• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

*AMD To Release 333MHz FSB Athlon XP Processors*

A

AGodspeed

Diamond Member
http://www.xbitlabs.com/news/story.html?id=1016666876

AMD representatives confirmed that Thoroughbred will not be different from Palomino in architecture, will work at 266 MHz bus frequency and will have a 256-Kb L2-cache. That is why new 0.13-micron Athlon XP will be fully compatible with the old mainboards. Nonetheless, AMD is planning to switch its CPUs to a faster 333 MHz FSB, however, it can happen most probably, only after the next Barton core release.
 
There is some debate as to whether AMD will actually release Barton.

Personally, I have my doubts as to whether AMD would release an entirely new Athlon platform at the end of that processor's lifecycle on the desktop...
 
you will only most likely to experience performance boost with high operating clock frequency with 333mhz FSB.
This was proven by Anand in his article on Athlon 2100XP.
 


<< There is some debate as to whether AMD will actually release Barton.

Personally, I have my doubts as to whether AMD would release an entirely new Athlon platform at the end of that processor's lifecycle on the desktop... >>

Lol, how is changing the multiplier going to be "an entirely new Athlon platform" KenAF? There's no architectural changes involved AFAIK.
 


<<

<< There is some debate as to whether AMD will actually release Barton >>



Link? >>

It's true actually. To give one of many examples....here. However, it's irrelevant at this point, since Thoroughbred will likely be dubbed the "Duron" once ClawHammer hits the market...
 
I am disappointed that AMD has not added more cache (from 256k ito 384k or 512k on L2) in Thoroughbred to continue a leading performance against Northwood. I guess Northwood continues holding its lead over AMD in performance until Hammer comes out in late 4 QTR this year...
 


<< I am disappointed that AMD has not added more cache (from 256k ito 384k or 512k on L2) in Thoroughbred to continue a leading performance against Northwood. I guess Northwood continues holding its lead over AMD in performance until Hammer comes out in late 4 QTR this year... >>

AFAIK the Athlon XP 2100+ is just as fast (if not slightly faster) than the 2.2GHz Northwood Pentium 4...
 
> Lol, how is changing the multiplier going to be "an entirely new Athlon platform" KenAF?
> There's no architectural changes involved AFAIK.

Well, I just don't see it happening. Everyone will require new motherboards for Clawhammer....and I just can't see AMD requiring new motherboards for new Clawhammer and Athlon processors at roughly same time. Many will complain that AMD is forcing them to upgrade their PCs just to use newer processors...and that's rather hard to justify at the 2% to 5% gain in real world applications found by Anand's Hardware in its tests.

Certainly, AMD will hit 166FSB with the Athlon or Athlon mobile, but I don't think it will occur until PC2700 DDR has completely replaced PC2100 in the marketplace (at equal/lower cost). PC2700 DDR is available now, but not at anywhere near the volumes of PC2100 DDR, and I have my doubts as to whether PC2700 DDR will be more common and at comparable cost for an equal level of demand. Current PC2700 prices are not valid as a benchmark, given demand for the memory is so low. Current PC2700 DDR is also CAS2.5 by SPD timings, compared to CAS2 by SPD for the better PC2100 DDR; when you account for the increased latency, it's almost a wash as far as performance goes (and you can't change the timings on memory on the motherboards used by a number of the large retail vendors with AMD systems).

I would also add that there is no evidence to support the conclusion that a greater FSB will be needed for faster Athlon processors. AMD has said a single 2.1Gb/s PC2100 memory channel with be more than sufficient for Clawhammer with performance rating of 4400, so I find it hard to believe that the Athlon, at any reasonably forthcoming frequency, would perform significantly better with more than the 2.1Gb/s it has now with 133FSB. If 2.1Gb/s is more than enough for Clawhammer, why should Athlon need more? The simple fact is that current AMD processors are not [memory] bandwidth limited.
 


<< > Lol, how is changing the multiplier going to be "an entirely new Athlon platform" KenAF?
> There's no architectural changes involved AFAIK.

Well, I just don't see it happening. Everyone will require new motherboards for Clawhammer....and I just can't see AMD requiring new motherboards for new Clawhammer and Athlon processors at roughly same time. Many will complain that AMD is forcing them to upgrade their PCs just to use newer processors...and that's rather hard to justify at the 2% to 5% gain in real world applications found by Anand's Hardware in its tests.

Certainly, AMD will hit 166FSB with the Athlon or Athlon mobile, but I don't think it will occur until PC2700 DDR has completely replaced PC2100 in the marketplace (at equal/lower cost). PC2700 DDR is available now, but not at anywhere near the volumes of PC2100 DDR, and I have my doubts as to whether PC2700 DDR will be more common and at comparable cost for an equal level of demand. Current PC2700 prices are not valid, as demand for the memory is so low. Current PC2700 DDR is also CAS2.5 by SPD timings, compared to CAS2 by SPD for the better PC2100 DDR; when you account for the increased latency, it's almost a wash as far as performance goes (and you can't change the timings on memory on the motherboards used by a number of the large retail vendors with AMD systems).

I would also add that there is no evidence to support the conclusion that a greater FSB will be needed for faster Athlon processors. AMD has said a single 2.1Gb/s PC2100 memory channel with be more than sufficient for Clawhammer with performance rating of 4400, so I find it hard to believe that the Athlon, at any reasonably forthcoming frequency, would perform significantly better with more than the 2.1Gb/s it has now with 133FSB. If 2.1Gb/s is more than enough for Clawhammer, why should Athlon need more? The simple fact is that current AMD processors are not [memory] bandwidth limited. >>

Several memory makers (Micron, Hynix, Samsung, etc.) have mentioned that by the end of 2002/beginning of 2003, PC2700 memory should be out en masse. Samsung has even demonstrated fully functional PC3200 RAM, and that's NOT DDR-II. I don't think latency will be a big deal with PC2700 modules.

No one is forcing anyone to upgrade to 166MHz FSB Athlon XP's. Maybe a few guys on these forums will bitch and moan about it, but that hardly concerns AMD. Also, if you go here, you'll find that Johan believes that a 166MHz FSB will yield significantly better results than what Anand got. Performance gains are in the double digits according to Johan's data.

And when did AMD say that single channel PC2100 RAM would be more than sufficient for ClawHammer? I find that odd?
 
anand's results seemed to be complete utter crap for the 166Mhz bus. Just take a look anywhere, the XP333 from Iwill proves it many times, this wouldn't be the first time anand numbers are wrong, hell, its not even surprising anymore.
 
I am disappointed that AMD has not added more cache (from 256k ito 384k or 512k on L2) in Thoroughbred to continue a leading performance against Northwood. I guess Northwood continues holding its lead over AMD in performance until Hammer comes out in late 4 QTR this year...


There is another spin to AMD not introducing Tbreds with a 512 cache. Price. The current 129mm Athlon competes well with available Northwoods and runs about 300 die per wafer. The 80mm Tbreds should continue that trend and run close to 414 die per wafer (unless my math is off). Ignoring yield and margin factors, this translates to a potential 38% reduction in costs that could be carried over to the buyer. Intel would have to counter that with a price drop of their own, or lose market share. Either way, the consumer will see the lower prices that follow with every process upgrade.

I also don't think DDR333 will be in high enough volume to supply the market for Thoroughbred and Northwood processors until at least end of Q2 of 02'. If AMD introduced the Tbred with a 333Mhz FSB now, only those willing to spend a premium for DDR333 would be able to pair it up with the new processor. It seems to me that introducing DDR333 with Barton makes more sense. It would be the platform evolution that would settle the Athlon core as the "Duron" of the next few years, leaving the Hammer to reign at the high end with a 600Mhz bus and single/dual channel DDR333.
 
AA,

> anand's results seemed to be complete utter crap for the 166Mhz bus.
> Just take a look anywhere, the XP333 from Iwill proves it many times,
> this wouldn't be the first time anand numbers are wrong, hell, its not
> even surprising anymore.

It makes a big difference whether you are testing at CAS2 with all sorts of tweaked bios memory options (i.e. on KR7A), or the standard CPD memory timings (CAS2.5 or CAS3) that you would get by default with the bios, and as standard on most retail (non-enthusiast) systems. In comparing results, you also need to be sure you are comparing processors at the same frequency.

Anand tested with SPD memory timings (CAS2.5 @ 166FSB), IIRC Anand kept the processor frequencies equal, whereas Aceshardware (and other reviews) did not for many of their tests.

Aceshardware tested using fully tweaked nForce and KT266A (KR7A) setups with CAS2. If you read the results carefully, Aceshardware found 1733MHz with 165FSB @ CAS2.5 to offer only 7% improvement over a 1666MHz with 133FSB @ CAS2, and that was in Serious Sam which is extremely sensitive to memory bandwidth. Other applications show a negligible improvement of 165-166FSB with CAS2.5 over 133FSB/memory with CAS2. When accounting for the difference in clock speed, Aceshardware tests show that 165FSB @ CAS2.5 offers a mere 3% benefit over 133FSB @ CAS2. Moving from CAS2.5 to CAS2 at 165FSB increases performance an additional 6%, but you won't see any of that from a retail Athlon system from the likes of Gateway, which use only CPD timings for memory. Only the enthusiasts with custom built rigs (and quality PC2700 memory capable of running @ CAS2), which form less than 10% of AMD's customer base, will see that gain.

Once again, you need to be careful in comparing the benchmarks at Aceshardware and some other sites, because the 133FSB systems are running at 1660MHz, while others are running at 1720MHz. Johan should have provided 133FSB results at 1730MHz for direct comparison (or used a lower multiplier on his high FSB tests), but he did not.
 
Godspeed,

> And when did AMD say that single channel PC2100 RAM would be more than
> sufficient for ClawHammer? I find that odd?

During one of their presentations...of course. "More than sufficient" wasn't the exact phrase, of course, but the jist of the message was that AMD didn't care whether PC2700 or PC2100 DDR was the norm at release, because Clawhammer would do [just about] equally well with both. Why do you find that odd?

As Paul Demone notes in his latest post/article, increasing the Clawhammer's memory bandwidth by 100% (as on Sledgehammer with its second memory channel), combined with four times the amount of L2 cache (1Mb vs 256Kb), only improves integer (application) performance by 5%. Paul DeMone is well recognized, and most people in the know seem to agree with him (and a new AMD interview seems to confirm his x86-32 Clawhammer performance projections). When twice the bandwidth (100% difference) and four times the L2 cache would only improve integer and general application performance by 5%, why do you think there would be much of an improvement from PC2100 to PC2700 (27% difference)?

As Mr. DeMone notes, the quadrupled L2 cache and doubled memory bandwidth would substantially improve floating point performance. That you would probably see in some games and traditional 3D rendering applications...but not in typical application performance (Office, Photoshop, web browsing, etc), nor in most games.
 
I think Barton is just get their plants fabing SOI because barton was supposed to have SOI last time i checked. They could very well implement 512k cache in barton since I think the Barton core would be more like a transitionary chip to hammer if it is released at all. Can't wait till hammer time though 🙂
 
It makes sense that Barton will be a 166Mhz FSB chip. By that time current DDR boards like the KT266A will be almost fully supplanted by the DDR333 chipsets from SiS, ALi and even VIA. PC2700 will be commonplace and competition from Intel will likely demand it. I don't see AMD implementing a 512k L2 until Sledgehammer, but I suppose anything is possible. With SOI the Thoroughbred should be able to shed some additional wattage...hopefully dropping down to 1.55 or below on the Vcore. 😎
 
this EOL on the athlon is a bit premature.... i think it will take a year, at least, for clawhammer to supplant the athlon core in units shipped by AMD. that is shorter than earlier, like when we had the 486 and 386 side by side for 3 years or so, but it won't be immediate upon hammer's release. thats just idiocy.
 
it's possible...but if and only if psuedo-pc2700 is widely available by multiple manufacturers, not just nanya for god's sake.

most manufacturers say they've announced pc2700 production ramping in proportion to market demand...which is nill at this point...so we get the chicken/egg dilemna.

no pc2700 until there's demand.....no pc2700 platform until pc2700 is abundant....
 
I think it's premature to assume that a 333mhz fsb won't make much of a difference. Let's look at 266mhz fsb as an example: The AMD 760 came out and was very fast. Then VIA released the KT266 which was fast, but slower than the AMD 760 and not much faster than the KT133A(PC133 SDram chipset). Then ALI came out with their 266 chipset which was even slower than the KT266. The SiS 735 came out and was significantly faster than the AMD 760. Then VIA's KT266A came out which was even faster than the SiS 735. Nvidia's NForce was released and also was much faster than the AMD 760, being about on par with the KT266A. What we see here is that as chipset manufacturers became more familiar with the 266mhz fsb and DDR SDram, they were able to eek out much more performance than what most were expecting. The same may hold true for 333mhz.

As for the ram situation, let's not forget all the production problems that RDram had.
 
Yeah I knew AMD wouldnt have jumped to 333fsb until proper chipset support was available! That will happen when the KT400 is released! It does look like Duron will move to 266fsb as indicated on the its roadmap!

WOW if they do release Barton then the Athlon will be in the mid to upper 2ghz's. That is fast processor!
 
Yeah I knew AMD wouldnt have jumped to 333fsb until proper chipset support was available!

Just like you knew AMD would be going to 512k L2 cache.

Hey lets just all wait for Thoroughbred! AMD is reportedly going back to 512k cache link

Anyway KenAF just beat me to the Ace's hardware info. If it is correct and the KT400 chipset comes out sometime this year , 333FSB athlons will still not be on the market.
 


<< Well, I just don't see it happening. Everyone will require new motherboards for Clawhammer....and I just can't see AMD requiring new motherboards for new Clawhammer and Athlon processors at roughly same time. Many will complain that AMD is forcing them to upgrade their PCs just to use newer processors. >>



They can and will do whatever they please.. think about Slot1 --> S370, P4 423 --> P4 478

It's not a big deal to them... companies do it all the time, they don't care about the consumer, they want to force ppl to buy more, so they can make more...
 


<< I am disappointed that AMD has not added more cache (from 256k ito 384k or 512k on L2) in Thoroughbred to continue a leading performance against Northwood. I guess Northwood continues holding its lead over AMD in performance until Hammer comes out in late 4 QTR this year... >>




They did that because it would have amounted to a very small performance increase.


Its much better to have s smaller die and be cheaper.


Check out this interview by Sanders where he mentions AMD's aim to cut costs.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top