Ron_Jeremy

Xeons Have Any Future?

Recommended Posts

The job I completed yesterday resulted in me being paid in hardware, rather than cash (call it more of a trade). Anyway, I dunno whether to keep this dual Xeon mobo (Asus PC-DL) or just sell it. It's practically a brand new oem board. I'd keep it if Xeons were soon to adopt a 800MHz FSB, figuring it would make for a decent low cost dually. But if Intel has no plans of moving Xeons to the faster FSB I will probably give it to a friend.

Does anyone know what's in store for socket 604 in the very near future?

Share this post


Link to post
Share on other sites

sell it quick. The amd opterons destroy the xeons in virtually any meaningful server benchmark. Check aceshardware or anandtech for reviews on this.

Share this post


Link to post
Share on other sites

The last that I read, which was a well ago so it may be wrong, was the intel was skipping the 800fsb for xeon and going to the 1066 fsb i believe.

Share this post


Link to post
Share on other sites

HA! I keep forgetting about Asus' little experiment! Too bad the multiplier is locked. :blink:

I had read/heard the same thing about skipping the 800MHz bus.

But if Intel feels pressure from AMD in the workstation / low server market, then they may magically pull 800MHz bus chips from the.....the air. :)

Who expected a P4 with 2MB of L2 cache???

Feel the irony: Intel is now the low-cost alternative to the Opteron. :D

I'm having trouble justifying the huge price premium for Opterons.

The 244's are now dropping down to 3.0GHz Xeon prices.

DogEared

Share this post


Link to post
Share on other sites

The problem with Intel's shared bus architecture is that increasing FSB speed makes designing multi CPU boards more difficult. If it does come out with 1066 FSB Xeons the boards will probably be rather expensive.

A Xeon will be fine for many tasks years to come though. A file/print server doesn't need all that much power. Hell, a Celeron or Duron can often do the job.

Share this post


Link to post
Share on other sites

You get what you pay for. If you pay for a celeron, you will get celeron performance. Would you want celeron performance in a production server? I wouldn't.

Xeon is pretty crappy choice no matter how you look at it. It has no 64-bit capabilities, which just screams "yesterday". It also gets it's ass kicked in 32-bit performance by Opteron. Plus Opteron runs at lower frequencies, which means less aggressive cooling.

And don't tell me "there is no need for 64-bit! there is not 64-bit software!". Already there are versions of Linux which are optimized for the AMD64 platform. And 64-bit does mean better performance, just read the damn article:

http://www.aceshardware.com/read.jsp?id=60000275

Why should we stick to 32-bit when we can have 64-bit? "It's good enough" was never a good excuse in the IT industry, nor is it in this case.

Share this post


Link to post
Share on other sites
Why should we stick to 32-bit when we can have 64-bit? "It's good enough" was never a good excuse in the IT industry, nor is it in this case.

"It's good enough" is even worse than an excuse, it's a reason. Why buy more than what you need now or in the forseeable future? 64-bit is a waste of money for many servers. Even Xeons are. We sell file/print server for networks of up to 30 users. A Pentium II is fine for that. Fast disks, fast network and truckloads of RAM are more important for them. 64-bit isn't the solution to everything. I'd be comfortable selling stock P4's for those kinds of networks.

Share this post


Link to post
Share on other sites

Yeah, I feel guilty buying dual Xeon Servers for 30 (concurrent) users Databases.

But that's what the big boys are selling....

Gotta admire Dell for selling $249 P4-based Servers! :)

As much as I'm looking forward to putting together a dual opteron workstation, I don't see any 64-bit software...unlike when the PentiumPro came out.

When I bought my PPro, I had a 32-bit OS (NT & OS/2) w/ 32-bit software.

Besides when it comes to 64-bit on the desk: "Been there done that."

I always laugh when I hear Apple and AMD arguing over first 64-bit desktop.

I seem to recall having Alphas and UltraSparcs on my desktop...

(Yes, the Alphas had a 64-bit version of NT -- kinda. ;) )

So in reality, AMD doesn't even make the "First 64-bit Windows Processor."

But if NVidia can get away with claiming to be the first GPU....somebody should explain to Texas Instruments that their graphic chips don't count. ;)

(That was kind of a rant wasn't it??? Sorry. )

DogEared

Share this post


Link to post
Share on other sites
But if NVidia can get away with claiming to be the first GPU....somebody should explain to Texas Instruments that their graphic chips don't count.  ;)

What TI chip was that?

As for nVidia claiming to make the first GPU?... umm I would have thought Tseng Labs or Trident (for 2D), and 3dfx (for 3D - well what is considered 3D now).

Other notables before nVida: Matrox Mill, S3 Trio, Tseng Labs ET6000, ATi Mach32/64... and of course 3dfx with the Voodoo1...

What ever happened to the days, where you had tens of video vendors, not just the 5 or so we have now? (ATi, nVida, Intel, SiS, Matrox)...

What ever happened to 3dfx (bought out by NV), Cirrus Logic (IIRC bought out by Intel), Tseng Labs (??), Bitboys (will they ever be a player), Paradise (??), etc... And whats the deal with Trident and S3 at the moment?

Lets see what other video makers we can name???

PS. Sorry for going off topic...

Now to get back on topic. IIRC Intel were going to release 800FSB Xeons, but due to market pressure (ie getting their butts kicked in the low end server market performance wise), may adopt the 1066FSB instead of the 800FSB. The Netbus arch is a high bandwidth architecture, which requires tons of FSB bandwidth... the jump from 533FSB to 800FSB on the same chip gave IIRC 10-15% performance increase, (ie 2.66GHz vs 2.6Ghz (w/HT off)).

If I had that board, I would just get some 533FSB Xeons and be done with it... Plenty fast for most tasks... (unless your trying to simulate a nuclear explosion :P ).

As for Opterons, yes they are good performers (no one should question that), but very few SME (which makes up 90-95% of the market), need 64bit systems, (and if they do, they will be running Sun/SGI/Apple). I'm personally focused on the SME (as AFIACT so is HMTK), and the AMD64 arch doesn't make sense for this market at this time... (esp with the lack of a MS system to take full advantage of it).

Share this post


Link to post
Share on other sites

Some of the architectural changes of the Xeon "nocona" to be released next year are here

Undoubtably, the new changes will provide significant improvement over current xeon's.

I guess it depends what you plan to use that board for...

Share this post


Link to post
Share on other sites
As for nVidia claiming to make the first GPU?... umm I would have thought Tseng Labs or Trident (for 2D), and 3dfx (for 3D - well what is considered 3D now).

Other notables before nVida: Matrox Mill, S3 Trio, Tseng Labs ET6000, ATi Mach32/64... and of course 3dfx with the Voodoo1...

Waitek? Google turns up nothing, but it accompanied the 924 on some old Diamond Stealth Pro VLB Vram cards and was a 2D hardware co-processor.

Frank

Share this post


Link to post
Share on other sites
As for nVidia claiming to make the first GPU?... umm I would have thought Tseng Labs or Trident (for 2D), and 3dfx (for 3D - well what is considered 3D now).

Other notables before nVida: Matrox Mill, S3 Trio, Tseng Labs ET6000, ATi Mach32/64... and of course 3dfx with the Voodoo1...

Waitek? Google turns up nothing, but it accompanied the 924 on some old Diamond Stealth Pro VLB Vram cards and was a 2D hardware co-processor.

Frank

Number Nine predates every suggestion I've heard here so far, as a full-fledged graphics co-processor. With 6 MB of ram I could download the entire X-client subsystem to the graphics card on my SCO Unix system. Woohoo!

- Rick

Share this post


Link to post
Share on other sites
As for nVidia claiming to make the first GPU?... umm I would have thought Tseng Labs or Trident (for 2D), and 3dfx (for 3D - well what is considered 3D now).

Other notables before nVida: Matrox Mill, S3 Trio, Tseng Labs ET6000, ATi Mach32/64... and of course 3dfx with the Voodoo1...

Waitek? Google turns up nothing, but it accompanied the 924 on some old Diamond Stealth Pro VLB Vram cards and was a 2D hardware co-processor.

Frank

Number Nine predates every suggestion I've heard here so far, as a full-fledged graphics co-processor. With 6 MB of ram I could download the entire X-client subsystem to the graphics card on my SCO Unix system. Woohoo!

I guess that just shows how young I really am...

Share this post


Link to post
Share on other sites

I am running a dual Xeon Database Server.

I am very impressed with the chips' ability to multi-task, especially when hyperthreaded. Basically you are getting a quad processor server for the price of a dual.

The ability to execute 4 threads simultaneously is the main buying reason for this Chip. Its not as fast as others when performing a single task that relies only on a single thread using the CPU and memory, but it does allow you to continue doing what you are doing unhindered by the fact that 1 of your 4 processors (Dual hyperthreaded Xeons show up as 4 proc machines in Windows) is at 100%.

I would suggest that you check out newegg. They are selling for the cheapest by far.

Intel Xeon DP 2.4 GHz 533MHz FSB / BX80532KE2400D / N82E16819117015

$231 + 0 shipping

So I guess you just need to examine what kind of use you will have for this box.

Gamming - Dump it

AV Box - Dump it

File Server - No benefit

Development Server - Only if you are doing load testing

Development Workstation - Only if you have no sensitivity to noise. The 2 cpu fans are pretty loud. I had one as a workstation for awhile, and you will love the fact that you never have to wait for 1 background process to finish before you can get Windows to acknowledge that you'd like to still keep typing and doing other things. It really cuts down on the seat time when you can fire off a whole bunch background compiles and unit tests while you work uninterupted on the next problem.

Production Server - Definately the only way to go

Share this post


Link to post
Share on other sites
I am running a dual Xeon Database Server. 

I am very impressed with the chips' ability to multi-task, especially when hyperthreaded.  Basically you are getting a quad processor server for the price of a dual.

Even in the best conditions, HT offers performance improvements of maybe 15%. In many situations, it lowers performance. Having four processors displayed in the Windows Task Manager does not mean that the system has four processors, or their equivalent.

Database serving tends to benefit from HT, but certainly not as much as it would from having two additional real processors.

The ability to execute 4 threads simultaneously is the main buying reason for this Chip.  Its not as fast as others when performing a single task that relies only on a single thread using the CPU and memory,  but it does allow you to continue doing what  you are doing unhindered by the fact that 1 of your 4 processors (Dual hyperthreaded Xeons show up as 4 proc machines in Windows) is at 100%.

Running more than a few tasks should not hinder even a uniprocessor system if its OS can multitask well.

Additionally, PC processors since the original Pentium can technically execute two tasks at once. It is a matter of thread-level parallelism vs. instruction-level parallelism.

As Honold often drills into everyone's heads, fluff like the number of virtual processors your OS thinks it has, how many instructions or threads the processor can theoretically work on at once, the clock speed, the amount of cache, and any other 'selling point' of a processor has no intrinsic value in and of itself. What really matters is the final performance figure, and for DB servers, the Xeon has had enough trouble competing with the vanilla AthlonMP. It isn't even in the same league as the Opteron. Granted, it is still a fine chip, and is part of a very mature and well-tested platform, but let's not fool ourselves by referring to useful but limited tricks like HT.

Anandtech's Xeon vs Opteron review:

Frankly, we were shocked when we saw the first performance results, and we ran and re-ran them to make sure our numbers were correct. In the end, they were.

The Opteron 248 setup managed to outperform Intel’s fastest, largest cache Xeon MP by a whopping 45%.

From the Ace's Hardware Xeon/Opteron review:

Again, a landslide victory for the Opteron.
Production Server - Definately the only way to go
A good way to go, but most assuredly not the only way.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now