The Athlons Limits

[Lynx]

Has anyone got any idea on the limits of the architecture of the Athlon. In other words how fast is the fastest it will go.

[oc jason]

you are totally right- all the rest of the hardware is so far behind. If they had hardware that kept up with 1200+ procs then wed see a big difference- i mean we need stuff like 15k rpm IDE drives and like bigger then 4 or 8 gigs- instead of the only big drives being 7200. andyes the DDR is still a 133 base- if they made a DDR witha 150 bus speed base it be like 300 mhz bus speed. Speed is starting to run away with all of our procesors. I dont think that they'll hit 2 gig on a conventional Palamino setup. Yes they have 1.5 coming out, then say its a 1.7 ( its just an overclocked 1.5) and therefore less of a potential than say a true 1.7. I think they will have to go to 13micron core to pass 2 gig in a conventional system

[peppy]

dont u guys think the new nvidia based chipset boards would be a big improvement?, i beleve it would be, with that new AMD HyperTransport, and TwinBank Memory Architectur?,and other stuff

[oc jason]

I believe that the new nVidia nforce based board will be a improvement but not enough to make it worth buying it right away just becuase its a new platform board and is pretty new. if it turns out to be a good board then maybe later ill get one but not for now

[AmbientFiction]

Would71 (Jul 12, 2001 02:13 p.m.):
OK, in this post I WILL address the mass storage problem briefly. (after all, this is an AMD CPU thread and not a mass storage thread.. but it has been mentioned here and I'll follow up a little)

I don't feel raw spindle speed alone will be enough to bring drives up to speed with everything else. I think they are going to have to gradually make changes in the design of the drives to gain the most performance. Over on the SR forums, I posted a really crude drawing of one idea, making a 5.25" drive that would be essentially a self-contained RAID 0, seen by the system as one drive. (please forgive the crudeness, you can tell this isn't one of my better skills)

[img="http://a676.g.akamai.net/f/676/987/12h/photos.netclubs.com/live/photos/6/c/d/k/dkkfc3en8vah3bp80188nc11hc/dual_ide.gif"]

Even changes like this won't be enough, but they need to consider more things like this, along with current product developments, to try and catch up a bit with the runaway CPU speeds. (and video cards, for that matter)

I have a good idea they should make scsi as cheap as IDE drives. That would help with gains in CPU speed.

[Hoot]

To me, the only sensible solution to a mechanical device, slowing down a system of solid state devices, is to replace the mechanical device with solid state equivalent. This can obviously be done already, but the cost per megabyte is no where near mechanical storage devices. With how cheap you can obtain solid state storage, it amazes me that most OS's still insist on using virtual memory. Talk about penalizing everyone for the few who can't afford an adequate amount of RAM.

Hoot

[Lynx]

Ironicaly I did not set this post up for this purpose but I was talking about this while playing golf yesterday. The main problem as we could see was the pci bus. It seems to be limiting the whole system. what we need is a new standard. If there are any people who read this and have some influence we need to sent the pci card the same way as the isa cards have just done and quickly or we are going to go no where. The pci bus limits everything. It is the basis why we have the low transfer rates on hard disks. It cannot cope with the ammount of bandwith that is nowadays passed through it. we need an improvement soon.

[AMDGuy]

Lynx (Jul 13, 2001 01:32 p.m.):
Ironicaly I did not set this post up for this purpose but I was talking about this while playing golf yesterday. The main problem as we could see was the pci bus. It seems to be limiting the whole system. what we need is a new standard. If there are any people who read this and have some influence we need to sent the pci card the same way as the isa cards have just done and quickly or we are going to go no where. The pci bus limits everything. It is the basis why we have the low transfer rates on hard disks. It cannot cope with the ammount of bandwith that is nowadays passed through it. we need an improvement soon.

I've heard about some new serial connection that is supposed to be in the development stage. This supposedly takes the place of IDE and such. These connections look like round cables that plug into the mainboard.

I have not seen this and do not have a link. I just saw someone talking about it in a hardware forum a while back.

[Lynx]

That wont help the problem much. What most people dont realise is that the pci bus is 32bits wide and is 33mhz fast. That gives a maximim transfer rate of 264mbps Stick only 1 udma100 hdd on there and bang a third of your bandwith gone. If you have a raid setup it is even worse. Some of you will be thinking well what about the other 164mbps. Well the CD/DVD drive takes up about 33 of it. A sound card will take a sizable protion and soon you have not much left. Haveing a PCI sound card. I wont even go there. There is not enough bandwidth. I have a friend who has three hdds in a raid array. It fills up the bandwith entirly and holds it all the time.

[Lynx]

Sorry I meant to say that the bandwith was 133mbps not 264mbps it is even worce

[Hoot]

What's worse that that is a UDMA100 HDD may transfer the data it has retrieved from the magnetic media and moved into buffer ram that fast, but once that buffer empties and more data has to be retrieved from the magnetic media, then data moves in milliseconds, not microsecinds. Now that is a seriously limiting factor. That's why it does not "handcuff" the PCI bus as much as predicted. The same applies to CD Rom drives.

It's kinda (loosely) like reaching down into the mud, scooping up a handfull, forming it into a ball and finally throwing a 100 MPH fastball with it. The speed of the pitch is not the time consuming part of the process.

Hoot

[AtomicGuY]

Limits of the Athlon, hmm, well currently, there is five main types of Athlons out there.

K7 - This one started at 500MHz and ended at 700MHz. (Slot A).

K75 - This one begain at 750MHz and died at 1000MHz (Slot A).

Thunderbird - This one started at 700MHz and is topped at 1.40GHz officially. There is a 99% chance that there will be a 1.53Ghz T-bird since the latest core revision can hit 1.53GHz without voltage change.

Palomino - This one starts the desktop officially at 1.53GHz and should end with roughly 1.93GHz

Theroughbred - Advanced .13 Micron core which should start at 2GHz. The Theroughbred will redefine the name AMD and bring them back into the game once again crushing the northwood from Intel. This advanced Socket A processor will feature the new 333MHz BUS standard as well as smaller die shrink meaning less voltage.

(Remember, the Socket A bus was a long investment and will show its true colors when processors speeds pass the 2Ghz mark and still float about this same architexture)