AthlonXP 3400+ VS Athlon64 3400+

[OC-Master]

Now that we know AMD has 400MHz FSB Bartons ready to roll with a 3200+ launch this month, this definately shows us they can easily do a 3400+ Barton.

Now,, the Athlon64 will debute with a 3400+ speed on the chart (2.00GHz).. How do you people figure the next generation will fair compared to the previous generations?

I fore one wonder if the Athlon64's PR rating is based upon the now dated T-bird or on the P4's clock frequency.

People are telling me that the AthlonXP should be faster in 32-bit computing still,, does this sound right??


OC-Master

 

[XJ]

nice avatar...

 

[treepop]

my experience is VERY LIMITED with 64bit processors....back in the day me and my friend got together and he purchased an ALPHA processor....the thing was there where no 64bit apps..I take that back...few 64 bit apps..and linux was was the only os we knew of that was 64 bit capable.....to make a long story short("TO LATE!"clue)we ended up running windows emulated...which created a signifigent loss of performance.....considering that this chip was 500mHz with a 350(I think..maybe 333)(was long ago)this thing was a beast especially at a time when the 350 was the big man on campus...and a voodoo2 with 12 megs of video ram PWNED....so in the end I will only go 64 when the market becomes a 64bit market...

 

[dropadrop]

I think they will both have areas where they will be better, but with a bit of overclocking the athlon64 should be on top. We have to wait and see how conservative amd are with their pr-markings, as xp3000 is really on the limit of living up...

 

[Mark Larson]

Arstechnica's Explanation of x86-64

In AMD's technical documentation it says tha

 

[emericanchaos]

People are telling me that the AthlonXP should be faster in 32-bit computing still,, does this sound right??

yes. in a different thread on here it was being debated. basically athlon64's are estimated 40% faster than xp's while running all 64bit. so any xp running at a higher clock than the athlon64 and running 32bit software will be faster.

 

[Pla]

64bit processing is not for speed.

It is to be able to handle more the 4GB of addressing space.
Even if there are 64bit apps they will be no better if they don't use more then 4GB of memory. The only exception is for scientific computing where you need as more as possible bits per int/float.
Servers would also need as much memory as possible.

May be in 5 year even games may need 16GB of mem... who knows, until then I will wait for some apps before jumping in the boat

Just my thoughts...

 

[Captain Hilts]

I'm sure that if there is a 3400+ Barton and a 3400+ Hammer, then AMD will make sure that they are clocked so that they perform about the same. If they didn't, that would pretty much defeat the purpose of the rating system.

 

[OC Detective]

I'm sure that if there is a 3400+ Barton and a 3400+ Hammer, then AMD will make sure that they are clocked so that they perform about the same. If they didn't, that would pretty much defeat the purpose of the rating system.

Exactly my thoughts as well. Where the Hammer MIGHT be a better bet is due to the greater headroom it may have over the Barton when it comes to overclocking. If there is a Barton XP3400 then it may well be near the limits of capability (similar to the XP2800 Tbred at present).

 

[OC-Master]

Exactly my thoughts as well. Where the Hammer MIGHT be a better bet is due to the greater headroom it may have over the Barton when it comes to overclocking. If there is a Barton XP3400 then it may well be near the limits of capability (similar to the XP2800 Tbred at present).

How do you think SOI technology will effect the overclocking ability of the hammer series?


OC-Master

 

[ninthebin]

dont know, I am yet to see a review that may better clarify how "pushed" they are. But from the roadmap fairly far.

 

[Severian]

In my opinion I expect clock-for-clock that the A64 will still beat a Barton hands down on 32-bit apps, simply because of the on-die memory controller. This reduces the latency from something like 135 ns to more like 80 ns (This is from memory, take it with a pinch of salt), which should make a huge difference. And yes, the SOI should also help overclocking a lot. The downside is now CPU speed and memory speed will have locked multipliers, so my guess would be that memory will hold people back a lot more than CPU speed.

A 2 GHz A64 with 400 MHz memory means that in order to get 2.4 GHz (ie, a decent overclock), the memory would be 480 MHz, which is slighly higher than PC3700 is rated. So, all summed up, I don't think that we'll be seeing A64 overclocks of over 2.5 GHz, regardless of even if you are using liquid nitrogen to cool your processor, because that won't be the bottleneck.

 

[Pla]

Very good point Severian!

 

[RangerJoe]

i would think AMD would try to go HIGHER in clock speed, even though the new cpu's are faster, lots of people buy just because of the actual speed, not only because of the PR rating.

 

[CrashOveride]

yes but it could very well be that the future of Computers speedwise is not in Mhz but rather efficiancy per mhz.... i am starting to think that we are going to reach a mhz limit reletivly soon and need to start being more efficient instead of just pumping mhz... but i could be wrong

 

[RangerJoe]

ya i know that, but alot of people enjoy looking at a 12.7 GHZ OF RAW PROCESSING POWER!!! WOOHOOO!!! but you know what i mean.

the point is, the computer illiterate people, that buy all the top of the line items, enjoy big numbers. thats how it ends. because the people that dont use benchmark programs, but use their computer to type papers, use AOL, or play counterstrike, will not be able to tell the difference between a 3.06ghz chip, and a 2.4ghz chip.

 

[CrashOveride]

good point RangerJoe, but man.. how unfortuant... i hate people! lol ok not really but i really dont think that the way the market is driving it is really the best way to go... meh

 

[Krowa 02]

When you start talking about having to overclock a chip in order for it to perform better, your basically saying its not good. Overclocking the A64 is a whole different story anyway, since it is a totally different design than the processors we have today, no northbridge. Also it wouldnt be worth the money, because you could overclock a different chip and get better results.

Bottom line is, I dont think hammer is going to live up to is expectations. 64 Bit is being rushed out the door, no need for it yet either. AMD is just bascially trying to get their own thing out there, since most of their ideas have been similar to intels, now they want intel to pay for a license to them. I hope that x86-64 isnt widely accepted, I would much rather see a new instruction set developed from the ground up for 64 bit, as AMD should have but didnt have the balls since nobody supported them.

What happens if A64 and Opteron work well with 64 bit and make it a wise choice? Its going to make it a little easier on AMD, but still alot harder to get people to switch over. Intel could just come out with a 64 bit chip and the average user will just get an intel chip.

Also from what I have heard, A64 is a neutered opteron. I would like AMD to stay around for a while though as competition is good for the consumer. However I dont think with the introduction of A64 and Opteron will bring better times for AMD. If AMD does go under, I expect IBM or a similar company to snatch them up, and hopefully they will combine technologies and make something worth while.

Thats it, thats my $.02

 

[DaddyB]

I think the opposite is true, it is becoming more difficult to make a chip more 'efficient' so they will have to ramp up clock speeds to stay competitive ('they' not refering to any individual CPU maker).

The Opterons have some weird 'PR" rating that will be similar to 842 and each number stands for something...what they stand for is not so clear, the first digit denotes how many CPUs can be used together (ie 244 means it can be used in a dual cpu system, 844 would mean it can be used in a 8 cpu system. The last 2 numbers represent performance in some way, the higher the number the better. I'm just saying that you won't be able to directly compare the PR of opteron to the A64 (which AFAIK will still follow the 3200+ type rating), I suppose you will have to compare clock speeds.

From the benchmarks on xbitlabs.com the A64 is only 5% faster at equal clock speed (running 32bit apps) but the memory bandwidth is 20% higher, I hope they can ramp up the clock speeds on the A64 or it is already in trouble.

Using the A64 for 64bit proggies/apps (on a 64bit OS) WILL definetly speed things up, making it 64bit instead of 32 bit doesn't just increase the amount of RAM the CPU can adress it also increases speed since it doubles the amount of data the CPU can address at any one time (ie instead of only being able to handle operations/data 32bits long it can now handle ones that are 64bit long). Since the number of bits is doubled it is theoretically possible to see a 100% speed increase BUT THAT WILL NOT HAPPEN, you will likely see a 20% increase in speed but 40%-50% is a best case scenario.

For an already proven example of this (you could look at the jump from 16bit to 32 bit CPUs which will also demonstrate exactly what I am saying) look at RDRAM Vs DDR RAM. RDRAM uses a 16 bit bus but it is dual channel, so if we look at the bandwidth of PC800 it is:

400mhz x 2 (1 read/write leading and 1 read/write falling edge of clock cycle) x 16 bit bus /8 (bits per byte)=1600MB/s x 2 (dual channel) = 3200MB/s (3.2GB/s)

Whereas DDR uses a 64 bit bus, so with dual channel DDR400:

200mhz x 2 (1 read/write leading and 1 read/write falling edge of clock cycle) x 64bit bus /8 (bits per byte)=3200MB/s x 2 (dual channel) = 6400MB/s (6.4GB/s)

So you can see that doubling the number of bits for the bus theoretically doubles the performance but that does not mean it will double the performance, for example dual channel PC800 has a theoretical bandwidth of 3.2GB/s but in practicle use you will only see just about 2.5GB/s and with dual channel DDR400 you will get around 5GB/s of the theoretical 6.4GB/s, so the fact that the number of bits in the bus is multiplied by 4 (in the case of RDRAM Vs DDR RAM) does mean dual channel DDR has double the bandwidth even though the clock speed of RDRAM is twice as fast (400 compared to 200mhz).

With CPUs there is more overhead and more factors to take into account then when dealing with RAM bandwidth, so you will not see a doubling of the speed but 64bit will definetly be faster then 32 bit in most cases (depending on the overhead, optimized 64 bit code etc.).

Edited for better accuracy and a few mistakes.

 

[CrashOveride]

i do not think that 64bit is being pushed out too early... im sure there will be plenty of servers that will be made of them... it allows for more than 4GB of memory wich is a big plus for large servers... i really thing that them releasing the Athlon64 this soon is going to help them out a lot... though i do think that, at least for a while (until more software for 64 bit stuff comes out) they will need to continue succesful productioin in 32 bit procs for the everyday/enthusiast user.

when the windows 64bit OS comes out i think that having good 64bit processors will be essential