Sign in with
Sign up | Sign in
Your question

dual article - comments?

Tags:
Last response: in CPUs
Share
March 14, 2002 12:31:20 AM

I'll take off with Fatburger's idea, it was surely a good one.

After reading the article first thing that comes to mind is..."Hyperthreading? wth" seems like an impressive flop. But seriously I can't wait to see how optimized programs respond to it, seems like theres lots of potential there. That ram solution was weird too, but I remember seeing stuff like that in older motherboards. Its also interesting to see that the AMD solution isn't to far behind considering the clock difference. I wish they would have done some benchmarks with an MP 2000+ and mp 1200 MHz, only mentioned that it worked.
Oh well nothing else comes to mind and I have to work on taxes, talk amongst yourselves :smile:

"Why can't I be the man? I mean, I DO have harmony balls..." -epoth

More about : dual article comments

March 14, 2002 1:21:54 AM

my point of view on dual, ...

it's pretty much worthless with PC apps...

its good for servers... when there is a lot of workl to be done simultaneouly.
March 14, 2002 1:31:56 AM

I wish the smuck would of waited on 13 mircon chips. Would that made the test better.
Related resources
March 14, 2002 1:35:15 AM

Why is that? Does AMD have a thoroughbred athlon MP on the roadmap?

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 14, 2002 1:40:56 AM

Why yes. Yes they do!

;) 

Mark-

When all else fails, throw your computer out the window!!!
March 14, 2002 1:45:09 AM

Atol why would waiting have made the test better? Please give your insight as to why you think it would have been better to wait.

I personally think the test is fine as is. It illustrates the strengths of both processors; both processors having plenty of strengths.

Mark-

When all else fails, throw your computer out the window!!!
March 14, 2002 2:02:22 AM

15 days for the new Amd Chip. They could not wait?
March 14, 2002 2:10:49 AM

Atol.. shipping to OEMs and chipmakers isn't the same as actual release. AMD likely won't release the chip for review or sale for several weeks after initial shipping begins. So.. sometime in April or maybe May for .13 T-breds.

Add to that the MP version may not ship for several more weeks. AMD didn't indicate which chips would ship first...XP, MP or Mobile.

Mark-

When all else fails, throw your computer out the window!!!
March 14, 2002 3:19:49 AM

It is also good for worstations where you have a lot to do at once also. Microsoft still isn't very good a writing apps that truly multitask on a single cpu.



-* <font color=red> !! S O L D !! </font color=red> *-
To the gentleman in the pink Tutu
March 14, 2002 3:38:20 AM

Bah! Another completely pointless review. Let me ask this who is going to buy either of these boards only to install a 40 GB 5T040H4 Maxtor IDE drive? I can't say for sure about Sandra but most office application benchmarks are highly dependent on i/o pefoprmance. Come on guys those 64bit/66mhz slots are there for a reason! Can't ANYBODY do some testing with a nice 64bit Highend raid card and lets see what these babies can do? Same ol same ol we get the lecture on how for most apps that a "dual" setup is not neccesary yadda yadda. Can some one tell me where I can buy a single CPU board with a 64 bit slot? Did they ever stop to think this might be one of the reasons people may consider this option? OK, yes, this is a rant but I must have been through 12 different reviews so far and nobody has even tried out the 64 bit slots, heck for all we no they may not even work!


It's not what they tell you, its what they don't tell you!
March 14, 2002 4:02:33 AM

a few points

1. it seemed that the dual xeons won the greater number of benchies due to its mem bandwidth & clock speed advantage. yay.
but toms article mentioned the price performance ratio... but gave no prices!

what IS the cost of 2 Xeons, that wierd mobo + ramcard and the ram verus the MP setup?

*heads over to pricewatch.com*

2 x MP2000+ = 2 x $314 = $628
also u could use dual XP's as well
2 x XP2000+ = 2 x $214 = $428
cant fins any p4 xeons advertised... hmmmm
does that tell a point?

2. was that a typo where they gave the fsb of the dual mp chipset as 100?

3. doesnt the xeon have L3 cache as well? kinda puts it in a different category entirely

I love helping people in Toms Forums... It reinforces my intellectual superiority! :smile:
March 14, 2002 11:18:45 AM

About the HyperThreading thing. Tom sais it does not do al lot to performance and that it even slows things down. Anybody see the benchmarks showing us the difference? I did. Only one, the Sisoft Sandra-stuff. The most theoretical benchmark in the bunch. I would like to have seen some real-life benchamrking-results to show the difference in performance between a HyperThreading-disabled and an enabled setup. Both in Single and Double-proc setup (or double and quadruple, as you like). Maybe anyone already found something like that?

Greetz,
Bikeman

<i>Then again, that's just my opinion</i>
Anonymous
a b à CPUs
March 14, 2002 11:29:51 AM

Again, Tom's hardware forgets the perhaps most interesting use of dual cpu's: software development. And the main reason for generally abysmal concurrent/parallel performance is Windows. An OS like FreeBSD takes advantage of dual cups much better. Naturally the software must be written in a scalable fashion! It is a completely different science writing concurrent software compared to sequential. The old catchphrase "multitasking" is only the tip of an iceberg. I can't see a reason for anyone using windows for typical home-use to presently invest in a multicpu system. Myself, though, I can really use another cpu, since I'm a BSD user ...

hack mode: A Zen-like state of total focus on The Problem that may be achieved when one is hacking.
March 14, 2002 11:59:38 AM

Current Xeon P4s do not have L3 cache, new cores are being released (1000s $) with 512 and 1Mb cache options, and SMT.

P4 Xeons are available, and listed I believe.

Board is normal, and is the expanded i860 to accept 4GB RDRAM. Board cost $665 upwards. 2.2A Xeon start at $700 each.

Can be had as a dual bundle for just $2049...

-* <font color=red> !! S O L D !! </font color=red> *-
To the gentleman in the pink Tutu
March 14, 2002 12:43:51 PM

Thanks Peteb, that's what I was looking for.

The xeon's are outrageously priced. Comparing them with an athlon MP is a joke. It's worse than comparing a 2.2Ghz P4 with a 66mhz FSB celeron.

These two cpu's can't be compared without pricing involved.
March 14, 2002 1:04:52 PM

that and 3dsmax i believe takes advantage of dual processor. But you're right. Other than it sounding cool i don't see the benefits of it. I would much prefere spending the money on a decent videocard or faster cpu or just strait up save the money.

If anything it will just make your computer a bit more unstable unless you have very good cooling. My opinion it would double the likelyhood of it crashing do to overheating.

<A HREF="http://www.anandtech.com/mysystemrig.html?id=9933" target="_new"> My Rig </A>
March 14, 2002 1:52:11 PM

My observations:

1) Ncogneto, you have a point. The use of an IDE hard drive itself was very odd. To make it one with with only a 2MB cache was even stranger. Anyone who is going to spend that kind of money on a system is no doubt concerned with one thing and one thing only: performance. The review should have used a SCSI hard drive that would blow away anything that IDE could ever hope to be.

2) Again, THG fails to mention the density of the RIMMS used.

3) Why does one of the graphics have the subtitle "The well-known Intel Southbridge 82801BA.", yet the picture of the chip reads as 82806AA?

4) Did anyone else notice that in three of the benchmark results (Sandra02/Mem, SM02/Ov, and SM02/OP) the dual MP 2000+ platform was actually beat out by the single MP 1900+ platform? Funky. It looks like AMD has a few quirks in their dual platform's operation to work out.

5) Is it me, or does the statement, "With its Athlon MP 2000+, AMD has added a high-performance processor to its portfolio. And what's more - it certainly holds its own against the Intel Xeon 2200.", seem <b>very</b> biased towards AMD, considering the benchmark results?

<font color=red>MP3: Dual Xeon2200 is 134.74% faster than Dual MP2000+</font color=red>
<font color=red>MP4: Dual Xeon2200 is 110.53% faster than Dual MP2000+</font color=red>
<font color=red>LW7b: Dual Xeon2200 is 146.04% faster than Dual MP2000+</font color=red>
<font color=green>C4D: Dual MP2000+ is 101.16% faster than Dual Xeon2200</font color=green>
<font color=red>3DSM: Dual Xeon2200 is 100.72% faster than Dual MP2000+</font color=red>
<font color=red>SM02/Ov: Dual Xeon2200 is 130.77% faster than Dual MP2000+</font color=red>
<font color=red>SM02/ICC: Dual Xeon2200 is 149.15% faster than Dual MP2000+</font color=red>
<font color=red>SMo2/OP: Dual Xeon2200 is 114.18% faster than Dual MP2000+</font color=red>

Out of the 9 application benchmarks, the Athlon only won 2 of them. It barely won Cinema4D by a whole 101.16%. And it 'won' Pinnacle Studio 7 only because the software for some reason (read software bug) couldn't even run using a Xeon processor. So we have no idea what the performance difference is according to the Pinnacle Studio 7 benchmark.

That leaves the Xeon having won the other <b>7</b> benchmarks. At one point the Dual Xeon 2200 beat out the Athlon MP 2000+ by a whopping <b>149.15%</b>.

And while I have excluded the SiSoft Sandra benchmarks results for being theoretical only, we still saw the Dual Xeon2200 best the Dual Athlon MP 2000+ in all three Sandra benchmark results.

Overall, across the 8 usable benchmarks, the Dual Xeon2200 performed on average 123.12% better than the Dual Athlon MP 2000+.

So would I saw that the Dual Athlon MP 2000+ "certainly holds its own" against the Dual Xeon 2200? Certainly not!

Now, before anyone turns my comments into an Intel vs. AMD debate, let me point out that the PR rating of the dual AMD platform <i>was</i> 200 points lower than the Xeon's MHz. So given even that, we <i>shouldn't</i> be expecting to see the dual Athlon MP platform performing equally with the dual Xeon platform anyway. After all, the AMD platform is two PR rating steps behind the Intel platform. I am <b>not</b> saying that Intel is better than AMD. I am merely saying that the conclusion drawn by <i>Frank Völkel</i> and <i>Bert Töpelt</i> is completely baseless and in fact contradictory to the benchmark results of the review they just ran. So I figure either they wrote the conclusion before they ran the benchmarks, or else they're extrordinarily biased.

<pre>If you let others think for you, you're the
only one to blame when things go wrong.</pre><p>
March 14, 2002 2:03:32 PM

Quote:
These two cpu's can't be compared without pricing involved.

While normally I would agree with you if we were talking about desktop processors, I would like to point out that in this case we <i>are</i> talking about high-end workstation platforms, <i>not</i> your average desktop.

The far vast majority of people looking to run a high-end workstation are corporations who care more about pure performance than they do about any price tag. We are talking about the kind of case where spending an extra grand or two just for the absolute latest-and-greatest to give a whole 5% to 10% peformance boost is an every day occurance. So because of this, I think price has little to no meaning in the comparison of a dual Athlon MP 2000+ platform to a dual Xeon 2200 platform.

In a somewhat related note: Has anyone read the <A HREF="http://www.theregus.com/content/3/24313.html" target="_new"> Why isn't SMT Xeon scaling?</A> article? This is the first time that I ever actually thought an article posted on <i>The Register</i> was written by a blithering idiot.

<pre>If you let others think for you, you're the
only one to blame when things go wrong.</pre><p>
March 14, 2002 2:41:11 PM

Thanks for posting this, Oni.

Quote:
"If you give me a VOB file, I'll thrash you to the ground!" - the messages sometimes get pretty hot at the Tom's Hardware Community, emotional reactions included.


Did anybody else see this at the beginning? Time to search, methinks.

Hahahahaha
THG gave us their WinXP serial number :lol: 

The mis-captioned picture of the Southbridge is because they accidentally re-used the picture of the memory controller. I just emailed Fredi about it.

Benchmarks...blah, blah, blah.

<font color=orange>Quarter</font color=orange> <font color=blue>Pounder</font color=blue> <font color=orange>Inside</font color=orange>
Don't step in the sarcasm!
March 14, 2002 4:31:02 PM

And again I say: Why doesn't THG hire an editor to find these things <i>before</i> they release the article?

<pre>If you let others think for you, you're the
only one to blame when things go wrong.</pre><p>
March 14, 2002 5:16:34 PM

The Xeon MP chips (n-way) will only have a 256k L2, but will include a 512k or 1024k L3 cache.

- JW

[edit]
FYI: The current Xeons are really "Xeon DP" (dual processor).
[/edit]

<P ID="edit"><FONT SIZE=-1><EM>Edited by JCLW on 03/14/02 02:25 PM.</EM></FONT></P>
March 14, 2002 5:30:59 PM

<A HREF="http://www.infoworld.com/articles/pl/xml/02/02/25/02022..." target="_new">Here's</A> a real hyperthreading benchmark running on some real hardware.

Results?
"The performance advantages of hyperthreading are undeniable. Our tests of a hyperthreading-enabled Intel Xeon DP server showed, on average, a 45.71 percent increase in SQL transaction performance and a 31.13 percent increase in three-tier Web application performance, versus the same system with hyperthreading disabled."

- JW
March 14, 2002 6:30:42 PM

Good Analysis.

This always happens, I don't know if they are trying to start argument on their forum or they are just bias.

KG
March 14, 2002 6:45:31 PM

I vote for the latter, but so long as THG keeps posting their benchmark results along with their bias, I won't worry over it.

<pre>If you let others think for you, you're the
only one to blame when things go wrong.</pre><p>
March 14, 2002 8:36:35 PM

"
MP3: Dual Xeon2200 is 134.74% faster than Dual MP2000+
MP4: Dual Xeon2200 is 110.53% faster than Dual MP2000+
LW7b: Dual Xeon2200 is 146.04% faster than Dual MP2000+
C4D: Dual MP2000+ is 101.16% faster than Dual Xeon2200
3DSM: Dual Xeon2200 is 100.72% faster than Dual MP2000+
SM02/Ov: Dual Xeon2200 is 130.77% faster than Dual MP2000+
SM02/ICC: Dual Xeon2200 is 149.15% faster than Dual MP2000+
SMo2/OP: Dual Xeon2200 is 114.18% faster than Dual MP2000+"

For one thing, they were comparing an Athlon MP2000 against a XEON 2.2ghz. Yes, I realize that AMD bases their XP PR rating relative to the Thunderbird performance, BUT if the comparison had been between the MP2000 and Xeon 2GHZ the results would have been much closer with Athlon winning several more benches. THAT's why I think they mentioned the XP2000 held it's own against the Xeon.

Mark-

When all else fails, throw your computer out the window!!!
March 14, 2002 9:25:03 PM

I agree with what you're saying, but I think a comparison between Intel's best and AMD's best is fair.
I don't know if you think otherwise or not.

<font color=orange>Quarter</font color=orange> <font color=blue>Pounder</font color=blue> <font color=orange>Inside</font color=orange>
Don't step in the sarcasm!
March 14, 2002 9:46:42 PM

That's not always true either.

Corporations are not always out to spend the maximum possible on a system. Usually they will run a little behind that curve on the 'sweet spot' of what can be done.

Currently the costs of workstations for corporate is ridiculous. Moving from the dual PIII to Dual PIV architecture in most cases is giving a $2k delta on price (bearing in mind we load up on RAM).

You need to factor in a lot of things where cost of cycles are involved and I an tell you we are virtually boycotting compaq and dell dual PIV workstations due to the enourmous hike in price from the PIII. This isn't because of just Intel prices, but because there are few competitors so they charge whatever they like. IBM fall at the same price roughly also.

If someone significant made dual athlon workstations right now, we'd have some in on eval. Our applications are all propriatary financial apps and raw processing power is the issue, not image manipulation. I expect the AMDs would hold up very well to scrutiny. Alas no-one does, so we can't.

There are a lot of people out there who would like/need dual performance, freelance graphics designers, students etc. who can use, would like dual processor rig, but are not going to be able to pay >$2k for 2 processors and a motherboard. The knowledge that they can pay $200 for the board and $660 for 2 MP2000s (sorry, did I just say you can buy <b>2</b> MP2000s for less than the price of <b>1</b> Xeon 2.2A? :smile:

Actually, since the MP2000 has only just been released there are no kits yet, but you can buy a dual board and 2 MP1900s for $760. That means you can actually buy a dual MP1900, board and 1GB registered DDR for a little over 1/2 the cost of dual Xeon2.2A and board <b>without</b> memory!

So I can probably approach building 2 workstations for the price of 1 under the AMD banner.

-* <font color=red> !! S O L D !! </font color=red> *-
To the gentleman in the pink Tutu
March 14, 2002 10:07:27 PM

Oh no disagreement there at all Burgermeister. But that 10% difference in top end also holds true. Intel has built a pretty powerful duallie in the Xeon Northwood indeed. But, for the price of one Xeon dual system you can get 2 Athlon Dual systems or pretty close to it! OR if you go with a lower end Xeon to better match the price of the Athlon, you end up with worse performance compared to the Athlon. That's the whole price performance advantage AMD currently holds.

Intel, otoh, has a more solid reputation in the dual market. They have the tier 1 OEMs and they hold that hand quite tightly still. So, AMD is still onthe periphery here, selling only a few thousand or tens of thousands of MP's a month I think. Perhaps they sell more, but I haven't seen anything to indicate they are gaining significant marketshare in the dual system market yet.

Mark-

When all else fails, throw your computer out the window!!!
March 15, 2002 2:14:44 AM

Not that we'll ever see one of <A HREF="http://www.digit-life.com/articles/ibmpower4/index.html" target="_new">these</A> on the desktop, but they sure look neat. 32mb of L3 cache? Yum. No matter how they perform, they are an impressive piece of engineering.

- JW
March 15, 2002 3:57:25 PM

Quote:
For one thing, they were comparing an Athlon MP2000 against a XEON 2.2ghz. Yes, I realize that AMD bases their XP PR rating relative to the Thunderbird performance, BUT if the comparison had been between the MP2000 and Xeon 2GHZ the results would have been much closer with Athlon winning several more benches. THAT's why I think they mentioned the XP2000 held it's own against the Xeon.

Had they tested it against a 2GHz Xeon, I would totally agree with you. They didn't. So I still say that their conclusion was baseless.

<pre>If you let others think for you, you're the
only one to blame when things go wrong.</pre><p>
March 15, 2002 4:05:16 PM

peteb, I haven't run into what you are describing. From what I've seen, either the customer wants the absolute best of the best and is more than willing to pay even up to five grand, or they want the absolute cheapest PoS that we offer. (I believe this would currently be a P3-800 on a 440BX motherboard with a whopping 256MB of 100MHz SDRAM.)

Of course, my experieces with customers could be biased that the customers in my own company either need as much power as possible, or just run MS Office apps. And my experiences with customers outside of my own company are purchasing these systems with analytical x-ray machines that can run over half a million with all of the bells and whistles. So I readily admit that this could throw my experience with customers into an 'all or nothing' state of mind.

So far though, I have yet to see a customer want something only slightly less than the best.

<pre>If you let others think for you, you're the
only one to blame when things go wrong.</pre><p>
March 15, 2002 4:20:07 PM

They almost look like a chip design that I came up with five years ago while playing Shadowrun as a decker while running the matrix.

Only my design had:
One Master Processing Unit which had a moderate cache for commands, a small cache for response pointers, and 12 ALUs for logic determination of handling the Sub Processing Units, and the appropriate registers and other units for handling the SPUs.

And four Sub Processing Units which each had individual memory controllers and seperate memory caches, registers, two ALUs, and one FPU.

The idea was that the MPU would look at thread priority and number of individual threads and distribute the use of the SPUs accordingly. Each SPU would determine the appropriate part of the calculations, and then the MPU would just grab the results from the SPU's registers, or even use the SPUs for putting together results from other SPUs.

And each SPU could be reset by the MPU if they ever locked-up. So the whole CPU would only crash if the MPU were to lock up.

So it is nice to see that someone is putting similar ideas into CPUs. Too bad I'd never be able to afford a cluster for home use.

<pre>If you let others think for you, you're the
only one to blame when things go wrong.</pre><p>
!