MHz

Edit: Quite the number of ClawHammer clockspeed rumours popping up of late, first MSI's 1.6GHz ClawHammer at Computex online advertisement, now this.

 

[RaynorWolfcastle]

My Hammer prediction:

Until AMD finally releases it, there will be a 10-20% increase in thread locksdue to flame-fests

-Ice

 

[Pariah]

There may be more to this story than AMD limiting the clockspeed to 800Mhz.

AMD concedes poor performance of early Hammer processors

I don't think this has anything to do with clockspeed. Apparently AMD is confirming the rumours that an internal memo sent last week stated that early performance results don't look good. If they want an end of year release they better get cracking on optimizations.