Sign in with
Sign up | Sign in
Your question

Thouroughbred to Barton, then what for x86?

Last response: in CPUs
Share
March 16, 2002 1:56:18 PM

As most of you already have seen or guessed, AMD is pushing for the T-bred this month or early on in april. After that, AMD is shooting for Barton with SOI process in Q3 this year. If AMD is entering the 64-bit world, and is going to mainstream the new technology for both servers AND practical home PCs, what is the future of the x86? Will Barton have any reasonable upgrades, or is it the end of the line for AMDs SocketA and K7 line of processors?

"When there's a will, there's a way."
March 16, 2002 1:59:02 PM

The general consensus from outside AMD seems to be that aside from die shrinks (.09 micron) Socket A will receive no major upgrades from here on out. Instead, AMD will likely move it down into the price category currently held by Duron; that being the budget processors., I think theis will keep it alive and kicking well into 2003 and possibly into 2004 as well.

Mark-

When all else fails, throw your computer out the window!!!
March 16, 2002 2:02:25 PM

Don't forget that the Hammer is a 64-bit extension of the x86 architecture.

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
Related resources
March 16, 2002 2:05:14 PM

Are you saying that the Hammer is sharing roots of a 32-bit processor? I've read into it's ability to work in either 32-bit, 64-bit, or compatability mode. Does this mean that Hammer can be considered a transition point between 32 and 64 bit? It would be nice for AMD to make the big leap in the programming world for all of us. Considering these facts, would you consider Hammer the processor of the near future AMD_Man?

"When there's a will, there's a way."
March 16, 2002 2:12:22 PM

I for one, would. The Hammer is made to make the transition fully functional, use 32-bit at extreme performance and be able to get into 64bit later on. That's the goal of Hammer, not be a 64-bit processor only. It is AMD's next BIG HIT like Athlon, majoy leap in performance.

--
For the first time, Hookers are hooked on Phonics!!
March 16, 2002 2:20:51 PM

Yes, I agree with Eden. With the Hammer you won't have to throw away your old software, but in the future, you'll be able to rejuvenate your system's performance through a simple RAM upgrade and a 64-bit OS and 64-bit apps.


Why a RAM upgrade? I'm predicting 64-bit apps will use up vast amounts of RAM that the current standard of 256MB or 512MB RAM can't handle.


AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
March 16, 2002 4:27:27 PM

You are correct there. Most applications will be twice the size, as all pointers and references will be take up 8 bytes of memory space (64-bits) rather than 4. A 64-bit processor can only fit half as much into its caches as a 32-bit processor, unless it doubles all of its cache sizes. A doubling of cache sizes has other inherent problems, such as taking more time to search through the cache.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 16, 2002 5:33:35 PM

So then, Ray, what would you say the standard amount of RAM will become once 64-bit starts to take hold? 1024MB? Or will it be an even higher amount? That right there might make it cost prohibitive for many until large capacity RAM becomes norm (with a 1024MB stick of PC2100 costing $385 and a 512MB stick of PC800(not 2x256, but 1x512) being $213 as of 3/16/02 on pricewatch.com)

And, will the dual channel nature of RDRAM allow for those massive amounts of RAM? The reason I ask that is that I'm only aware of RDRAM up to 512MB, which means in a four slot MB, you'll only be able to get 2048MB (way more than enough by today's standards, but enough for high end 64-bit processing?) Does the RDRAM standard have a solution for this possible problem? Not that DDR might do any better, since most DDR boards are limited to 3 slots, since any more can cause timing issues... but still with 1024MB sticks out, that will make 3072MB. What are your thoughts, Ray?

-SammyBoy
March 16, 2002 5:39:51 PM

I believe RAM prices will go down again by then and more dense RAM sticks will become available.

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
March 16, 2002 5:50:06 PM

Quote:
I believe RAM prices will go down again by then and more dense RAM sticks will become available.


Unfortunatly, ram prices will never again fall to the levels they are at now and were at, for the past 6 months ram companies have been selling at a loss due to the dip in the industry, things are returning to normal and ram prices will probably double before its over.


Buy the ram NOW, before its too late!

"The Cash Left In My Pocket,The BEST Benchmark"
No Overclock+stock hsf=GOOD!
March 16, 2002 5:59:39 PM

Matisaro, that's now. When there's a need for more RAM, RAM prices will go down again. RAM is a commodity so its price adjusts according to the purchase patterns of customers.

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
March 16, 2002 6:15:02 PM

Quote:
So then, Ray, what would you say the standard amount of RAM will become once 64-bit starts to take hold?

This is difficult to say. The moment you begin running 64-bit applications, the memory requirement for your system immediately doubles. Thus, in order to get performance even equal to a 32-bit system, you will need to double the system memory.

Note that when we moved from 16-bit processors to 32-bit processors we jumped almost instantaneously from 1MB of memory to about 4-8MB of memory as being standard. The cost of these 32-bit computer systems (using the 80386 processor) was quite prohibitive for a while, especially due to the high pricetag associated with so much memory.


Quote:
And, will the dual channel nature of RDRAM allow for those massive amounts of RAM?

Current Intel-based RDRAM chipsets are not designed for a 64-bit processor, so we need not worry about this. 64-bit processing will not be attractive in the consumer market for quite a long time.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 16, 2002 6:23:57 PM

Quote:
Matisaro, that's now. When there's a need for more RAM, RAM prices will go down again. RAM is a commodity so its price adjusts according to the purchase patterns of customers.


Yes, but the level of the commodity is usually much higher than it is now, the cost of production usually ensures a high pricetag regardless of demand, recently the dramurai have been selling at a loss, thus we have very low prices.


The prices are stabilizing once more, but unless ram companies are going out of buisness, they will not drop this low again.

"The Cash Left In My Pocket,The BEST Benchmark"
No Overclock+stock hsf=GOOD!
March 17, 2002 3:16:16 PM

Out of curiousity, what does the Itanium run off of then, memory wise?

-SammyBoy
March 17, 2002 4:45:57 PM

it uses pc100 sdram memory I believe

<i>My life wasn't complete untill I tried sse-2 optimized pong</i>
March 17, 2002 4:53:10 PM

and PC1700 (100Mhz DDR)...

This post is best viewed with common sense enabled<P ID="edit"><FONT SIZE=-1><EM>Edited by iib on 03/17/02 08:54 PM.</EM></FONT></P>
March 17, 2002 5:24:43 PM

No wonder it turned out this much crappy to most people...


--
For the first time, Hookers are hooked on Phonics!!
March 17, 2002 6:08:08 PM

By June what will be Amd's newest Processor.... Will the Thoroughbred be the latest in June, or will something Be better by June?

Once you go AMD, You never Go back!!
March 17, 2002 6:14:51 PM

Rob since Thoroughbred won't be available for sale until April or May even, it's safe to say Thoroughbred will still be the latest AMD processor available through June. In fact, I don't expect to see Barton until at least September or October. Hammer isn't expected until the very end of the year (November or december) and we may not see them actually on sale until January.

Look at AMD's roadmap. It's pretty clear and doesn't really change all that much except to extend the roadmap to new quarters over time.

Mark-

When all else fails, throw your computer out the window!!!
March 17, 2002 6:45:08 PM

This list shows the maximum amount of RAM each type of processor supports. It all has to do with the largest number the processor supports, which means it's the highest adress it can create for a memory position. Yes, I know that definition sucked, but you get the idea.

16 bit processor - 64KB
32 bit processor - 4GB
64 bit processor - 16,777,216TB (16,384PB)

Now did I figure out the 64bit processor right? My method worked for the 16 and 32 bit processors so I'm assuming it works with the 64 bit processor as well.

2^x(bit) then divide by 1024 until you get useable figure.

if(GetSystemMetrics(SM_PROCESSOR) != AMD_PROCESSOR)
{
SendMessage(hwnd,WM_CLOSE,0,0);
}
March 17, 2002 6:57:54 PM

true... Itanium pretty much sucks... its floating point preformance are matched by the latest x86 processors (P4 and Athlon) and its integer preformance are far infrior to x86...
it doesn't have a real scalibe multi-processor abiltys (not even glueless SMP - forget about HAMMERs HT and Bus-to-Bus protcols) and has very weak I/O... (PC100 or PC1700).
it is considered a pretty weak cpu amoung the RiSC top guns (DEC alpha, IBMs Power, Suns Ultra-Sparc).

did I mention it cost 5 times a x86 top-of-the-line processor eats twice the power and outputs twice the heat with virtualy no real software support?

go go intel...

-----
yes my stocks are on AMD...


This post is best viewed with common sense enabled<P ID="edit"><FONT SIZE=-1><EM>Edited by iib on 03/17/02 11:02 PM.</EM></FONT></P>
March 17, 2002 7:07:40 PM

Ok so i guess i'll be getting the ThoroughBred.
Will a 2.0ghz or 2.2 ghz be available?? Any Comments would be helpful...one other thing..Kt400a will be out by June?

Once you go AMD, You never Go back!!
March 17, 2002 7:16:16 PM

Quote:
16 bit processor - 64KB
32 bit processor - 4GB
64 bit processor - 16,777,216TB (16,384PB)

We used 1MB of memory in our 16-bit systems. The segment registers allowed us to access more than a single 64KB segment. The 20-bit memory bus allowed us full access to that 1MB of memory.

Our modern 32-bit processors come with a 32-bit memory bus. This gives us access to 4GB of memory, as you stated.

64-bit processors will allow access to up to 17,179,869,184 GB of memory, which is probably what you stated, though I do not feel like translating to terabytes.

Increasing from a 32-bit to a 64-bit processor (without a redesign of the instruction set) is similar to switching gears from 2nd gear to 3rd gear. The moment you do that you lose some power. But you have greater potential. Until we reach the limit of the potential of 32-bit processors, I do not think the loss in power is worth it. In essence, switching from 2nd gear to 3rd gear too soon will kill your performance.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 17, 2002 7:26:42 PM

Hammer puts you on automatic Gear...
you can switch to 3rd when ever you fell like it (if MS will produce a 64bit Hammer OS) and you can still go 2nd Gear even with a 64bit OS! depending on the application you run...

or go 2nd gear all the way if you dont fell like your ready for 64bit - just use a regular windows OS or Linux OS...



This post is best viewed with common sense enabled
March 17, 2002 7:36:00 PM

Yes, but Hammer's 3rd gear has nowhere near the performance of the Itanium's 3rd gear. ;)  I will grant you though that Hammer's 2nd gear is a hell of a lot better than Itanium's 2nd gear. This is why Itanium is not targetted at the 32-bit market.

(For those having trouble following, 2nd gear is 32-bit performance and 3rd gear is 64-bit performance.)

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 17, 2002 7:48:26 PM

The Itanium costs thousands of dollars, while a mid-range ClawHammer will probably cost $200-$300USD. We, as enthusiasts aren't concerned about the performance of the Sledgehammer or the Itanium.

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
March 17, 2002 8:00:37 PM

Well this forum is for everyone, not just the PC enthusiast's market. Unless the forum is renamed to "x86 CPUs" I believe there is a place for discussion of these processors here. ;) 

As far as the cost... I will probably pick up an Itanium just to play with it once it reaches the $1000 range.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 17, 2002 8:02:37 PM

Quote:

Yes, but Hammer's 3rd gear has nowhere near the performance of the Itanium's 3rd gear. ;) 


dont be so sure about it... the Itanium as it stands today offers weaker Integer preformance and aroud equal Floating prefornce comperred to your evry day x86 processor. fell free to look up Itaniums Spec_Int and Spec_FP scores.
see how it ranks among RiSC processors (say IBMs Power4) and x86 processors - its not that Hot (well - it's hot - as-in you can fry eggs on it faster then on an athlon :)  )

also look at "Real World Technolgies" take on the Hammer
<A HREF="http://www.realworldtech.com/page.cfm?AID=RWT0102020339..." target="_new">Here</A>
one of the finest sites which deals with High-End processor preformance evaluation.

heres some of what it has to say:

Quote:

Although the Intel McKinley and AMD Hammer are both 64 bit MPUs, these devices are directed at different markets. While the large and expensive McKinley will target medium and high-end server applications, the first member of the Hammer family, code named "Clawhammer", will target the high end desktop PC market. That is not to say that McKinley will outperform the Clawhammer device. Indeed, I expect the AMD device will easily beat the much slower clocked IA64 server chip in SPECint2K and many other integer benchmarks, as well as challenge much faster clocked Pentium 4 devices in both integer and floating point performance.


This post is best viewed with common sense enabled
March 17, 2002 8:07:48 PM

Question on that Ray.......double of the cache size has its own problems. I'm guessing it has to do with timing and electric migration and such. What about the sledgehammer that's going to have up to 8 megs of l2 cache. Do you see AMD having problems with that much cache and performance; does that much cache become a hinder in performance in anyway?

Curious, I'm doing research just for knowledge on this.

MeldarthX
March 17, 2002 8:18:18 PM

Quote:
Question on that Ray.......double of the cache size has its own problems. I'm guessing it has to do with timing and electric migration and such. What about the sledgehammer that's going to have up to 8 megs of l2 cache. Do you see AMD having problems with that much cache and performance; does that much cache become a hinder in performance in anyway?

Yes, the larger yoru cache the more time it takes to search through it. If something is not in the cache then you incur the penalty of having to search through it all before even trying to fetch it from memory. Anything that is not found in its 8MB L2 cache is going to have latency issues before a memory fetch is even attempted.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 17, 2002 8:40:09 PM

Currently, cache runs at the CPU frequency... but is it possible to double, quad, etc., pump the cache so that it runs faster than the CPU? Granted, that seems like an easy way to overcome the increased cache size penalty, but is it feasible?

-SammyBoy
March 17, 2002 8:48:18 PM

Quote:
dont be so sure about it... the Itanium as it stands today offers weaker Integer preformance and aroud equal Floating prefornce comperred to your evry day x86 processor. fell free to look up Itaniums Spec_Int and Spec_FP scores.
see how it ranks among RiSC processors (say IBMs Power4) and x86 processors - its not that Hot...


At 800MHz, the currently released Itanium comes in at a SPECint2000 of 370 and a SPECfp2000 of 711. An Athlon XP 2000+, operating at over twice the clockspeed as the Itanium, comes in at a SPECint2000 of 710 and SPECfp2000 of 619. When the McKinley is released as the next version of the Itanium, it will offer about twice the performance.

Now then, how many of you want a processor that offers twice the performance as one with a score of 711 on the floating point spec. Think about the performance in all of your 3d games and applications. ;)  I am going to grab one when the prices hits 3 digits just to play around and see what I can do. It will be fun.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 17, 2002 8:53:37 PM

Quote:
Currently, cache runs at the CPU frequency... but is it possible to double, quad, etc., pump the cache so that it runs faster than the CPU?

If this was done, then the cache would be the limiting factor in the processor's clockspeed. You would have to reduce the clockspeed of the processor for this to work, so it is not really feasible. The same transistors are used in the cache as are used in the rest of the processor. They all have about the same speed limitations.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 17, 2002 9:11:31 PM

"this forum is for all cpu discussion"

you mean, even apple cpu's???? *gasp* lol

and i see intel and amd diverging. intel going the way of the p4 with amazing high clockspeeds while amd goes for the hammer architecture and ondie dual processors

I love helping people in Toms Forums... It reinforces my intellectual superiority! :smile:
March 17, 2002 9:16:35 PM

Quote:
At 800MHz, the currently released Itanium comes in at a SPECint2000 of 370 and a SPECfp2000 of 711. An Athlon XP 2000+, operating at over twice the clockspeed as the Itanium, comes in at a SPECint2000 of 710 and SPECfp2000 of 619. When the McKinley is released as the next version of the Itanium, it will offer about twice the performance.


OK Clawhammer on release is expected to have a SPEC2000 of over 1400 at a clockspeed of just over 2ghz and a price similar to todays Athlons. Not sure of the FP numbers, though.

Working the numbers out...let's say Itanium gets down to $1000 price range by the time Hammer intros. How much will Itanium boards go for compared to Hammer boards?

With a 2P Hammer system probably going for significantly less money than an Itanium 1 or 2 P AND Offering better performance all around than an Itanium costing more....which would I get?

Hammer probably.

Any timeframe on when the Itanium will hit the 1000 mark?

Mark-

When all else fails, throw your computer out the window!!!
March 17, 2002 9:21:21 PM

Quote:
OK Clawhammer on release is expected to have a SPEC2000 of over 1400

*laugh* Says whom? You seriously expect a Clawhammer at PR3400 to perform 2 times better than the Athlon at PR2000? The PR rating improvement is only 1.7 times. I mean no disrespect, but this is just not going to happen. The author of whatever article you are quoting is full of himself.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 17, 2002 9:26:46 PM

For all we know that PR refers to the relative performance of the clock speed of the Athlon XP not the Athlon. If that's true then a Hammer is as fast as an Athlon XP at 3.4GHz which is over twice as fast as an Athlon XP 2000+. Plus, AMD's PR ratings are generally conservative.

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
March 17, 2002 9:28:47 PM

I highly doubt AMD is going to change the definition of their PR ratings. The shifting of definitions would absolutely kill their credibility.

As far as their PR ratings being conservative, I disagree. AMD handpicks the benchmarks that allow their processors to perform the best, enabling them to pick higher PR ratings. They publicly state their PR ratings compare their processors to their own older versions, yet put up benchmark charts showing a comparison to a Pentium 4. The only reason they publicly state it as a comparison to their own processors is to avoid legal issues. We all know the intended purpose is to base comparisons against Intel processors.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 17, 2002 9:38:47 PM

Hmm, I don't understand. Most review sites agree that the Athlon XP 2100+ is best performing overall processor in most tasks. I consider that conservative.

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
March 17, 2002 9:41:38 PM

And what tasks are these? Do you need MS-Word to run more quickly?

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 17, 2002 9:54:48 PM

See <A HREF="http://www6.tomshardware.com/cpu/02q1/020313/index.html" target="_new">here</A> and <A HREF="http://active-hardware.com/english/reviews/processor/xp..." target="_new">here</A> and <A HREF="http://www.anandtech.com/cpu/showdoc.html?i=1595" target="_new">here</A>.

As you can clearly see, the Athlon XP is faster overall in all aspects of performance except Lightwave, which is specifically P4-optimized.

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
March 17, 2002 10:20:35 PM

Quote:
As you can clearly see, the Athlon XP is faster overall in all aspects of performance except Lightwave, which is specifically P4-optimized.



First off, your middle link pairs up the Pentium 4 with DDR SDRAM. This kills performance. That one can be tossed.

That leaves a link to a THG review, and another review from some other site. I never count THG reviews as they are always flawed/biased/take your pick. So I am not going to bother reading through it.

Let us take a look at the numbers from your third link. Ok this is an Anandtech link. Let us count up the benchmarks to see where we won, lost, and tied...

Here are the results indicating who won each benchmark:

Pentium 4 2.2GHz: 6
Athlon XP 2100: 4
Tie: 1


Would you care to retract your comment? Perhaps you can point out some further benchmarks from non THG sites where they use an i850 chipset motherboard for the Pentium 4?

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 17, 2002 10:34:12 PM

I would, but since I've seen you and Matisaro argue for dozens of posts, I'll just hand this one to you simply to avoid the risk of losing your respect or the respect of other valued members of this community.

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
March 17, 2002 11:05:16 PM

So long as you refrain from "name-calling", cussing, lying, etc. you will not lose my respect.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 17, 2002 11:17:37 PM

I don't consider Lightwave to be Pentium 4 optimized, otherwise a Celeron 1.3GHZ should not be beating an Athlon XP 1500+. The software is clearly rejecting AMD processors, as benchmarks indicate, and like I said, Celerons are nowhere near Athlon performance, considering they don't even contain any SSE optimizations like AthlonXPs. Thus that test should be omitted and even sent back to Newtek for complaint.

--
For the first time, Hookers are hooked on Phonics!!
March 17, 2002 11:20:47 PM

Fine, correction, it's Intel-optimized or AMD-deoptimized, take your pick.

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
March 17, 2002 11:27:15 PM

While I agree that Celerons based on the Pentium III core do not have SSE2... what makes you think they do not have SSE? One interesting problem in today's software is that many applications that do incorporate optimized instructions for different processors will have trouble with processors that have multiple types of optimizations. They will check for 3dnow support and load that module if supported. If not supported, they will then check for SSE support and load that module if supported. This is simply the way it had always been done.

If today's optimized software is using 3dnow (not 3dnow pro) for any processor that supports it, then a benchmark that does the same thing should be used to accurately reflect real software. The 3dnow modules most software used were not built with SSE enhancements as well. Thus, you have to pick one or the other. Older and most current software being sold were not created with a processor in mind that supported both 3dnow and SSE. If AMD was planning on adding SSE support they really should have informed the software industry long ago so we could have taken this into account. The use of processor optimizations is only as good as the information provided by the processor manufacturer at the time the software is being written.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 17, 2002 11:31:54 PM

3DNow seems to be a subset of SSE and so I believe SSE should be preferred over 3DNow in modern apps. Also, I've heard about a few apps that check for the 'GenuineIntel' text before using SSE.

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
March 17, 2002 11:35:37 PM

I agree that SSE should be preferred over 3dnow in future software. However, AMD's processors will have to run software that is available today and written in the past. This is much the same problem that the Pentium 4 encountered when it was first introduced. Most applications used the wrong optimizations for this new processor, and most review sites held this against the processor. It is now the Athlon XP's turn to deal with it.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 17, 2002 11:42:33 PM

Actually, I don't think 3DNow and SSE are really related. They are different routines that do some of the same things. 3DNow PRO? actually incorporates the full SSE(1) instruction set, making SSE a subset of the larger 3dNow instruction set.

SSE2 incorporates SSE in it's more extensive instruction set

OK OK instruction sets are confusing me now....I think I need to take more ginseng!

Mark-

When all else fails, throw your computer out the window!!!
!