Sign in with
Sign up | Sign in
Your question

what will AMD do?

Tags:
Last response: in CPUs
Share
August 11, 2004 12:17:28 PM

the K8 arc is very good. but what will they do when they can't get clock speeds any higher. because the core looks like it can't be tuned anymore. so all they can do is ramp clock speeds up. So intel might have been the clever ones here. gaining clock speed first then tuning. whats everyone think?

More about : amd

August 11, 2004 12:53:28 PM

Both companies will make dual-core chips by 2005.

That's not the complete answer, but that's a good part of it. It is, indeed, getting increasingly harder to increase clock and IPC...
August 11, 2004 2:16:56 PM

Have you seen this post:
<A HREF="http://www.aceshardware.com/forum?read=115094143" target="_new">http://www.aceshardware.com/forum?read=115094143&lt;/A> ? I think the real question is: what the hell will <b> intel </b> do ? 152W maximum power consumption, ouch !!

AMD is still competing very well (not to say trouncing intel) using "last years" 130nm process, they still have their 90nm trumph. They have also reduced their L2 cache size, so if everything else would fail, especially at 90nm, they could easily up it again. Its likely at least Opteron will be seeing some bigger cache (2 Mb or so), so it shouldn't be hard to move this down to the A64/FX line if somehow required.

But I don't think it will be necessary; by this time next year, I'd except some 3-3.2 GHz A64's, which is still 25-33% faster than today. I seriously doubt intel will be able to match that with Prescott, a similar increase would put it at 4.5 -4.8 Ghz.. doesnt seem likely or even remotely possible unless they can really do something spectaclular about Prescotts heat output (which is not likely on its own, as intel has publically abandonded the speedracer design in favor of hgher IPC/multicore approach for the future because of this reason).

Both companies will go dual core. intel perhaps more as a mainstream solution (out of necessity ?), while for AMD I expect only the opteron and FX range to go dual core. If intels dual core solution is Prescott based (which I expect, at least initially), it will mean it will be seriously power limited, probably not faster, if not lower clocked than todays mainstream P4s and therefore, relatively sucky at single threaded performance (still most desktop software today). If it is Dothan based, it remains to be seen how high they can clock it to be competitive with K8/P4, but unlike most here, I'm not expecting clock speed miracles from this improved Pentium Pro core. There is just no way it will clock as fast as K8 on the same process node.

So.. ? what will intel do ? Probably spend fortunes on getting the entire world to move to multithreaded apps, tell us we don't need such fast cpu's anyway, and diversificating their business model by selling more TVs, projectors and Intel Bunny keyhangers. Time to sell your intel stock IMO. I think intel made a serious judgement error 5 or so years ago by expecting Itanium would take over the world, and their plan-Bs (netburst, Pentium Pro) just don't have the legs to keep up with AMD.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
Related resources
August 11, 2004 2:22:26 PM

Well, it would seem they have a cooling problem. Prescott, in its E0 revision, will probably be able to clock as high as 4+Ghz without any problem except lotsa heat...

A dual-core prescott would certainly require water cooling. In any case imaginable, dual-core is more of a P-M's job. Water cooling is enough for up to 300+W !!!

If not, Intel better start including some truly advanced HSFs as stock coolers. Maybe even water cooling kits. Heck, in the EE, this would even <i>justify</i> the EE price tag, which is as of yet not explicable.

Intel is bound to find a better way to do things; they'll burn market share and image until they do.
August 11, 2004 2:31:12 PM

You say that you expect some 3Ghz+ A64's *drool* is this with or without the dual core?
Is it ambitious to think we will have dual core, 90nm, 3Ghz, 2MB Cache by 1Q of 2006? Or will dual cores be 130nm and slower? or what is known?

What do you call a group of midgets? A pack, gaggle, a pride, or maybe just a murder?
August 11, 2004 2:35:27 PM

Both dual cores will be based on 90nm. It is likely that they'll clock a little lower too, at least when introduced. Not much more is known, though.
August 11, 2004 2:35:34 PM

> Prescott, in its E0 revision, will probably be able to
>clock as high as 4+Ghz without any problem except lotsa
>heat...

I think even current Prescotts should be able to do that, or at least close to that, if somehow intel could drastically reduce power consumption/leakage. maybe they should ask AMD to fab one using their SOI process ? God knows how good it would be :) 

anyway, what I wanted to say, we are now beyond the point where heat is just a nuisance. 150W of peak power (170W/cm²)and 150+ (!) ampere are just incredibly hard to deliver and cool. Nothing is impossible, but many things are economically not feasable for mass market. 3.6+ GHz prescotts, 150+A motherboards and watercooling may well be some of those things. Especially when you have to compete again a product that peaks roughly 50-100% lower than those values. So intel may release some 'extreme' editions of their chips, that work with very highend motherboards, PSU's and (water)coolers, but none of these are option for the mainstream users that buy $500 PC's. That is where intels major future problem lays, and perhaps where (dual core) Dothan could provide an adequate anwer.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
August 11, 2004 2:42:02 PM

Quote:
the K8 arc is very good. but what will they do when they can't get clock speeds any higher. because the core looks like it can't be tuned anymore. so all they can do is ramp clock speeds up. So intel might have been the clever ones here. gaining clock speed first then tuning. whats everyone think?

AMD, Intel, and I'd imagine pretty much <i>every</i> high-end CPU manufacturer has these problems. The answer is usually to use better manufacturing materials and techniques to further the limits of the clockspeed. Silicon germanium, strained silicon, low-k dielectrics, high-k gate dielectrics, depleted substrate transistors, silicon-on-insulator, smaller process technologies, and so on can all be used to make processors more energy efficient (and in turn to run cooler), allowing for even higher clockspeeds.

<pre><b><font color=red>"Build a man a fire and he's warm for the rest of the evening.
Set a man on fire and he's warm for the rest of his life." - Steve Taylor</font color=red></b></pre><p>
August 11, 2004 2:44:57 PM

>ou say that you expect some 3Ghz+ A64's *drool* is this
>with or without the dual core?

Without.

>Is it ambitious to think we will have dual core, 90nm,
>3Ghz, 2MB Cache by 1Q of 2006?

Optimistic for sure, but not entirely impossible.

> Or will dual cores be 130nm and slower? or what is known?

Dual cores will definately NOT be 130nm, they will be 90nm from both intel and AMD. And its safe to assume dual core chips will clock lower than their single core brethern because of power consumption and delivery issues. Dothan might be an exception to this, but dual core Prescott will be overwhelmingly power limited, and A64/FX at least somewhat IMHO.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
August 11, 2004 3:13:40 PM

Agreed, you do have a point. You exaggerate somewhat, however, when you say
Quote:
Especially when you have to compete again a product that peaks roughly 50-100% lower than those values.

100% lower is nil, of course, which is an absurd. You could say 40-60% lower, or even, say, 50-80% lower, but don't say 100% lower. Even if it is roughly. (worse than that, only "roughly 50-150% lower"... :smile: ). Just being picky here, sorry.

In all other points, I agree. The problem is not giving the high-end consumer that <i>can</i> buy almost anything interesting you throw at him, including a 4+Ghz 64-bit enabled dual prescott with water cooling and an ultra-high-end mobo to give it the hundreds of amps needed. The problem is delivering to the masses.

(note the hundreds of amps figure - that is a lot. MRI scanners, for instance, use 400A for their superconducting magnets. Not many electronic devices exist that need more current than that, let alone one delivered so sustained and "delicately" as required of mobos and PSUs.)

And multicore dothans or tweaked dothans are much, much, <i>much</i> saner designs than multicore prescotts. That much is obvious.
August 11, 2004 3:18:34 PM

>100% lower is nil, of course, which is an absurd

I meant to say the other is 50-100% higher, even though that may indeed be an exagerration on its own. I believe AMD's TDP is roughly worst case, so you'd be comparing 89W to 150W, so AMD is 59% lower or intel is 68% higher :)  Or not.; whatever LOL :) 

= The views stated herein are my personal views, and not necessarily the views of my wife. =
August 11, 2004 4:43:00 PM

In a few years we will need to connect a molex to the cpu...

Athlon 2700xp+ (oc: 3200xp+ with 200fsb)
Radeon 9800pro (oc: 410/360)
1024mb pc3200 (5-3-3-2)
Asus A7N8X-X
August 11, 2004 5:03:28 PM

On this power consumption issue - the only way intel could get away with prescotts for the $500 pc buyer would be if they just kept quiet! No one would buy a pc if they knew their energy bills would be 10% higher, especially if they're buying $500 pcs.
Also, no one really wants to buy a pc which is uncomfortably loud - which I suppose prescotts would be in dual core form - the fan noise of my pc isn't bad, but I notice it, and its only an xp2000!

Dothan seems to be the only solution for intel atm - and its a good one! If a cpu can use less power andd match (or as near as to make no difference) a p4 or A64, run coooler too, then that's gotta be the magic answer to their problems imo - it beats the A64 clock for clock (if they could ramp up the mhz to 2.4 ghz, which I'm not sure about...) but in dual core situation, what do you guys think? A dual A64 3200 vs a dual Dothan 1.7ghz, who's out on top?

XP2000, 512 ddr 2700ram, GF4 MX440, XP Pro
August 11, 2004 5:15:58 PM

I think that's a very tough comparison. The Dothan is not equipped with all the bells and whistles. So I'm guessing a dual-core A64 3200+ (@2.0Ghz) would kill a dual-core Dothan@1.7Ghz. Not so sure about a dual Dothan 2Ghz, though.

But things would be different if Intel tweaked the dothan core for desktop operation. Because we don't know dothan architecture at all (it's somewhat secret), we cannot ascertain what would happen to a dothan core if it were, say, equipped with 1066Mhz FSB and dual DDR2-533 or DDR2-667. All we can say with good certainty is that it would run much cooler.

I hope, for Intel's sake, that Smithfield is not a pure netburst offspring. I'm willing to risk saying that a 2.4Ghz Dothan is not an impossible beast, if the core gets tweaked. If that's not 24x100, but rather 18x133, then each core would get roughly and crudely 533Mhz bandwidth each (1066Mhz total)... or, heck, even better, if one core didn't need the bus, the other would get the full 1066Mhz!!!

This would even negate (or at least terribly diminish) the performance penalty of a shared bus amongst the cores!
August 11, 2004 5:21:26 PM

Quote:
note the hundreds of amps figure - that is a lot

Hundreds of Amps?. I think we were talking 100-150 WATTS, since no PSU on earth can provide that many amps. Still, a good point is made - if you CPU itself is sucking 8-12 Amps off the 12V rail, that leaves you with precious little amps for the rest. A proc like that would have to siphon off additional power from one of the rails, or else we'd see insane requirements like a 550W PSU for the new intel procs.

"It's too late now anyway. That song is stuck in my head and the only way to get rid of it is to blow it out. With a bullet!! - Carl
August 11, 2004 5:25:36 PM

Hm, yes, hundreds of Watts, however, you have to consider that the voltages are low. So, with 1.2 - 1.4 volts, a 150W maximum thermal dissipation exceeds 100 amps. In a single processor configuration.

However, when going multi-core, you'll probably exceed that 100 amps figure... That's why I mentioned a number in the hundreds.

Where's dothan to save us from the gazillion amps, Intel?
August 11, 2004 5:27:26 PM

Sorry - didnt make myself that clear, I didnt mean dothan in its present state, I meant once it was 'primed for battle' ie with the bells and whistles and so on.

XP2000, 512 ddr 2700ram, GF4 MX440, XP Pro
August 11, 2004 5:32:05 PM

Ah yes, a dothan primed for battle (with Em64T enabled, please) should give AMD some cause for concern. 1066Mhz FSB, NX, 64-bit, 2.5+Ghz clock if possible - or whatever, and so on...... Unfortunately, none of us has any really solid idea of the platform's potential as a desktop part (or the lack thereof).

However, it has to be primed for battle before AMD gets their piece together and churns out higher-clocked 90nm A64s, or else it won't have any effect on the tides of battle!

<P ID="edit"><FONT SIZE=-1><EM>Edited by Mephistopheles on 08/11/04 04:33 PM.</EM></FONT></P>
August 11, 2004 9:23:14 PM

> it beats the A64 clock for clock (if they could ramp up
>the mhz to 2.4 ghz, which I'm not sure about...) but in
>dual core situation, what do you guys think? A dual A64
>3200 vs a dual Dothan 1.7ghz, who's out on top?

Two things:
1) Actually, A64 is slighly faster than Dothan clock per clock if you compare A64's typical (desktop) configuration (ie not using single channel on a 512 Kb dual channel A64) with Dothans typical (mobile) configuration (slow FSB, single channel) unlike what some hardware sites did recently. So be carefull not to think "A64==Dothan clock per clock and if we up the FSB it will be faster". No if you up the FSB and give it dual channel, it will probably match A64 (or if you prefer, if you castrate A64 to more or less match Dothans features, they perform about equal).

2)Second point is clockspeed. Its hard to guess how high Dothan could clock when you liberate it from its low thermal enveloppe. One thing is for sure, clockspeed will not be determined by its power draw (like Prescott), but it seems most of us have forgotten that in the past, cpu's where almost never thermally limited. it wasn't until Tbird or so, that it even became a concern. Dothan's core is a tweaked Pentium Pro/2/3, no matter how you slice it. They added SSE2, a P4 FSB as well as some other, but mostly minor performance tweaks and lots of power consumption tweaks. The pipeline and the execution units are almost a carbon copy of the P3 from what I've seen.

Not that this is necessarely bad at all (this is after all, IMO the best core Intel ever produced), but I fear it has its limits. Athlon could clock (significantly) higher than P3 at a given process, and K8 is even a (small) step up from there. I don't believe Dothan could keep up in clockspeed, which means, it likely won't be able to keep up in performance. Consider K8 will most likely achieve 2.6 GHz on 130nm, and 3+ GHz on 90nm. Banias (even though it was a mobile part, I fully realize that) topped out at 1.7 GHz on 130nm, and Dothans roadmaps so far only goes to 2.1 GHz on 90nm. That is a BIG difference, and just dropping the strict thermal limitations (like vcore) IMO won't make this happen.

dothan is a wonderfull cpu, it does extremely well what is was designed to do, but that is: provide ample computing power at the lowest possible power consumption (especially in minimum and typical power consumption it is in a league of its own given the performance. TDP is less spectacular). However, it was not made to be a top performing part, and I sincerely doubt it will ever become one without a major overhaul, and that includes more than just a FSB bump (which won't change much, P3 wasn't too FSB hungry, and with 2MB cache I expect the difference to be ever smaller for Dothan). But that is just my guess..

= The views stated herein are my personal views, and not necessarily the views of my wife. =
August 11, 2004 9:29:33 PM

oh, one more thing: I don't think there will be any significant difference between A64 and DOthan going from single to dual core. Both should gain roughly the same on the same sort of apps, there is no architectural difference that would give one cpu an advantage in dual core over the other. AMD's ondie memory controller doesn't give an extra benefit in dual core AFAICS. One thing that might be different though, is that a A64 could be thermally limited going dual core whereas Dothan might not be. That could give an edge to Dothan, but wether it will be enough to outclock and outperform a dual core K8, I have my doubts. Much will depend on how cool or hot 90nm A64's turn out to be.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
August 11, 2004 9:53:42 PM

Your point about A64 may be thermally limited at dual core while dothan is not - surly (as you have said) in order for Dothan to compete with A64 (of course, this is all theoretical...who really knows?) they would need to drop the thermal envelope issue, as heat is less of an issue on desktop parts - which would mean to get more speed, you'd get moore heat.

As you say, we don't know how Dothan would scale, so let's hope it is able to be converted into a performance part, as its architecture seems much more efficient than the P4s, and could give amd a shock.

What I'd like to know is, if the P3 was so good (and I know it certainly was), and the p4 is relatively worse (though faster due to its supperior mhz) why did intel choose that design route. If dothan / banias is just a tweaked p3, they must have realised the p3 was awesome - so why abandon it for p4? I suppose it reached the end of its' development cycle, but surly they could have carried across its stronger features?

XP2000, 512 ddr 2700ram, GF4 MX440, XP Pro
August 11, 2004 10:15:05 PM

> they would need to drop the thermal envelope issue, as
>heat is less of an issue on desktop parts - which would
>mean to get more speed, you'd get moore heat

True, no doubt. But you are starting with a very low TDP (21W or something ?) and an even much lower typical/iddle power consumption (next to nothing compared to Prescotts ~50-60W at iddle). Even if you multiply by 2, increase vCore, up the FSB, bump the frequency.. you're still looking at relatively low numbers IMO. 800 MHz FSB would add something like 5W (going by memory, I looked this up for P4 a long time ago), you could do the math on Vcore if you make it equal to Prescott (power draw increase with the square of Vocre) and you could do the math on frequency (power draw scales linearly with clockspeed). anyway, you could double Dothans TDP for vcore/clockspeed, then double it again for dual core, add 5W for FSB and still end up below Prescott. that is assuming it would need (or be able to use) a vcore as high as Prescott, which I doubt.

>What I'd like to know is, if the P3 was so good (and I know
>it certainly was), and the p4 is relatively worse (though
>faster due to its supperior mhz) why did intel choose that
>design route

Old discussion. P4 wasn't that bad really, when you consider performance on a given process size. For instance, P3 topped out around 1 GHz on 180nm, while P4 managed 2+ GHz. Willamette 'sucked' but not so bad it couldn't beat a P3 at half the frequency. Likewise, 130nm P3 (tualatin) reached 1.4 (?) Ghz, if you like 1.7 Ghz in the form of Banias, while P4 (northwood) reached 3.4 GHz with some headroom still.

P4 sucked when you compare performance/watt though, which is why it has always sucked as a mobile part. Thing is, performance/watt was not an issue on the desktop until recently, so P4 made a lot of sense in its time, but as it is now, it seems to be overshooting its design goals. Performance/watt is becoming determinative for performance of the P4 tout court. Not good for netburst.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
August 11, 2004 11:14:54 PM

Do there best and change K8 to K9 as fast as possible.I i wish them good luck with K9 1 arch from AMD since 10 year<P ID="edit"><FONT SIZE=-1><EM>Edited by juin on 08/11/04 11:36 PM.</EM></FONT></P>
August 12, 2004 2:46:05 AM

Quote:
the K8 arc is very good. but what will they do when they can't get clock speeds any higher. because the core looks like it can't be tuned anymore. so all they can do is ramp clock speeds up. So intel might have been the clever ones here. gaining clock speed first then tuning. whats everyone think?

I agree, AMD has been at 2.4ghz wall for sometime now from Athlon XP 3200+ to A64 FX-53, A64 3700+ s754, and A64 3800+ s939. It seems that can't go past 2.4ghz under current 130nm process and cpu architecture. Pretty much , if you can't make it faster. You make it wider! 64-bit wider. so in order for A64 to perform on par with Intel chips, they need 64-bit OS, 64-bit drivers, and 64-bit software support to be as fast as Intel 3.6ghz 32-bit chips. alot of compromises.

also, onboard memory controller also helped AMD increase IPC and performance to lift memory bottleneck limitations. AMD has tweaked everything possible to achieve the most out of every mhz. AMD introduced S939 and dual channel non registered memory. NEXT is for AMD is dual core, yep you guessed it. Dual core means no AMD chips will go past 2.4ghz. They will attempt to increase performance not by mhz but by forcing cpu to send larger packets. more effecient. What AMD needs now is 90nm , and higher clockspeeds.

------
Prescott 3.2E 1MB L2 HT
1GB PC 3200 Dual channel(PAT)
Asus P4P800 Bios 1016
PNY Geforce 6800 GT 256MB DDR3
60,823 Aquamarks
August 12, 2004 3:34:43 AM

You know that 64 bit make thing slower

i need to change useur name.
August 12, 2004 4:04:08 AM

Quote:
You know that 64 bit make thing slower
Quote:
and since Itanic is "true 64 bit", it is truly slow.
Okay, so we know that is bs. Always repy to bs with bs.
Kannyguy- Amd hit a wall with the xp chips at 2.2. The A64 scales past that, but Amd couldn't cope with the demand, if they put out an A64 @ 2.6 They may even be able to bring out a chip on 130 nano that runs at 2.8. With 90 nano, they will be top dog for all of next year, and intel's dual chip just wont help. (until progs catch up at least)
August 12, 2004 4:46:07 AM

my athlonXP will do 2.4ghz below stock voltages, and so can most peoples who have mobiles, so they havent exactly hit a wall. if i go slightly above stock voltage, i can do 2.5-2.6ghz

-------
<A HREF="http://www.albinoblacksheep.com/flash/you.html" target="_new">please dont click here! </A>

Brand name whores are stupid!
August 12, 2004 7:12:40 AM

well i dont think that its becuase amd cant release 2.6 or maybe even 2.8ghz parts, i think its more like they have no reason to. its not like 200 mhz will allow the chip to beat out the p4 part in areas where the p4 tends to lead anyway. all it would do is gain a larger lead where it already leads and maybe get closer elsewhere. Amd may just nto think the pressure is there that they have to rush a faster speed part and can take thier time getting to 90nm.
August 12, 2004 7:16:21 AM

>o in order for A64 to perform on par with Intel chips, they
>need 64-bit OS, 64-bit drivers, and 64-bit software support
>to be as fast as Intel 3.6ghz 32-bit chips.

Ahem.. lets rephrase that: In order for A64 to perform as slow as inexistant intel chips, it requires an unfinished 64 beta OS and mediocre 64 bit drivers with major performance penalties. For everything else, I'd say intel has to catch up to AMD's performance lead. But I know you know that, but to choose to ignore reality in your own fantasy world.

> Dual core means no AMD chips will go past 2.4ghz.

Maybe not for the dual core AFX chips, but do you want a bet we will see >2.4 GHz this year ? and >>>2.4 GHz next year? You want to bet AMD will increase their clockspeeds significanlty more than intel over the next 12 months (percentage wise) ? I say >25% for AMD and <12% for Intel. I take paypal.

> They will attempt to increase performance not by mhz but
>by forcing cpu to send larger packets. more effecient.

Larger "packets" ? LOL. Anyway, I don't care where the performance comes from, if someone builds a 300 Mhz cpu the outperforms all the others, thats the one I want.

>What AMD needs now is 90nm , and higher clockspeeds.

They need 90nm to be able to more cheaply produce K8. They don't need clockspeed or performance boosts nearly as bad as intel does.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
August 12, 2004 9:03:34 AM

Agreeed - I mean when did intel reach 3.2ghz? It was a very long time ago! Now they're at 3.4, while athlons are at 2.4, when they were at 2.2 with the xp - so I think that argument is negated really.

One thing I dislike about amd right now (and this is true of the XP chips too) is that they have A64s 2800, 3000, 3200, 3500, 3700 - yet thats between 2ghz and 2.4 ghz! Its isnt quite as bad as the xp chips in all honesty - between 2400, 2500, 2600, 2700, 2800, 3000 3200, there was what, 2000mhz difference? I know the ins and outs of it (different cores, yadayadayada) but that is the basic fact. Do these minor improvements warrant whoole new retail parts, imo it doesnt.

I hope the 64 doesnt get to that stage really, though AMD are doing a damn good job, so I suppose the ball's in intel's court!

XP2000, 512 ddr 2700ram, GF4 MX440, XP Pro
August 12, 2004 12:26:55 PM

You know Itanium int performance is not better that P4 NW 2.8 GHZ.While having twice or more resource at every level.It is a probleme only having 64 bit mode.Opteron also reach best INT performance on 32 bit.While is dont affect FPU as much as data size is still a max of 80 bit for IA64 and AMD64.

i need to change useur name.
August 12, 2004 12:32:52 PM

They need 90nm to be able to more cheaply produce K8. They don't need clockspeed or performance boosts nearly as bad as intel does.


Performance dont make everything.

i need to change useur name.
August 12, 2004 12:44:17 PM

>Performance dont make everything.

You're quite right. other things matter too, things like SFF, low noise, low power consumption, or the ability to run future 64 bit software, or the ability to use SP2 NX secutity feature.. those things matter as well ! I guess intel can relax now ?

= The views stated herein are my personal views, and not necessarily the views of my wife. =
August 12, 2004 1:16:22 PM

Or having a real chipset chipset division or compiler that goes with your chip.

i need to change useur name.
August 12, 2004 2:14:15 PM

Yes, that too, and having a blue corporate colour scheme, and selling bunny man as well !

= The views stated herein are my personal views, and not necessarily the views of my wife. =
August 12, 2004 3:55:18 PM

Or a CEO that think the world end after the blue sheme corp that you were taking about.

i need to change useur name.
!