Slashdot Mirror
Anand Reviews Athlon 64 FX-53
trickofperspective writes "Anandtech has a review of AMD's latest processor, the Athlon 64 FX-53. Long story short -- the FX-53 is a "very solid processor," but you'd be better off waiting a couple months for Socket 939."
I had a conversation with a neophyte that was looking to "build their own computer" yesterday...He was obsessed with the idea that megahertz=performance...I tried to tell him that an FX-51, 52, or 53 would be a much better performer, all around, than any Pentium 4, "Extreme Gaming Edition" (as he put it) or not...but in the end, he was swayed by things like "Hyperthreading" and "Netburst"...AMD is having a hard time fighting against Megahurtz Madness and Buzzword Bufoonery.
Back in the 70's IBM mainframes were liquid cooled, the problem was, they often sprung leaks...
Mod +5 Drunk
What they probably mean is the chip has the capability to set segments of memory with a do not execute bit, for parts of memory such as the stack. That reduces the amount of things that a hacker can do if he finds a buffer overflow to exploit.
I'm very much against this, if I don't want to be safe, thats my problem. If a doctor says I need anti-biotics for something, I'm not required to take them.
I understand that kind of thinking can hurt other people when my machine is on a network, and in my laziness my machine becomes infected and starts a DoS attack and such. But still, not every machine is on a network, and I don't want to have to pay for the extra security.
Let the processor do the processing.
Josh
If you knew how the protection is done you woudlnt be so critical of it. The core of AMD's Enhanced Virus Protection is the a NX bit which specifies whether a page of memory is executable or not. This way, even if buffer overruns occur in that area of memory, it wouldnt be executed. I am not really sure how this is a bad thing.
My mom never taught me to sign.
The Cray-2 supercomputer was also liquid cooled. They used a fluid called flourinert which is electrically non-conductive but a good thermal conductor. Flourinert was origionally developed as an artificial plasma substitute for heart surgery. It is also insanely expensive, around $500/gallon!
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
Now who is going to have the first kernel which sets it all up properly to be secure? Linux? OpenBSD? FreeBSD? Or will it be that backward little company in Redmond who have major quality and security problems with everything they do?
I can't find anybody anywhere that is saying you won't be able to stick a 939 pin Athlon in an 940 pin socket. Plus, Opteron chips will still use the 940 pin platform. You'll be able to upgrade still if you buy now. The only downside is that you'll have to use Registered ECC memory.
I definitely don't need one right now, but in a year, when it is a mainstream product, I will find some excuse to persuade myself to buy one. By that time the OS (Linux of course) will have been very well debugged.
I wonder when the move to 128-bit will come?
It's THG, but... http://www20.tomshardware.com/cpu/20010917/
BTW, that only applies to non-64-bit Athlons/Durons. Yes, Intel is hotter than AMD. No, Intel processors don't hit 370 C when their heatsinks are removed, as the Pentium III shuts down, and the Pentium 4 slows down. The Athlon MP/XP had thermal overload protection in their spec, but some boards don't include it.
That's just lame.
the growth in cynicism and rebellion has not been without cause
It should be simple to pass the flag into gcc and see if the kernel still works. (It should.)
Unless they do odd things like generate code on the fly, it shouldn't be a problem.
JIT runtimes might have problems, though, if you forced the flag in, say, your Gentoo installation.
tasks(723) drafts(105) languages(484) examples(29106)
Why can't AMD stick to the traditional K6, K7 and keep on adding up. All this Fx and Opteron and 8 million other names are confusing as hell.
Whatever happen to marketing making your purhase decisions easier. It's doing to exact opposite nowadays. Intel ain't doing a whole lot better.
I'm curious if that bit is preserved when pages get moved into swap space. It'd be interesting to study the kernel mechanism for Linux.
tasks(723) drafts(105) languages(484) examples(29106)
Actually, I do nearly the opposite. This summer I'll be purchasing one of the 'obsolete' motherboards with one of these 'obsolete' chips (or the FX51 depending on how thrifty I'm feeling.). Then I get the following:
;-)
- Marked down prices from vendors clearing their old inventory.
- Mature technology without the usual sharp edges associated with cutting edge technology.
- Middle to high end performance.
Of course, then I'm stuck with this cpu for the next few years. That is, of course, until the newest, whizbang socket 1039's hit the market and I pick up a socket 939 mobo priced to move
BTW, Thank you for encouraging everyone else to adopt the newest techology as soon as it hits the streets. _Someone_ has to beta test the hardware after all...
I never upgrade components anymore because it just became too expensive. Anytime I upgraded more than 2 or 3 things I felt compelled to build a machine just to put the old parts in and ended up just spending more and more money. I now have 5 machines at the house and I've supplied machines to most of my family members for cheap. Sure I recovered some of the costs by selling them the old machines but I never recovered all of my expenses and I eventually ran out of family members =)
I wouldn't buy a 940 board for an FX system - 940 will be the thing to get for Opterons, if you want multi-proc, and/or registered memory to get LOTS of memory, for a DB machine or something. It's good to have a choice, as long as both choices are going to be around for awhile. Really, you'll want to target your platform (CPU & mobo choices) to what you want to do with your system.
:(
I guess it's good this socket switch is happening at the same time as the switch to PCI-E, so you can get all this stuff outta the way at once.
Personally, there's no way I'd get a 939 board withOUT PCI-E at the same time - you're just going to have to ditch that soon enough, as well.
Really, I'm more interested in the PPC 970FX & 980, anyway. C'mon - 24.5 Watts for a 2gHz 970FX?! That's pretty amazing. I don't know if the 980 will have an on-die memory controller, but here's hoping.
As for other upcoming CPUs, the Intel Dothan also sounds very promising, as does the possibility of a multi-core version of the Dothan. I just wish Intel would wake up and make the Pentium-Ms for desktop platsforms. I also think the 2Meg L2 of the Dothan is overkill, and makes the chip more expensive than it needs to be. A 1Meg L2 is plenty for Dothan, and would reduce the die size considerably, and thus the cost. Intel's being a little strange on that front, though the Celeron-M is a good move.
I'm still curious as to why Intel and AMD haven't added AltiVec to their x86 processors.
The Quantum Computers I've seen use a gas in a pressurized chamber with lasers to "read" and "write" the quantum states of the gas molecules. (Yes pedants, I realize this is far oversimplified, but I'm making a simple observation here) There is your gas processor, I suppose.
Slashdot is proof that Sturgeon's Law applies to mankind.
In January, I built an Athlon64 FX51 system. I debated the merits of getting the 940 vs. waiting for the 939 and decided that I will have to upgrade the motherboard in a year anyway. Why? PCI Express should be readily available, and market tested by then.
Besides, everytime I have actually upgraded a processor, I have also had to buy new memory and motherboard to accompany it. There's no sense putting a shiny new processor in and having it use old, slow memory on an old slow, bus.
However, just last week, I had to replace the mb in my kids' computer - I got a kt600 board in anticipation of upgrading their Athlon XP 1900+ and PC2100 at some point.
How much do you want to bet that by the time I need to upgrade that processor and memory, I will still need to buy a new mobo in order to utilize the new features of whatever processor or memory architecture I use.
BTW, I LOVE MY FX51 MACHINE! I built the whole system for just $3200 ($1150 was the display). For work stuff, I run Gentoo Linux compiled for AMD64, and for games, regular 32 bit Windows XP Pro. Absolutely no complaints. Wolfenstein Enemy Territory shows 60-90fps at 1600x1200 full detail.
Hope this isn't a RTFAC (C=Carefully), but I don't see in this article if they specify the tests were done with a 64 bit version of the OS and 64 bit binaries for the apps, or just a 32 bit versions.
/think/ it would make a big difference benchmark-wise.. correct me if I am wrong.
Maybe I'm missing it, but in the "OS" category, they only say "Windows XP Profession SP1", no reference that I can see if it's a 64 bit version of the OS or not. Ditto for the apps.
My understanding is (with the current state of affairs, 03/18/2004) if you run to the store and buy an AMD 64 machine, you'll get a 32 bit version of Windows and apps.
But you only get the benefits of eg. >2^32 ram access IF you run a 64 bit OS and 64 bit apps.
Being able to access >2^32 of ram in an app is really useful for 3D rendering of very large projects (a business I happen to associated with), so if the tests are done with 32 bit OS/Apps, the benchmarks would seem to be not so useful for that purpose.
Considering this is a 64 bit processor being evaluated, it would seem lacking not to mention this.
Most folks in 3D evaling AMD 64's are sticking 64 bit os's on there right away (Suse, Gentoo, Fedora/Yarrow, etc), and doing tests with that.
Am I missing the part where they talk about 32 vs. 64 bit OS in these tests? I would
ps. With all those blinking flashing (*!&@# banner adds, it's often hard to RTFAC. I wonder, do schools now give reading comprehension tests in rooms with flashing lights and spinning graphics to simulate 'real world' scenarios? >;)
Where are the 64bit processor benchmarks where the tests have been compiled for them?
I mean, given that the x86 64bit decendants have more registers and all, running some stupid Sysmark or Unreal Tournament on top of them is like comparing V4 and V8 engines in such a way that the V8 only gasoline to four of it's cylinders.
What I want to know is the P4 flag ship lined up with the AMD 64bit flag shit on linux with a kernel compiled for 64bit and apps compiled for 64bits.
I have not been able to locate a single such benchmark as of yet. Anyone? Please...
1 Earth is warming, 2 It's us, 3 it's royally bad, 4 we need to take action NOW
I suspect that these chips exist entirely for marketing purposes & are there really to have the 'fastest desktop cpu on the market title'. But a much more VALUABLE question to ask is 'what's the best chip for $400, $300, $200, $100?, heck even $50!!'.
every time i look at these benchmarks, i have to cut out the EE's & the FX's & just see how the P4's compare to the althon-64's & XP's. because in my mind these 'elite' chips don't exist. They're toys. there's no remotely reasonable reason to buy them.