Prescott vs Athlon64?

[unclesasamon]

Hey,
I'm in the business of buying a new computer in the next few months and I was wondering if you guys can give me an unbaised opinion of whether I should get the Athlon64 or the prescotts coming out in December? Thanks!!!! :D

[Icarus]

no one can really tell yet. we'll have to wait and see some benchmarks before we decide which is better. but for now, I personally think that the Athlon 64 looks more promising.

[Chris_F]

I can't wait to see the new Prescott processors. I would get one once they release the T socket.

[AMD_Me]

AMD is running low, so they will release socket 940/939 athlon 64s, then move to socket 754. With yields and overclockability low (guranteed bad ocability becuase we have seen the opteron) the choice for me, a once die hard amd fan, is clear. get the Prescott.

[modenaf1]

Originally posted by AMD_Me
AMD is running low, so they will release socket 940/939 athlon 64s, then move to socket 754. With yields and overclockability low (guranteed bad ocability becuase we have seen the opteron) the choice for me, a once die hard amd fan, is clear. get the Prescott.

why are they gonna move from 940 to 754? isnt the higher the number the better? the data can be sent with more pins?


anyways, i hear that opterons dont even need active cooling!

[NookieN]

Originally posted by modenaf1
anyways, i hear that opterons dont even need active cooling!

That is extremely unlikely. The TDP for the current for the higher speed Opteron is about 85W. There's no way you can passively cool that (aside from maybe a 2Kg heatsink). Athlon64 will probably run slightly cooler, but not if it has a higher clock rate.

[Mark Larson]

Originally posted by NookieN


That is extremely unlikely. The TDP for the current for the higher speed Opteron is about 85W. There's no way you can passively cool that (aside from maybe a 2Kg heatsink). Athlon64 will probably run slightly cooler, but not if it has a higher clock rate.
Its not TDP, and that's the maximum number for the whole range - AMD specify their max number so that the system integrator can install a cooling system that doesn't need to be upgraded over the lifetime of the processor (Intel also do the same for the Itanium2 - the number is 130W)

[c627627]

Unless you have specific needs, AMD Athlon XP and the upcoming Athlon FX that will fit into nForce2 mobos will still be there and highly overclockable... for dirt dirt cheap.

Rememebr that right now:
1700+ T-Bred B > OC to 3000+ = $49
Thermalright all copper SK-7 heatsink = $14.99
nForce2 mobo with great onboard sound and lan and firewire and capable of seeing all multipliers on that unlocked T-Bred B = $83
======
$146.99...

Why not buy a few or 10 of these for the price of that Prescott (unless you have specific needs, of course.)

[climbski]

No one will know until we get em in our rigs and see what they REALLY do.

[BigJk-47]

I'd go with whatever one has the coolest name and most commercials :p

J/K

[cack01]

I would take the prescott.

The A64 is just the tone downed version of the Opteron. The Opteron is beating the P4 on some stuff and losing by a lot on others. Also as was mentioned, it does not O/C well. This means they have some big hurdles to overcome in order to scale the speed higher.

The Prescott is basically the P4 with more cache and faster clock speeds. If Intel gets the heat issues undercontrol, I do not see how the prescott can be beat in 32 bit mode. Also it has some new instructions, but of course we will probably never use them.

[Chris_F]

There is no need for a 64-bit processors at this time.

I see some really high clock speeds for the Prescott. I also see some really nice overclocks. :)

Just use my avatar to guide you to the right CPU.

[unclesasamon]

Yea, I was thinking prescott, I just don't want to wait :mad:. Oh well, if it's worth the wait, I'll be a happy camper :cool:

[MetalStorm]

In my opinion, I think the prescots will be the better buy, heat really isnt a problem in my book as you can always have a bigger heatsink or a larger fan... Anyway, if you really like playing UT2k3 then maybe get the A64 as UT2k3 is able to do 64bit and I guess should go quite some faster, thats about the only advantage I can think of at the moment!

Otherwise, go with what c627627 said, get a 1700+ overclock it, there you go, an easy 3000+ cpu for less and you can do it now, no waiting involved, I never thought I would ever get a "top of the line" computer as they cost so much, which is why I went the way of the 1700+ damn its fast ;)
Really, an overclocked 1700+ is as fast as anything you can buy today, but it costs a lot less...if you dont mind the noise (having said that, I cant see how the prescots can be anything but noisy). For me the choice was simple, I wanted a fast computer, and I wanted it now, and thats exactly what I have.

I hope that was of some use... :)

[Aslan]

At this point, IMO it looks like the Prescotts will be better, but honestly, you can't judge either of the CPU's at this point, let alone make a conscious decision of which one you should purchase.
I know waiting is a pain, but it's really the only thing you can do if you want either of these CPUs. :)

[climbski]

A nearly pure guess is that the prescot will be better ... but I am LMFAO at anyone who really is sure at this point. Since nobody has either one to test...LOL

Sorta Like askin which will be better the Pentium XXIII or tthe AMD ATHLON 512 at 1.2 Terahertz???

[MetalStorm]

Originally posted by climbski
Sorta Like askin which will be better the Pentium XXIII or tthe AMD ATHLON 512 at 1.2 Terahertz???

What?! Its obvious, the Athlon is the clear winner! duh! :p

[Chris_F]

1.2 Terahertz? :eh?:














I wish! :clap: :beer:

[Graphic67]

Just some random thoughts on these two yet-to-be-seen processors....

Internet "news" sites have reported that Prescotts are going to be 100W cores in the first release. Tier one OEM's will not like having to employ extreme cooling solutions for stock speed machines. Some speculative reports (from Japan) show later Prescotts reaching 140W.

The Opteron/Athlon 64 will be easier to cool than the current Athlon XP due to the integrated heat spreader which was not available on the XP cores. Some Opteron server configurations use pasive heatsinks on the cores, relying on airflow through the 1U case.

There are necessary differences between a server chip (Opteron) and a desktop chip (Athlon64) regarding stability and these differences extend into the chipset area as well. The Opterons with server grade chipsets and motherboards would not be expected to be very overclockable. Opterons on more aggressively designed boards for the enthusiast market (that's us by the way) should have the controlls necessary for overclocking. I can not say that I have read even one article dealing with overclocking the Itanium.

AMD issued a statement some weeks ago claiming that they would have 750,000 Athlon 64's ready to ship at the official September launch.

This is speculation on may part: The 940 pin and 939 pin Athlon 64 will be used in most desktop systems with the 754 pin version finding it's home in notebooks (one was demonstrated at the Million Man Lan in June) and as the budget 64 bit chip (64-bit Duron?))

[Mark Larson]

The Opteron/Athlon 64 will be easier to cool than the current Athlon XP due to the integrated heat spreader which was not available on the XP cores
That is false.
Some Opteron server configurations use pasive heatsinks on the cores, relying on airflow through the 1U case.
That is because of the need for redundant cooling and lack of space for HSFs in 1U servers.

[Graphic67]

M. Larson...

Would you mind explaining your "That is false." statement?

Cooling of 1U Opteron servers:
Sanmina/Newisys server uses side-mounted active cooling.
CCSI server uses passive cooling with case blowers.
IBM eServer 325 (cooling unknown)

Cooling of 2U Opteron server:
XI Computer uses 'traditional' top mounted active cooling.

[Mark Larson]

Would you mind explaining your "That is false." statement?
Its simple really. A heatspreader doesn't mean that the chip will be easier to cool as-is.

(However, a side-benefit of the IHS would be that greater force can be exerted more reliably on the die by the heatsink, but this force can not be greater than the same amount as with a naked die - the pressure on the die is the same notwithstanding the existence of the heatspreader.)

[gamefoo21]

every time i see that damned gator icon... oooh your icon makes me mad.... oooh how i hate that proggie...

anyways back to the thread at hand...

its true the new athlon 64's are just toned down versions of opterons.

the Athlon FX's are toned down versions of the A64.

what they get added on... SSE2, 64Bit abilities, Onboard mem controller only opterons and the most expensive a64's get dual channel the rest get single, and they get new sockets.

Opteron: 1meg cache
A-64: 512kb cache
A-FX: 256kb cache(its a new duron)

prescotts are p4's with Htt2, SSE3(NPI's :P), Increased Cache Size, 90nm architecture, and a new socket type.

hopefully the celerons based on this design will get 512kb cache to compete better with the FX's.

But right now its too soon to tell if the A-64's will out do the P4-D's.

If i made some mistakes please tell me.

[climbski]

Heatspreaders actually decrease heat transfer ie they cause higher temps. As has been proven by removing Intel heatspreader. (difficult not impossible) Heatspreaders do make it easier to lock the multi's and restrict access to bridges.... Not a good sign when a chip company adds heatspreaders.

[cack01]

The main thing I see that is the problem, is that we are not sure how well the future chips will scale.

We know the Opteron scales horribly, thus it will be a poor overclocker. We know little about the prescott scaling, but it will probably NOT be like what we saw from the move to .13 . It does have heat issues, and therefore it is resonable to assume that it will not scale well eighter.

Thus if we look at just stock speeds, I still feel the prescott will win, as the A64 is basically a nutered dog and even its big brother is having problems winning right now. I just do not see how it keeps up with a P4 w/ twice the cache and 600 more mhz.

[NookieN]

Originally posted by gamefoo21
the Athlon FX's are toned down versions of the A64.


I'm pretty sure the "FX" is a chopped version of Barton, and not related to the Opteron.


Originally posted by climbski
Heatspreaders do make it easier to lock the multi's and restrict access to bridges.

Not really. Intel chips are locked with or without the presense of a heatspreader.

[Zuzzz]

The heat spreaders on Intel P4's do NOT hide the bridges. There is nothing but the surface of the chip die underneith. . persoanly I would remove the IHS and apply good thermal paste (AS3)and replace it. . then install the HSF. . IMHO. To me, the chip die is a bit fragile to go directly against the HSF(If you are VERY careful and good fine. but Im old and a bit clumsy lol). Your results may vary.

[Gregory_WE]

I can't wait cause I think I'm going to be doing my 2-year cycle major upgrade next summer and by that time, Prescotts should be out and totally under control (stable, not extremely high priced), not to mention quite a few mobo options... also I'll probably be able to get stuff like PCI Express maybe DDR2? *drool*

[OC-Master]

Originally posted by NookieN


That is extremely unlikely. The TDP for the current for the higher speed Opteron is about 85W. There's no way you can passively cool that (aside from maybe a 2Kg heatsink). Athlon64 will probably run slightly cooler, but not if it has a higher clock rate.


Thats not true, over 90% of all Opteron based systems use passive cooling with no fans at all over the heatsinks. The reason the Opterons can do this is due to the fact they have a technology advance Silicon On Insulator (SOI).

Even at 2GHz, an Opteron can still run on passive cooling with the AMD stock heatsink solutions.

Infact, several HSF companies have gone the extra step to produce passive sink cooling solutions for Opteron based PCs. OCZ is a great example!
http://www.xbitlabs.com/news/other/display/20030604182552.html



OC-Master

[NookieN]

Originally posted by OC-Master
Thats not true, over 90% of all Opteron based systems use passive cooling with no fans at all over the heatsinks. The reason the Opterons can do this is due to the fact they have a technology advance Silicon On Insulator (SOI).


Can you show me some of these systems? Sure there are plenty of 1U heatsinks that don't have a fan bolted to the top of them, because they wouldn't fit into 1U as such. However, pretty much every 1U Operton setup that I've seen has fans or blowers to the side of the heatsink pushing air through the fins. That's _not_ passive cooling.

SOI reduces power output somewhat, but it doesn't guarantee that the chips won't require active cooling. Even the AMD thermal design guide calls for a fan on the heatsink.

[climbski]

You guys are still argueing about things that don't exist???? LMAO!! By the way how many angels can dance on the head of a pin???

[climbski]

SORRY!!! Why did this double post???

[Graphic67]

Originally posted by climbski
You guys are still argueing about things that don't exist???? LMAO!! By the way how many angels can dance on the head of a pin???

The discussion had taken a turn into the cooling of the Opteron which does exist. Here is one link showing the server chip cooling: http://168.143.107.122/articleview.cfm?articleid=1270&page=5

[cack01]

Originally posted by Graphic67


The discussion had taken a turn into the cooling of the Opteron which does exist. Here is one link showing the server chip cooling: http://168.143.107.122/articleview.cfm?articleid=1270&page=5

Thats active cooling if I have ever seen some.

When ever there is the slightest breeze of air, that is called active cooling. Some times you can argue "well the air isn't being forced over the heatsink". True, but by most people's definition its active cooling.

[Jay23]

Originally posted by cack01


Thats active cooling if I have ever seen some.

When ever there is the slightest breeze of air, that is called active cooling. Some times you can argue "well the air isn't being forced over the heatsink". True, but by most people's definition its active cooling.

i dont mean to go off topic or get into an arguement so i apoligize ahead if that happens..

Anyways. I think you're just saying that because amd's cpu uses passive cooling and the intel's(least from what i hear) doesnt :rolleyes:

[NookieN]

Originally posted by Jay23
Anyways. I think you're just saying that because amd's cpu uses passive cooling and the intel's(least from what i hear) doesnt :rolleyes:

This is basically just arguing semantics. Passive cooling means you are not using _any_ mechanical means to cool something. If you have a mechanical device moving air across the heatsink, that's active cooling. Even if you just have a very large heatsink that catches the air moved by the case fans, that's still technically active cooling (just a less powerful, but probably quieter version of it).

[Jay23]

Originally posted by NookieN


This is basically just arguing semantics. Passive cooling means you are not using _any_ mechanical means to cool something. If you have a mechanical device moving air across the heatsink, that's active cooling. Even if you just have a very large heatsink that catches the air moved by the case fans, that's still technically active cooling (just a less powerful, but probably quieter version of it).

ill end it this with this last comment.

When you think of passive, what do you automatically think of? Passing over(when ur thinking of air) and when think of active or what not its basically the fan on the heatsink. So the most common thing someone will think is passive cooling is what he was talking about. Im not saying your definition is false, thats just what my opinion on the matter is :)

[Graphic67]

If your northbridge has a passive heatsink, does it become active cooling because you have one or more case fans?

[ol' man]

Those server opterons were cooled actively. Notice the "blowers"? Those are not fans they are blowers, very similar to the blower that the invidia 5600fx Ultra originally had. You know the vid card that was called the dustbuster and other various names.

If you look close to the opteron cooling solution you will see it is called the "wind tunnel". Since they use the term wind tunnel I would assume that"wind" is moving pretty fast.
http://www.socketa.com/articleimages/cpu/hammer/opteronlaunch/ccsiinterior.jpg

[cack01]

Originally posted by Jay23


i dont mean to go off topic or get into an arguement so i apoligize ahead if that happens..

Anyways. I think you're just saying that because amd's cpu uses passive cooling and the intel's(least from what i hear) doesnt :rolleyes:

No, I said that AMD uses active. I agree with NookieN.


ol'man is showing the pic that I am refering to as active cooling.

[OC Detective]

Originally posted by cack01


No, I said that AMD uses active. I agree with NookieN.


ol'man is showing the pic that I am refering to as active cooling.
Using your somewhat restrictive definition give me an example of any system that uses passive cooling? i think some people are getting confused with the adverb passive and thinking its may be a derivation of the verb "pass" - its not!
Simple analogy a passive smoker doesnt smoke directly but takes in smoke from other smokers around them - the cooling system on this is not DIRECTLY onto the cpu but takes in cooling from other areas. It is passive cooling.

[NookieN]

The terms active and passive cooling are not specific to computers. Passive cooling has been a school of thought in building design for a long time. A building is passively cooled if can be cooled entirely by things such as natural ventilation, evaporative cooling, or radiative cooling. By the same token then, passive cooling in a computer system would rely only on the natural abilities of the heatsink and case to dissipate heat without the action of fans or forced water.

To say that an Opteron could be passively cooled would mean that it could run reliably with no fans in a case with no forced air. My guess is that such a setup would have problems, but I don't know that for certain. I doubt anyone using an Opteron in a mission-critical application would want to find out.

Frankly though, I don't see why this is turning into a debate between AMD and Intel followers. The same cooling mechanism people are talking about for the Opteron has been used in 1U Xeon systems since before the Opteron was released. Having a heatsink with a blower to the side of it is not due to the chip used, it's just a peculiarity of the 1U format.

[cack01]

Originally posted by OC Detective

Using your somewhat restrictive definition give me an example of any system that uses passive cooling?

Best thing I can think of would be an old NES. I think the VIA chips can be cooled passive.

I guess a good definition(my personal definition) of a chip being passive cooling would be that the chip can be cooled by a heatsink which does not need air going over the fins. Thus the heat is passed from chip to sink and then the heat is radiated into the surounding air. Thus the energy is passed though *equalization*)
*I know there is a better term than equalization, but I can't think of it. :)

[ol' man]

Originally posted by OC Detective

the cooling system on this is not DIRECTLY onto the cpu but takes in cooling from other areas.

So your saying the big copper heatsinks directly attached to the opterons are not cooling them directly.

[unclesasamon]

so....about those prescotts and Athlon64's :p

[Graphic67]

Originally posted by unclesasamon
so....about those prescotts and Athlon64's :p

It's all rumor and speculation until sometime in September...

[Akira283-IGN]

Originally posted by Graphic67
The Opteron/Athlon 64 will be easier to cool than the current Athlon XP due to the integrated heat spreader which was not available on the XP cores.


What you obviously don't realize is that the "heatspreader" actually increases temperatures. Overclockers.com (this website) proved this in an article several months ago. think about it - all the thing does is add a layer of metal between the CPU core and the heatsink (which is already a big chunk of metal). By adding another layer, you're actually decreasing the effective heat transfer. A better name for the metal cover would be "core protector."

[Graphic67]

I appreciate the corrections on the matter of the heat spreader. As one who did not follow the debate associated with the Pentium heat spreader, I was not aware of the down-side to its use with regard to heat transfer.

On a heat related note: the Inquirer has a brief article today claiming confirmation from Intel that the Prescott core is "close to" the 100W figure previously reported. Also, in the same article it was also confirmed that revision 1.5 Canterwood and Springdale boards will be required to run prescott. Here is the link to the article: http://www.theinquirer.net/?article=10862

[Overclocker550]

The a64's arent looking good. The xp3400 may be a 1.8GHz a64 and its hardly faster than an xp3200 barton at 2.2GHz and besides an xp2100 at 2.4GHz would match both of those other cpus for a fraction of the price too. The xp3200 is about on par with a 2.8c so the a64 1.8GHz may match a 3.0c northwood, but imagine a prescott at 3.4GHz, itll own the a64's unless amd can get em around 2.4GHz

[Graphic67]

A report from a german site (planet3dnow.de) lists some interesting Athlon 64 model numbers...

http://www.planet3dnow.de/cgi-bin/newspub/viewnews.cgi?category=1&id=1060102789

To be announced September 23rd:

* AMD Athlon 64 FX-51 : 2.0 GHz, Dual-Channel DDR, Sockel 940
* AMD Athlon 64 3200+ : 2.0 GHz, Single-Channel DDR, Sockel 754

Versions in 2004:

* AMD Athlon 64 FX-55 : 2.4 GHz, Dual-Channel DDR, Sockel 940
* AMD Athlon 64 FX-53 : 2.2 GHz, Dual-Channel DDR, Sockel 940
* AMD Athlon 64 3700+ : 2.4 GHz, Single-Channel DDR, Sockel 754
* AMD Athlon 64 3400+ : 2.2 GHz, Single-Channel DDR, Sockel 754

Note that the single channel 754 pin version is the only one sporting a Performace rating. The almost-opteron dual channel FX uses an arbitrary model number.

[OC Detective]

Originally posted by ol' man


So your saying the big copper heatsinks directly attached to the opterons are not cooling them directly.
Its a case of specific heat transfer rather than cooling. The copper heat sink sitting by itself is not an active cooling system - a fan on the other hand would be.

[MrMarbles]

Originally posted by Gregory_WE
I can't wait cause I think I'm going to be doing my 2-year cycle major upgrade next summer and by that time, Prescotts should be out and totally under control (stable, not extremely high priced), not to mention quite a few mobo options... also I'll probably be able to get stuff like PCI Express maybe DDR2? *drool*

I was thinking exact same thing. Ill be upgrading I guess some time in second half 2004 and Prescott/DDR2/PCI Express combo sounds really nice.

[Chowdy]

IMO in the computer world, passive cooling is a cooling aparatus which has the single purpose of cooling a single object without the use of mechanical means (IE a bare copper heatsink on a cpu or NB etc....)

Active cooling is the opposite, a cooling device which both increases surface area (as does a heatsink) but also increases the amount of air which passes over that surface by various means (fan). An example of active cooling would be a heatsink with a fan on a processor, gpu, nb, etc....

The opteron cooling solution that you show which seems to be a blower blowing air accross the heatsinks is kind of passive, yet not. As, i could probably fit a ducting systm to my computer without altering any of the case fans and run my slk800 without the panaflo that it's attatched to. That'd be passive cooling, only really efficient as it directs the air which is also used to ventilate the case to cool the cpu.

just my $0.02

[Overclocker550]

oh looks much worse than I thought. an xp3200 athlon 64 running at 2GHz? man thats not much better than a 2.2GHz barton. amd has no chance against those prescotts

[Benno]

all is not lost!

http://www.theinquirer.net/?article=10880

[Sir Barton]

Originally posted by MetalStorm

Otherwise, go with what c627627 said, get a 1700+ overclock it, there you go, an easy 3000+ cpu for less and you can do it now, no waiting involved, I never thought I would ever get a "top of the line" computer as they cost so much, which is why I went the way of the 1700+ damn its fast ;)
Really, an overclocked 1700+ is as fast as anything you can buy today, but it costs a lot less...if you dont mind the noise (having said that, I cant see how the prescots can be anything but noisy). For me the choice was simple, I wanted a fast computer, and I wanted it now, and thats exactly what I have.

I hope that was of some use... :)

amen!

[Chrisdafu]

This is exactly like the Doom3 and HL2 debate, I wish people would stop talking about it until they are out, or at least one of them.

[Graphic67]

Originally posted by Chrisdafu
This is exactly like the Doom3 and HL2 debate, I wish people would stop talking about it until they are out, or at least one of them.

It could be worse... we could be arguing about which future cpu will be better at running those future games... :D

Thanks for the idea...

So, who want's to start?

[Chowdy]

From the benches i've seen, a uniprocessor setup with an opteron 1.8ghz spanked a 3.0ghz P4, if i remember correctly. If that gives any idea to how the Athlon64 will perform, then i think a 2.2ghz Athlon64 would be quite speedy.

[NookieN]

Originally posted by Chowdy
From the benches i've seen, a uniprocessor setup with an opteron 1.8ghz spanked a 3.0ghz P4, if i remember correctly.

The term "spanked" is kinda subjective. If by "spanked" you mean "was slightly faster in some programs, are noticeably slower in others", then yes that sounds about right.

[overdoze]

I looked at a few benchmarks for 32bits and 64bits results. I can take an educated guess on prescot vx athlon64:
For performance point of view. Not price since we don't know how much they will cost and I know it is the most important factor in most cases.

if you play games get the prescot. Faster 32bits mode SSE3 and lots of cache.
If you are planing to build a servers with dual cpus and your apps are multi tasking, intensive apps, or a computer servers) get the dual athlon64 or dual opteron.

For me, as I use most the computer for folding, seti and running computing simulations jobs that requires at least 2 to 8GB of ram. I would stick with 64bit no question ask. Since the prescot system maximum ram is only 4GB. The questions to me then is whether to get opteron or which version of athlon64 ...
Tsk Tsk Tsk ... decisions :)

[OC-Master]

Originally posted by overdoze
I looked at a few benchmarks for 32bits and 64bits results. I can take an educated guess on prescot vx athlon64:
For performance point of view. Not price since we don't know how much they will cost and I know it is the most important factor in most cases.

if you play games get the prescot. Faster 32bits mode SSE3 and lots of cache.
If you are planing to build a servers with dual cpus and your apps are multi tasking, intensive apps, or a computer servers) get the dual athlon64 or dual opteron.

For me, as I use most the computer for folding, seti and running computing simulations jobs that requires at least 2 to 8GB of ram. I would stick with 64bit no question ask. Since the prescot system maximum ram is only 4GB. The questions to me then is whether to get opteron or which version of athlon64 ...
Tsk Tsk Tsk ... decisions :)


Hmm,,

well the Athlon64 actually has more cache than the Prescott. The Prescott has a total of 1056KB of cache while the Athlon64 has 1152KB of cache.

SSE3 is absolutely no match with 64-bit once games start coming out with 64-bit. 64-bit technology adds an automatic 20% boost to total system performance when both the OS and game supports it and most other cases even faster than 20%.

All the benchmarks that are out are clearly showing that even in 32-bit modes, the 64-bit Athlon64 and Opterons are smoking P4s and the AthlonXPs.

The only thing the Athlon64s need now are 0.09 Microns with SOI and a solid 3GHz. Pair that with dual channel DDR400 and your doing awesome.

Sounds like Sandiago is needed LOL


OC-Master

[gamefoo21]

last i seen opeteron's were getting there ****zy kicked by 32bit xeon's :O. and if it is really limited to 4gb then why can i go out and get a dual xeon board with 16gb of ram :P. its the os/northbridge that determines how much ram you can access cuz the old 486's could access terabytes of memory.

ok i just looked it up... seems like the e8870 can have 128GB of ram. and it only needs one cpu. so **** on that theory. shouldn't compare desktop chipsets to workstation chipsets.

eg:

E7505: 1-2 proc 16GB of ram. thats 8GB of ram for one processor if you use 2 of them.
E8870: 1-4 proc 128GB of ram. so if we split fine hairs here thats 32GB of ram for 1 processor if your using four.

i still don't understand why amd'rs have this fallacy. and really the server editions of windows allow you to address all that memory due to x86 extensions they carry.

[OC-Master]

Originally posted by gamefoo21
last i seen opeteron's were getting there ****zy kicked by 32bit xeon's :O. and if it is really limited to 4gb then why can i go out and get a dual xeon board with 16gb of ram :P. its the os/northbridge that determines how much ram you can access cuz the old 486's could access terabytes of memory.

ok i just looked it up... seems like the e8870 can have 128GB of ram. and it only needs one cpu. so **** on that theory. shouldn't compare desktop chipsets to workstation chipsets.

eg:

E7505: 1-2 proc 16GB of ram. thats 8GB of ram for one processor if you use 2 of them.
E8870: 1-4 proc 128GB of ram. so if we split fine hairs here thats 32GB of ram for 1 processor if your using four.

i still don't understand why amd'rs have this fallacy. and really the server editions of windows allow you to address all that memory due to x86 extensions they carry.


32-bit is limited to 4GB unless other methods are used. The maximum dimm size allowedable is 1GB (512MB on each side with each BGA chip being 64MB. 64 X 16 chips gives you your 1GB stick of memory.

On a reference motherboard design on the x86 platform, dimms are stated to have up to four slots for equiping memory. That gives you a total of 4GB of ram.

32-bit processors can also address up to a maximum of 4GB of ram. The only reason Xeons can support more than 4GB of ram is because they have more than one chipset and processor on board.

By having two processors installed, 32-bit systems can actually use 2GB ram dimms but with the catch being, each processor can access only half the dimm's memory at once. That gives you a maximum of 8GB of ram on 4 dimm slots.

Now, using multiple chipsets (dual), you can have up to 8 dimm slots for a total of 16GB of ram. Now, HLE Server Works chipsets can handle up to 12 dimm slots meaning up to 24GB of ram. Thats how you obtain more than 4GB on a 32-bit platform.

Then there is 64-bit PCs that can use a single stick of ram that could easily contain 4GB of ram on a single dimm for a total of 16GB of ram with only 4 slots.

We get are maximum memory values from a simple bit x bit value. 32 X 32 = 1024MB (unbuffered) maximum size for a dimm for a single 32-bit chipset and CPU.

64 x 64 = 4096MB (unbuffered) maximum size for a dimm for a single 64-bit chipset and CPU.


OC-Master

[chemmajik]

Some people should really keep up on there study's, the latest 246's running overclocked at 2.2 & 2.25 on AMDzones revised review have WON 15 out of 17 benchmarks with over 1000 more memory bandwidth alone then a 3.2ghz Pentium4, can a Prescott really make up that much, until I see it I don't believe it. Its either put up or shut up time now Intel lovers haha!

[NookieN]

Originally posted by OC-Master
well the Athlon64 actually has more cache than the Prescott. The Prescott has a total of 1056KB of cache while the Athlon64 has 1152KB of cache.

It's not very accurate to directly compare the Opteron and P4 caches. The P4's cache is inclusive, so it's best to say it has 1024k. The L1 instruction cache on the P4 is also measured in micro-ops, which doesn't necessarily correspond to Kbytes. Yes, the Operton actually has slightly more cache, but since it's exclusive there are various performance penalties when data is in one cache but not the other.

SSE3 is absolutely no match with 64-bit once games start coming out with 64-bit. 64-bit technology adds an automatic 20% boost to total system performance when both the OS and game supports it and most other cases even faster than 20%.

That's a bold statement to make considering that no chip with SSE3 exists yet and there are next to no applications with 64-bit support. Actually I don't think SSE3 will do much either, it's not intended to be anywhere near as important as SSE2. But simply compiling appliations with AMD 64-bit support isn't going to magically make them 20% faster. The few applications that currently exist in an AMD 64-bit form are almost certainly faster because they have been optimised for Opteron, not just because they're 64-bit.

All the benchmarks that are out are clearly showing that even in 32-bit modes, the 64-bit Athlon64 and Opterons are smoking P4s and the AthlonXPs..

Let's see some proof. I want to see a diverse range of application benchmarks that doesn't come from a site known to predominantly favor either AMD or Intel. Most of what I've seen shows that a 1.8Ghz Opteron is somewhat faster than a 3.06Ghz P4 at some things, and slightly slower at others. Sure AMDZone has shown that an overclocked 2Ghz Opteron beats a non-overclocked 3.2Ghz P4. I'm not entirely sure what that's supposed to prove though other than that the Opteron can be overclocked by an amazing 10%. Woohoo!.

[OC-Master]

Originally posted by NookieN


It's not very accurate to directly compare the Opteron and P4 caches. The P4's cache is inclusive, so it's best to say it has 1024k. The L1 instruction cache on the P4 is also measured in micro-ops, which doesn't necessarily correspond to Kbytes. Yes, the Operton actually has slightly more cache, but since it's exclusive there are various performance penalties when data is in one cache but not the other.



That's a bold statement to make considering that no chip with SSE3 exists yet and there are next to no applications with 64-bit support. Actually I don't think SSE3 will do much either, it's not intended to be anywhere near as important as SSE2. But simply compiling appliations with AMD 64-bit support isn't going to magically make them 20% faster. The few applications that currently exist in an AMD 64-bit form are almost certainly faster because they have been optimised for Opteron, not just because they're 64-bit.



Let's see some proof. I want to see a diverse range of application benchmarks that doesn't come from a site known to predominantly favor either AMD or Intel. Most of what I've seen shows that a 1.8Ghz Opteron is somewhat faster than a 3.06Ghz P4 at some things, and slightly slower at others. Sure AMDZone has shown that an overclocked 2Ghz Opteron beats a non-overclocked 3.2Ghz P4. I'm not entirely sure what that's supposed to prove though other than that the Opteron can be overclocked by an amazing 10%. Woohoo!.


And the point of this is?

There is no way that Prescott will be able to out perform a Athlon64, its as simple as that. Prescott uses the same old Net Burst architecture with the Quad pump interface. Athlon64 uses a brand new technology driven architecture. 64-bits is only the begining... Were talking intergrated memory controller along with the Northbridge now mounted inside the CPU.

The only thing that is saving the Net Burst architecture is MHz. its pretty hard to beat a 4GHz CPU with a 2.5GHz or 2.0GHz CPU in the competition.


OC-Master

[Jarlax3]

Head acheing....

Ok, back to something close to the topic. Does anyone have solid dates as to when the processors will be released? I heard September, October and December but which is right and for what processor????

[NookieN]

Originally posted by OC-Master
Athlon64 uses a brand new technology driven architecture.

Uhhh, what? "Technology driven architecture"? Come on now, be serious. What CPU design _isn't_ "technology driven"?

The architecture of the Hammer/Opteron/Athlon64 is nothing revolutionary. It's basically a K7 with a few tweaks, plus 64-bit extensions, the integrated memory controller, and HTT. Those improvements are all a step forward. They bring new strengths to the Athlon, but also new liabilities.

The only thing that is saving the Net Burst architecture is MHz. its pretty hard to beat a 4GHz CPU with a 2.5GHz or 2.0GHz CPU in the competition.

So... you want to be able to compare the chips Mhz for Mhz? What would be the point of that? It's old news that P4s run at higher clock rates than Athlons of comparable performance. That doesn't mean Prescott will be a bad chip. But I'm also not going to claim that it will be the greatest thing ever.

The Opteron definitely has the potential to be a very good chip. It also has the potential to ruin AMD. The chip is going to have to prove it can live up to the hype though.

[OC-Master]

Originally posted by NookieN


Uhhh, what? "Technology driven architecture"? Come on now, be serious. What CPU design _isn't_ "technology driven"?

The architecture of the Hammer/Opteron/Athlon64 is nothing revolutionary. It's basically a K7 with a few tweaks, plus 64-bit extensions, the integrated memory controller, and HTT. Those improvements are all a step forward. They bring new strengths to the Athlon, but also new liabilities.




The K7 and K8 have absolutely nothing in common besides the word "Athlon".

A brand new architecture is what the Athlon64 is. You cannot mix this technology with previous technologies such as an AthlonXP based rig. Different socket design, different bus setup, different memory controllers.. ext..

The Pentium5 (Prescott) is the same thing over again. It works on Socket 478 and uses the same old quad pumped architecture as previous northwoods.

The only real thing that will happen for prescott is the move to Socket T which is about revolutionary as it will go.


OC-Master

[]-[itman]

Originally posted by NookieN
Sure AMDZone has shown that an overclocked 2Ghz Opteron beats a non-overclocked 3.2Ghz P4. I'm not entirely sure what that's supposed to prove though other than that the Opteron can be overclocked by an amazing 10%. Woohoo!.

Actually, if you look at the numbers, the non-overclocked 2ghz opteron beats the 3.2ghz p4 on all but 3 out of 17 tests and 2 of those were from the SPECView program. Even the 1.8ghz 144 opteron held it's own against the 3.2ghz p4. If AMD gives the 2ghz opteron's a 3200+ rating, it looks like the chips will have breathing room against the 3.2ghz p4's and will probably be able to hold it's own quite well against the Prescott if they can get the scaling of the 64bit chips up like they hope.

Of course, that's all if AMDZone's testing holds up.

[Mark Larson]

Uhh, OC-Master, you're mostly wrong.

And NookieN, the integrated memory controller is truly a revolutionary thing.

Does anyone know whether Prescott will have PAE chipsets so that >4GB addressing isn't only reserved for Xeons and Athlon64s?

[Yuriman]

This is pretty pathetic. I belive when they come out I will get both and post benches on the web. If the prescot wins in MY benches, ill stick with those till the next generation chip(most likely), same goes for amd.

[NookieN]

Originally posted by OC-Master
The K7 and K8 have absolutely nothing in common besides the word "Athlon".

No, that is incorrect. They are quite similar. Here's one reference (http://www.extremetech.com/article2/0,3973,1156821,00.asp) for you. Note the comparison table between the Athlon and the Opteron. The basic design is very much the same.


Originally posted by Mark Larson
And NookieN, the integrated memory controller is truly a revolutionary thing.

It's certainly new to the x86 world, and it's a nifty thing. But I still don't think I would call it revolutionary. First, the Sun UltraSparc-III (and related chips) has an on-die memory controller, and it preceded the Hammers by quite a bit.

Aside from that, I don't think an integrated memory controller is a new and radically different idea in chip designs. It's basically the next logical step in an effort to make memory accesses as painless and transparent as you can to the execution core.

Let's think about how this was accomplished in the past. First, there's cache to speed access to adjacent and/or frequencly accessed data. Cache alone isn't very effective when you're working with large amounts of memory, so you add non-blocking caching to continue executing during misses and TLBs to avoid looking at the page table all the time. After that, we can add prefetch to get data we think we need before we need it. After that what's the next logical step? Well to some people moving the memory controller onto the chip itself seems like a reasonable progression.

Having the memory controller on-die definitely has big advantages. But it doesn't come without trade-offs. It ties a particular version of a chip to a particular memory type, so anytime a new memory type comes along the design must be updated. An on-die memory controller also gives the chips ungodly pin-counts. This is expensive both to package and to test. Most high-volume CPU testers in use today only support between 384 and 512 data pins. I'll bet AMD had a fun time figuring out how they were going to test 940 Opterons. Of course many of those pins are for power, but even so I don't doubt they're expensive and time-consuming to test.


Originally posted by Mark Larson
Does anyone know whether Prescott will have PAE chipsets so that >4GB addressing isn't only reserved for Xeons and Athlon64s

Since the launch platform for Prescott is Springdale/Canterwood, the answer would be no in that case. The question mark then is will Grantsdale support it when socket T is launched. My guess is "probably yes," but there isn't a whole lot of technical info available yet.