Sign in with
Sign up | Sign in
Your question

Future cpus?

Last response: in CPUs
Share
July 29, 2002 11:58:12 PM

From what I've heard so far, AMD will have Barton after T-brd, and Hammer after Barton. (Correct me if I'm wrong) What are the spec. of them? FSB? Any changes to cache? It seems to me that T-bred can be used on current XP motherboards, but what about Barton? T-bred seems like it's just an XP but smaller. Therefore it has different power regulations. Otherwise, everything else is the same (or seems to be the same. Again, correct me if I'm wrong!) What about Barton? What are the differences between Barton and T-bred and T-bred and XP?

What about Intel? I hardly find anything regarding Intel's future processors in this forum. Can anyone help me or provide me a link?
Thanks a lot!

More about : future cpus

July 30, 2002 12:10:14 AM

Uhm, let's see here...

Barton
.13um process
*rumored and desired* 166Mhz FSB, more likely 133Mhz FSB
*possible* 512kb of L2 cache
improve tlb and branch prediction
heat spreader *hopefully*

K8 Hammer
Hypertransport and Integrated memory controller, *possible* "no" FSB
.13um process, later on .9um process
not sure about L1 cache, but initial 256kb L2 cache and later on, 512kb for Clawhammer, and up to 1Mb for Sledge

Pentium4 Prescott
800Mhz FSB
.9um process
cache most likely to remain the same

Not sure after that, but I believe Intel will move on to their Tejas core with the 1200Mhz FSB in .9um process


"When there's a will, there's a way."
July 30, 2002 12:26:26 AM

Pretty much yep! Though the 166FSB (DDR 333) for the Barton is defintely a rumour and not likely, but I'd personally love to see it. Bartons cache by all accounts is supposed to be 128K L1 and 512K L2. Until offical spec release or actual release of all of these chips it's all hear-say.

Hammer will have an on-chip memory controller, but it will have the option of being turned off, thus allowing the easy transition to future memory technologies such as DDR II w/o needing an entirely new CPU. >> Quote from a article at extremetech "The first Hammer chips will include a memory controller that will allow up to DDR-333 memories to be used, according to Charles Mitchell, a strategic marketing manager at AMD, Sunnyvale, Calif. The next memory controller will support DDR-II, specifically the DDR II-400 speed grade. Mitchell said the switch will likely take place in 2004." <<

Really not too much info on prescott aside from .09 nm production. Some say it might have Yamhill (IA32/x64) extensions, others believe it won't. Something it will probably have is more cache than northwood. As well there may be a FSB increase, though the 800FSB is defintely a rumour and seem unlikely that intel would skip the 166MHz bus and jump to 200MHz (166 = 667MHz quad pumped / 200 = 800 quad pumped). Again we'll just have to wait and see as it's simply a guessing game until offical docs or the actual chips are released.

You can read that entire artcile about the hammer and memory controllers here > http://www.extremetech.com/article2/0,3973,386094,00.as...

"Opinions are like assholes, everyone has one"<P ID="edit"><FONT SIZE=-1><EM>Edited by nja469 on 07/29/02 08:36 PM.</EM></FONT></P>
Related resources
July 30, 2002 12:32:41 AM

Ok...well, I think Intel definately will include Yamhill extensions, but have them initially disabled with the first Prescotts, then decide later on whether to change that or not. They might do what they did with the original Pentium4, and ramp it up from 133Mhz FSB, to 166Mhz, then 200Mhz finally. Not just a rumor, but maybe just a gradual change.

"When there's a will, there's a way."
July 30, 2002 2:06:54 AM

Yeah, Prescott is supposed to be released @ 3.2GHz, but I'm sure the clock speeds will ramp up quite quickly in face of danger from Hammer. Also, I think they've almost said for sure that it's gonna start with a 667FSB, then ramp up as you said. Also, there's a good possibility of a 1MB L2 cache and 32K L1. 16K of it for instructions I believe.

:lol:  Finally, I get a capitalized title!! WOOT!! :lol: 
July 30, 2002 3:15:26 AM

They HAVE to use gradual FSB changes. Rambus is a slow turtle in an endless sahara desert right now. Nothing on the horizon that we can hear about and be sure about it.
The only way to use synchronous operation , which Intel seems to be very pressing on it, is to use DDR333, which is official and ready. DDR 400 may not even last, in light of DDR-II 400, though I dunno what DDR II 400 means in the end. However if Intel needs Rambus, they'd have to release 166MHZ synchronisity with their RDRAM, therefore PC1300, not PC1200. I dunno how high is RDRAM gonna scale before they need to do some rearrangements. And as for 800MHZ FSB, DDR400 is perfect while RDRAM may never even see it before years, since that requires PC1600. Of course I am just thinking that way, but I am also basing this on real facts which are that Rambus is SLOW as hell these days. Has PC1066 become mainstream? Has there been at least 8 out of 10 PC stores in a region, selling it? Doubtful...

--
The sound of determination is the echo of will...
July 30, 2002 3:32:39 AM

Ok for the record (this is not rumor this is just the whole truth)

Barton WILL have 512L2 cache

Barton WILL NOT pump up the FSB, still stuck at 133 :( 

Some other minor optimizations- they should be able to ramp up clock speed a bit, but nothing drastic. Socket A will be obsolete by end of '03

Never underestimate the bandwidth of a station wagon full of tapes hurtling down the highway
July 30, 2002 4:59:31 AM

O.k, now I have another question though...
From what I've read so far, it seems to me that for all amd's 133 fsb cpus, the maximum data transfered from the cpu is only 2.1GB/sec...and hence ddr 266 will be just enough for that. It seems that it is also one of the reasons why moving from ddr 266 to ddr333 doesn't yield a significant performence improvement with athlong cpus. (I am not familar with p4 cpus and hence I can't tell whether the current memory bandwidth available to p4 is enough or not) Based on these, it seems to me that with all these coming mobo with higher and higher memory bandwidth isn't really going to make a big change in terms of performence. (Well, I think it's another story if you're overclocking)
Am I right about these?
If that is right, that means moving from ddr266 to ddr333 isn't really going to improve much, and same with ddr400 or even the coming nforce2 with dual channel ddr400 right?

I don't know whether I'm right on these or not. These are just conclusions based on what I know. At first I was quite excited about the nforce2 with its dual channel ddr400 support when I saw the reviews on nforce2 chipset, but when I got this conclusion today... I'm just not all that excited about the coming future memory modules until cpu can actually use all those memory bandwidth...
July 30, 2002 5:13:10 AM

Another thing to add, I'm not saying all these future memory modules are useless. I'm sure they're good for overclockers as I think people often overclock by pumping the fsb and the ram though I never overclocked my systems. However, a lot of my friends don't overclock and I think the majority of the market don't over clock as well...therefore, it seems to me these memory are not attracktive to general users (people who don't overclock). Then again, surely these mobos features ddr333 or above or the coming ddr400 will be advertised about how fast the memory on these boards run regardless they don't improve the overall system performence that much. owww....
July 30, 2002 11:32:01 AM

*shrugs* History manages to repeat itself. Here's a few examples.

The first majorly popular overclocker, the Celeron 300a, with a 66Mhz FSB and a 4.5 multiplier. It easily was raised to a 100Mhz FSB, which hugely increased performance due to the additional Mhz, increased memory bus, and already nice on die L2 cache (one of the first socket 370 Intel chips).

Then, the Athlons, Pentium3's, and future CPU's came on. Most of them benefited from a higher FSB, in a sense that their memory could never be fully quenched. Fast memory is always a good thing, and the FSB HAS to deliver it to the CPU efficiently.

If the Athlon Barton does indeed use a 166Mhz FSB (which would benefit us all), it provides us with two interesting standingpoints. Barton will immediately benefit from the increased memory channel. It's likely that the Athlon can still make use of faster memory, beyond 2.1Gb/s. Also, if 166Mhz is considered "stable" for the Barton, it opens the way to a possible 200Mhz FSB, which has only been considered a myth of recent times.

"When there's a will, there's a way."
July 30, 2002 2:09:57 PM

I believe for one, possible, just *maybe*, they will start with the same 8kb L1 cache and trace execution with 512kb of L2 cache, then ramp up to 16kb or 32kb of high speed L1 cache. Maybe even increase the number of execution trace cache entries. L2, yes it will likely be moved up to 1mb of even more.

"When there's a will, there's a way."
July 30, 2002 2:29:42 PM

From something I just thought of, I think the Intel P4a and p4b has a clock speend of 100mhz and 133mhz respectively. They are quad-pumped(though I have no idea how it is quad-pumped), therefore, the p4a processors have 400fsb while p4b have 533fsb. As for AMD athlon, it has a clock speend of 133mhz and is double-pumped(again, no idea how. Explanations will be appreciated), and hence athlon has fsb of 266mhz. With athlon at 266fsb, cpu maximum transfers 2.1gb/sec of data wich ddr266(pc2100) is enough to handle all that. As for p4b, it maximum transfers 4.2gb/sec and therefore pc1066 which is capable of 4.2gb/sec max. would be just enough for the p4b series.
This is just to correct my previous post.

What about Itanium? I've heard of this cpu in this forum..what is it?
July 30, 2002 2:45:27 PM

Yeah, the Pentium 4's utilize a quad pumped FSB, meaning it provides a bus with four times the amount of memory bandwidth. The Pentium4b for example, has a base FSB of 133Mhz, quad pumped to 533Mhz. That value is then double pumped, as with DDR, to an end grade of 1066Mhz.

The Athlon has a 133Mhz FSB, double pumped. Ram is double pumped simply by utilizing more of a single clock cycle. SDR takes advantage of only the rising edge of a clock cycle, at the speed of 133Mhz. DDR uses the rising and falling edges of a clock cycle, for twice the speed, hence double pumped, or 266Mhz.

"When there's a will, there's a way."
July 30, 2002 3:08:12 PM

I see, thanks a lot for the info!
Does anyone know the release date for Barton then?
I'm planning to get a new pc before the end of this year...but does anyone know what cpus will be available by the end of this year?
July 30, 2002 3:11:12 PM

I believe it will be out around September...i'm not positive on the release date. I know for sure though, Prescott should be out by early/mid '03

"When there's a will, there's a way."
July 30, 2002 3:22:43 PM

Barton will be released in Sept??
You mean this coming Sept?
July 30, 2002 3:37:02 PM

Yes, I think

Sometime Q3 of '02 (this year, late fall!)

"When there's a will, there's a way."
July 30, 2002 3:58:53 PM

I just build a system for one of my friends, based on a A7V333-RAID with 256 Mb of Samsung PC2700 running an unlocked and UNDER volted Duron Morgan 1200 ( 7.0 x 166, vcore @ 1.45v ).

It's a lowly 80$ ( Canadian currency ) Duron and it can run the 166 FSB with no problems. There seem to be no physical limitations that could keep the core from running unlocked on a FSB that is 66MHz out of specs.

My guess is that AMD won't officially launch a 166 FSB chip for marketing reasons, but then, once you unlock the chip, the only limit seems to be the MOBO northbridge and supported FSB multipliers in the BIOS.

Fok Speling Misstake
July 30, 2002 5:06:01 PM

Oh wow, I didn't know that!!
Will current mobo be able to support that then?
If it's releasing in Sept, how come no one is really sure about its spec.?
July 30, 2002 7:50:54 PM

Anyone?
Does anyone know whether Barton can be used on current mobos or comanies have to come up with new design/new chipset for Barton support?
Also, can someone confirm me about Barton being released in the coming Sept. and how come its spec. is still uncertain now though it will be released in Sept.?
July 31, 2002 12:51:13 AM

I have not seen one Duron in my life that can do a FSB above 133Mhz. Get me some pictures of the computer and WCPUID, and i'll believe it. Especially undervolted, you seem to be making a poor lie. When I see it, I will congratulate. An idiot posted an Athlon XP 1500+ they "said" they got up to a 200Mhz FSB. An obvious lie, that would be an Athlon XP at 2.0Ghz with a 200Mhz FSB. Too many lies to accept it at face value.

*shakes head* the limit isn't exactly the northbridge or the chipset. It is the capabilitites of the chip. *most* AMD DDR chipsets can do a range of FSBs, from 100Mhz to 255Mhz (the highest I think was 255Mhz on a kt333 chipset).

Quote:
and supported FSB multipliers in the BIOS


FSB's don't have multipliers, they are a speed, in Mhz.

"When there's a will, there's a way."
July 31, 2002 1:19:03 AM

Too bad I can't attach a file to my post =/ I'll send you a screen capture from wcpuid right now, I hope you provided a working e-mail in your THGF subscription, if not, feel free to foward me your e-mail address to sidvicious@ca.inter.net or PM me... For the 166 x 7.0 Duron, you'll have to wait until I get in toutch with my friend.

BTW, sorry for my lack of correct english grammar, english is a secondary language for me. That explains why you did'nt get my meaning when I was talking about the multipliers support from BIOS ( i'm using a A7V333-RAID, the lowest multiplier for this particular board is 6.5 ) that's why I was refering to BIOS multiplier support.


Fok Speling Misstake
July 31, 2002 1:24:54 AM

Quetzacoatl_@hotmail.com

I'm seriously interested in seeing a 166Mhz FSB Duron. Don't worry bout your english, it doesn't matter.

"When there's a will, there's a way."
July 31, 2002 1:46:27 AM

Argh ! I just sent the e-mail with the .bmp attachement to RedWyvren87@hotmail.com I'll be sending you a capture from my friend's Duron @ 166 x 7.0 as soon as I see him.


Fok Speling Misstake
July 31, 2002 1:50:32 AM

Okyday, sounds good

I maintain two email addresses by the way, if you didn't notice

"When there's a will, there's a way."
July 31, 2002 3:53:02 AM

Interesting reading on HardOCP about AMD's fsb, looks like the t-bred 2400+ and 2600+ may have 166 fsb. Its a he said she said, but AMD knows the problems with the tbred and has a new stepping coming out fixing them and keeping them at least close with Intel until Hammer comes out.

If this is true, we could see a large jump with new tbred in performance.

MeldarthX
July 31, 2002 4:10:50 AM

You must tell me where you get your informstion because theyare far from what i got.

The day i meet a goth queen that tell me Intel suck.I turn in a lemming to fill is need in hardware.
July 31, 2002 11:40:25 AM

You want to here my reading sources??? Ok...

THG (well duh...)
[H]ardOCP
Anandtech
Ars Technica
ZD Net
The Inquirer
Local computer shop-Blue Point Computers
<A HREF="http://www.bluepointcomputers.com
" target="_new">http://www.bluepointcomputers.com
</A>
3DGameman kickass reviews

I can't think of any others on the top of my mind, but if I remember, i'll post an edit.


"When there's a will, there's a way."
!