afraid to get outdated

K

kevinf

Limp Gawd
Joined
Feb 21, 2004
Messages
375
is it a good idea to buy a processor right now, because i have heard about the socket 1027, i am on a budget and i am planning on getting the 3000+ venice, should i buy it now or wait until the new sockets
 
That depends on what processor you have, and what you mostly use it for. Enlighten us.
 
i am about to get the motherboard with the processor, i do gaming, web surfing, and homework on it
 
The next socket will be M2, whcih has 940 pins...the one you are thinking about is for server application (I believe). If you want the most future-proof system you can get, an AMD X2 on a socket 939 is your best bet, especially with the 3800+ very soon...but as far as a single core goes, socket 754's price makes alot more sense.
 
but if i get the skt 939 cpus wouldn't i be able to upgrade to x2's in the future
 
Sure can, but I don't see the reason for buying a single-core CPU now and then getting an X2 later, when you can just get an X2 now, or the X2 3800+ that's coming out in a few weeks and save the difference in money?
 
No matter what you buy, something new is right around the corner. Seriously. They plan (with roadmaps) for new Chipsets, CPUs, etc years in advance

so Just buy the best you can now and dont worry about it.
 
ya that is my point i just don't really have that much money right now so maybe i'll start with this one for now
 
USMC2Hard4U said:
No matter what you buy, something new is right around the corner. Seriously. They plan (with roadmaps) for new Chipsets, CPUs, etc years in advance

so Just buy the best you can now and dont worry about it.
Click to expand...

that makes me suspicious. Why wouldn't they just come out with the most powerful stuff they can right away? I thought I read about a "Hammer" AMD CPU a couple or three years ago that was supposed to be 3.8 GHz. Then they kept on saying they had "manufacturing problems". M'yeah. :rolleyes:
They should just make the fastest available thing now. Like if they know multi cores is what they want to do, why don't they just come out with 16 core right now? Cripes, it would avoid a couple of unnecessary upgrades.
 
forumposter32 said:
that makes me suspicious. Why wouldn't they just come out with the most powerful stuff they can right away? I thought I read about a "Hammer" AMD CPU a couple or three years ago that was supposed to be 3.8 GHz. Then they kept on saying they had "manufacturing problems". M'yeah. :rolleyes:
They should just make the fastest available thing now. Like if they know multi cores is what they want to do, why don't they just come out with 16 core right now? Cripes, it would avoid a couple of unnecessary upgrades.
Click to expand...

that wouldn't be good marketing. the object is to stay in business not release the most powerful chip you have right away. If they did that, not only would it be crazily expensive, but we'd have to wait a few years for them to develop something better. It's not good business sense at all. Besides, no body says you HAVE to upgrade all the time.
 
forumposter32 said:
that makes me suspicious. Why wouldn't they just come out with the most powerful stuff they can right away? I thought I read about a "Hammer" AMD CPU a couple or three years ago that was supposed to be 3.8 GHz. Then they kept on saying they had "manufacturing problems". M'yeah. :rolleyes:
Click to expand...
hamme... hmm, that sounds familiar..

sledgehammer! clawhammer! ;)
they definitly came to market, just not as "fast" as people were rumoring them to be
 
forumposter32 said:
that makes me suspicious. Why wouldn't they just come out with the most powerful stuff they can right away? I thought I read about a "Hammer" AMD CPU a couple or three years ago that was supposed to be 3.8 GHz. Then they kept on saying they had "manufacturing problems". M'yeah. :rolleyes:
They should just make the fastest available thing now. Like if they know multi cores is what they want to do, why don't they just come out with 16 core right now? Cripes, it would avoid a couple of unnecessary upgrades.
Click to expand...

They don't do that because it's really hard. They are themselves on an upgrade cycle --- they have to keep buying the latest manufacturing technology to keep putting more transistors in CPUs. This manufacturing technology moves pretty fast (they have to replace their machines every couple of years), but it moves incrementally. The industry would love to go to a 45nm process right now, but the technology for it is still in development, and the engineers won't be done developing it 'till 2008-2009.

Dual core is pretty much the biggest CPUs they can make right now. An Athlon X2 has a die size of 200 square millimeters, and anything much over 250 and your price per core runs into Itanium (~$2000+) territory. However, quad core will be entirely possible when they move to the 65nm process, which will halve the size of the die and allow you to put four cores in the same space as the current X2s. So in order to get ready to use the technologies the manufacturing engineers will be done with in the next few years, the CPU engineers plan ahead. They did the same with the X2. It was planned in early 2000 when Hammer was being designed. But back then, they were on 130nm and there was no way they could fit two cores into a reasonably-sized die. However, they knew 90nm would be coming out eventually, so they designed the Athlon64 with that in mind.
 
A sempron @ 2GHz for socket 754 would be very cheap and perform/OC great, but not future proof with upgrades. A socket 939 3000+, on the other hand, would be a bit more expensive OC well (but less stock performance actually), but would be upgradeable to X2 later on to blow your pants off (or just get X2 now and save 147 altogether. Your choice really.. blow a bit more money now and save big later, or save a little now and blow more money later.
 
rhashem said:
They don't do that because it's really hard. They are themselves on an upgrade cycle --- they have to keep buying the latest manufacturing technology to keep putting more transistors in CPUs. This manufacturing technology moves pretty fast (they have to replace their machines every couple of years), but it moves incrementally. The industry would love to go to a 45nm process right now, but the technology for it is still in development, and the engineers won't be done developing it 'till 2008-2009.

Dual core is pretty much the biggest CPUs they can make right now. An Athlon X2 has a die size of 200 square millimeters, and anything much over 250 and your price per core runs into Itanium (~$2000+) territory. However, quad core will be entirely possible when they move to the 65nm process, which will halve the size of the die and allow you to put four cores in the same space as the current X2s. So in order to get ready to use the technologies the manufacturing engineers will be done with in the next few years, the CPU engineers plan ahead. They did the same with the X2. It was planned in early 2000 when Hammer was being designed. But back then, they were on 130nm and there was no way they could fit two cores into a reasonably-sized die. However, they knew 90nm would be coming out eventually, so they designed the Athlon64 with that in mind.
Click to expand...

And Intel, the big dog didn't.. How ironic.
 
Intel designed the P4 back in the 180 nm age, not 130 nm. It's a generation older than the Athlon64. Intel's answer to the Athlon64 is still in development. It will come eventually, don't worry.
 
indeed..

though, i'd reword it to say that they designed netburst back then. pescott is amazingly different than the willamette of old. one would be hard-pressed to call the two part of the same 'architecture' :eek:
 
(cf)Eclipse said:
indeed..

though, i'd reword it to say that they designed netburst back then. pescott is amazingly different than the willamette of old. one would be hard-pressed to call the two part of the same 'architecture' :eek:
Click to expand...
Well they are the same in the aspect that they both suck...

Northwood / Gallatin was the best version that their was. It was Outstanding for the Netburst line but that still isnt saying much.

For a while their was potential with it. You could give me a 4Ghz, 65nm, Gallatin (20 Stages, 512K L2 Cache, 2MB L3 Cache) and if you can add NX Bit, SSE3, 64 Bit, LGA775 and support on the 955X.... I would take it in a heart beat.
 
Thanks for the explanations.

But, does anyone know if Intel succeeded in including that thing AMD talked about in their lawsuit in their chips? I mean, they said it was some kind of code that was supposed to slow down AMD chips but not Intels.
I don't want to see that kind of code being used against any maker, even if we start seeing more competitors in the market.
 
It's more that Intel didn't write their compiler to optimize for AMD CPUs, whereas they knew their own processors quite well and optimized for them.
 
forumposter32 said:
But, does anyone know if Intel succeeded in including that thing AMD talked about in their lawsuit in their chips? I mean, they said it was some kind of code that was supposed to slow down AMD chips but not Intels.
Click to expand...
the compiler.. what makes the programs that run. it can be optimized in different ways.. this part doesn't really have much to do with the cpu itself ;)

USMC.. definitly. i'll admit that i used to be a fan of early northwoods.. i thought intel had finally made good on netburst. the problem was that a similar athlonXP at the time was ~3/4ths to 1/2 the price, hence why i went amd. now with dual core, the price swings back into intel's favor, but performance and heat.. :eww:
 
(cf)Eclipse said:
though, i'd reword it to say that they designed netburst back then. pescott is amazingly different than the willamette of old. one would be hard-pressed to call the two part of the same 'architecture' :eek:
Click to expand...

Well, I consider Netburst the architecture (the general structure and layout of the execution core). Northwood and Prescott are just updates of the original architecture.
They all process instructions in pretty much the same way.
Pentium-M is clearly a different architecture. Instructions travel a completely different path than in any of the P4's.
And Conroe will again be very different.
 
(cf)Eclipse said:
now with dual core, the price swings back into intel's favor, but performance and heat.. :eww:
Click to expand...

The intel DC and AMD DC are completely different products. It's like comparing an Athlon MP 1.4ghz SMP to a 3.6ghz Xeon SMP. I can't believe how much the intel guys claim the dual core price crown, as if the tables have turned. These aren't products even remotely related to each other. Intel is selling a chip with a 18month old clock speed (and the same architecture, just 2 cores), against a chip with a 3month old clock speed (and a whole new architecture mapped out for 2 cores). AMDs solution is geared towards better and cheaper SMP (and some cash for them). Intel's DC is pure marketing, just like netburst. In fact, I'd argue that what current Intel Dual Cores do for the industry is equivalent to what netburst did for the industry, put out a half rate performing product that looks good to a fool with a loose pocket book, and makes them some money. A+ for marketing (2 is better than 1 right!), but fortunately all (or some) of us here are above that. That's why were here, is to discuss stuff as peers, and purchase based on that, not on marketing.

Oh no :eek:

My single 2.9ghz core is so lonely :(

I'd better upgrade :rolleyes:
 
mikelz85 said:
In fact, I'd argue that what current Intel Dual Cores do for the industry is equivalent to what netburst did for the industry, put out a half rate performing product that looks good to a fool with a loose pocket book, and makes them some money.
Click to expand...
agreed. there's no good reason for the cpu communication through the northbridge, esp since they tecnically need a new chipset anyway :(

quite a hefty update to the architecture ;)
 
but why a new chipset? I like the current crappy one that makes me buy a new board, new ram, etc! I can't see why anyone would want a simple drop in upgrade that uses the same board, same ram, and even most likely, the same HSF. Yuck :mad:
 
serbiaNem said:
And Intel, the big dog didn't.. How ironic.
Click to expand...

When Intel designed the Netburst, they didn't count on hitting the frequency wall they did. The current processes aren't ramping up as quickly as they expected, and leakage currents (creating waste heat) are much higher than they expected. Intel bet on single-threaded performance, and they bet on manufacturing technology keeping up as they did in the late 1990s. The bet just didn't pan out, is all.
 
LstOfTheBrunnenG said:
It's more that Intel didn't write their compiler to optimize for AMD CPUs, whereas they knew their own processors quite well and optimized for them.
Click to expand...

The Athlon is really insensitive to instruction scheduling (unlike the P4). It's a practical decision for AMD --- there are way more P4s than Athlons, so code will be optimized for the P4. The K7 architecture needed to run non-K7 optimized code well. If ICC generates significantly slower code on an Athlon than a P4, there is a very real possibility that its a deliberate pessimization, and not just a result of Intel knowing their own processors better.
 
Not entirely... AMD cut some corners in the design, so certain instructions cannot be reordered. Loads and stores mainly.
All Intel processors since the Pentium Pro are able to reschedule loads and stores, so it could be that the compiler doesn't actually bother to schedule those, since they will be sorted out on the CPU anyway. Athlons will have to run that code in-order, and that may hurt performance.
 
Scali said:
Not entirely... AMD cut some corners in the design, so certain instructions cannot be reordered. Loads and stores mainly.
All Intel processors since the Pentium Pro are able to reschedule loads and stores, so it could be that the compiler doesn't actually bother to schedule those, since they will be sorted out on the CPU anyway. Athlons will have to run that code in-order, and that may hurt performance.
Click to expand...

but yet when ppl trick the ICC in to thinking theres a intel CPU the code runs faster on the AMD CPU any way so hows that work?
 
Do you have any proof of that, something that can be reproduced?
Then we can perhaps analyze what is going on and why.

I have taken a program of mine that I compiled for P3 with ICC. Then I replaced all occurences of "GenuineIntel" with "AuthenticAMD". This should be enough to fool the program into turning on the optimizations on an AMD CPU instead of on an Intel.
There was no increase in speed on my Athlon XP, nor decrease.
The same program actually does run faster when compiled with VS.NET, but apparently it is not because it was deliberately slowed down on my Athlon. I don't have an actual P3 to test it on, so I don't know if that one would be faster with the Intel-compiled version, or if the modified AMD version would be slower on it.
When I compile for P4 and run it on a Northwood, it does go quite a bit faster than VS.NET. But I cannot run that version on my Athlon XP, and I don't have an Athlon64 to test it on.
 
tsuehpsyde said:
The next socket will be M2, whcih has 940 pins...the one you are thinking about is for server application (I believe). If you want the most future-proof system you can get, an AMD X2 on a socket 939 is your best bet, especially with the 3800+ very soon...but as far as a single core goes, socket 754's price makes alot more sense.
Click to expand...

S939 isn't that much more. On the low end anyway. On the high end you shouldn't be doing Socket 754 anyway.

If buying a motherboard today, you should go with Socket 939 and PCI-Express. Buying into the old doesn't make sense. Even if you save a small amount of money buy doing it.
 
Hey Scali, in VS. NET I think their optimizations have more to do with the actual code (such as dead-store elimination or function inclining) rather than the actual hardware you are running it on (such as static scheduling).
 
Both compilers do both, although I expect Intel to do better at the instruction scheduling.
Apparently VS.NET is so good in the overall optimizations, that instruction scheduling isn't that important.
A few years back, Intel would win all the time, but MS seems to have improved a lot.
The funny thing is that the Intel compiler takes much longer to compile, so you expect it to do far more optimizations and all.
 
Sir-Fragalot said:
S939 isn't that much more. On the low end anyway. On the high end you shouldn't be doing Socket 754 anyway.
Click to expand...
the 4000+ is the same clock and cache as the 3700+... imo, that is a good deal, considering how much dual channel helps, esp after an overclock :p
one can't argue against the dual core argument for s939 though
 
I remember with quake 3 it would not run sse support on amd chips and people made a patch to force amds to look like intel chips and the performance was increased.
 
Scali said:
Do you have any proof of that, something that can be reproduced?
Then we can perhaps analyze what is going on and why.

I have taken a program of mine that I compiled for P3 with ICC. Then I replaced all occurences of "GenuineIntel" with "AuthenticAMD". This should be enough to fool the program into turning on the optimizations on an AMD CPU instead of on an Intel.
There was no increase in speed on my Athlon XP, nor decrease.
The same program actually does run faster when compiled with VS.NET, but apparently it is not because it was deliberately slowed down on my Athlon. I don't have an actual P3 to test it on, so I don't know if that one would be faster with the Intel-compiled version, or if the modified AMD version would be slower on it.
When I compile for P4 and run it on a Northwood, it does go quite a bit faster than VS.NET. But I cannot run that version on my Athlon XP, and I don't have an Athlon64 to test it on.
Click to expand...

ill have to dig up the post on slashdot but the guy tricked it in to thinking it was a p4 and there was a noteble increas in speed when run on the A64
 
Simple fact nothing is future proof except GOD, so just get the best you can.
 
You must log in or register to reply here.
Back
Top