First "real" Nocona vs. Opteron review?

[Markfw]

Originally posted by: rck01
All were 2-way running Windows XP (32-bit) SP1. Tests consisted of a mixture of client/server database, workflow, multimedia playback and media encoding (components of our Clarity Studio suite). Though its fast on linear tasks (OfficeBench is 20-25% faster), Opteron folds as we scale the number of concurrent workloads...

RCK

Not quite sure I read your slang on the "Opteron folds" but I re-read the review, and the nocona only wins one or two benches, and the server benches are TOTALLY dominated by the Opteron. Its not even a competition. Opteron blows it away.

 

[rck01]

Not sure what you're referring to - our review hasn't been published yet, so...

RCK

 

[mamisano]

So, you are again handicaping the Opteron by using a NON-Numa aware, 3 year old OS. In essence, you are disabling one of the main features of the chip...

 

[clarkey01]

I feel like getting an AMD tat on my face

 

[Vee]

Originally posted by: rck01
All were 2-way running Windows XP (32-bit) SP1. Tests consisted of a mixture of client/server database, workflow, multimedia playback and media encoding (components of our Clarity Studio suite). Though its fast on linear tasks (OfficeBench is 20-25% faster), Opteron folds as we scale the number of concurrent workloads...

RCK

(I'm curious to see how exactly that number of concurrent workloads was scaled, and why it is relevant.)

Why are "multimedia playback and media encoding" such heavy components of your Clarity Studio suite? They don't appear to me as being good (or even relevant) choices for showing the things about IT-systems that you pretend to show?

Checking out CSA, I seem to find that you are serving Dell and Intel as an advanced sales organization?
"Helping" business to "upgrade" their hardware?

 

[rck01]

Typical. Your reaction to a poor showing by your CPU of preference is to dismiss the test as being irrelevant. A bit pathological, don't you think?

Unfortunately, in my position I don't have the luxury of becoming emotionally attached to products. My customers - who are primarily in the financial services sector - have zero tolerance for delays. Time is literally money for these people, and my workloads model their runtime environment (which is a huge target market for workstation vendors).

Bottom Line: I'm strongly recommending that my customers avoid Opteron-based workstations for demanding, multi-process, multi-tasking workloads, and I'm echoing these sentiments in my InfoWorld Test Center contributions on the subject.

RCK

 

[justly]

Sounds as if your testing is done on prebuilt systems, are these being compared by price, or with identical configurations/components (its obvious they will not be identical seeing that they are from different OEMs).

Either way, your conclusions that one processor is better than the other can easily be questioned depending on the configuration or benchmarks being used (not to mention any preinstalled apps that the OEM includes in the system).

Since you provide NO specs other than the brand of processor being used (and the OEMs that configured them) what makes your ?opinion? any better than the thousands of other opinions on this forum? If it is that you work for a company that specializes in this then I have to say so what, I have seen many ?professionals? in their respective trade that couldn?t do their job worth a xxxx. As an example, I had a electrician tell my wife that there was no way to make a ceiling fan light work from a wall switch without having the fan also turn on and off from the same switch (I made it work within 5 minutes of getting home from a business trip) and this is a much easier task than trying to decide what computer configuration is best.

This is not to say that you are incompetent (obviously the licensed electrician was), it is just to indicate that from the lack of information you provided there is no compelling reason to believe you over anyone else.

BTW how much does ?multimedia playback and media encoding? come into play for ?My customers - who are primarily in the financial services sector? ?maybe I don?t fully understand what the ?financial services sector? does but it sounds like you are trying to sell a sports car to a farmer that needs a truck.

Enough rambling, I?ll shut up now.

 

[clarkey01]

Nocona does ok, oh well lets give AMD time to revamp Opteron and see what happens.

 

[woodscomp]

Intel is superior to AMD in the long run, becasue Intel produces there own chipsets that work very stable with there processors.

When AMD realizes this and gets back in the business of being a total solution provider then large OEM's might make the jump to offer AMD based systems and servers.

Remember Gateway tried this a few years ago and it flopped. Why? Because the chipsets suck.

AMD systems fair well when they are released but they still lag behind Intel in long term stability and reliability.

 

[Vee]

Originally posted by: rck01
Typical. Your reaction to a poor showing by your CPU of preference is to dismiss the test as being irrelevant. A bit pathological, don't you think?

Oh, no. I don't think so at all! I haven't "dismissed" anything yet. But I'm old enough in this game to know to examine benchmarks very closely. And I know enough of cpu architectures to understand what *can be made*.
"your CPU of preference"? - I do think your reaction was interesting, but do you hold me for a fool?

Unfortunately, in my position I don't have the luxury of becoming emotionally attached to products. My customers - who are primarily in the financial services sector - have zero tolerance for delays. Time is literally money for these people, and my workloads model their runtime environment (which is a huge target market for workstation vendors).

Sounds just like the usual sales pitch to me. :roll:
I have a part time boss who is very impressed when people talk like that.

Bottom Line: I'm strongly recommending that my customers avoid Opteron-based workstations for demanding, multi-process, multi-tasking workloads, and I'm echoing these sentiments in my InfoWorld Test Center contributions on the subject.

RCK

Interesting. I'll be sure to check it out. ...with a jaundiced eye.

 

[DrMrLordX]

Originally posted by: rck01
My customers - who are primarily in the financial services sector - have zero tolerance for delays.
RCK

If they don't tolerate delays, then you can't recommend Nocona-based systems to them!

Intel is having problems with potentially severe bugs in their Lindenhurst chipset.

http://news.com.com/Intel+chip+glitch+hobbles+server+advance/2100-1010_3-5289900.html?tag=nefd.top

So, how can you recommend Nocona-based systems to firms that are intolerant of delays? None of the hardware is, as-of-yet, widely available, and the motherboard chipset has flaws that may not be fully resolved until Q4 of 2004! How long do you think they want to wait for new hardware? They can buy top-of-the-line Opteron systems right now.

Furthermore, why do you want to jump all over people for questioning your benchmark methodology? People tend to question any/all benchmarks unless they produce results indentical to previous, known-good benchmarks. If you produce research that shows unexpected results, you'd better be prepared to back up your numbers rather than jump all over people for daring to ask questions.

I think that mamisano, Vee, and justly have all raised valid inquiries regarding your test setup. I'd like to ask another question of your benchmark: Why do you test 3 Nocona systems and only one Opteron system? That is not representative of the current product availability on the market. There are not three times as many Nocona systems for sale or in use as their are Opteron systems. Opterons are more common in sales channels and in IT departments.

 

[Mik3y]

Originally posted by: rck01
Typical. Your reaction to a poor showing by your CPU of preference is to dismiss the test as being irrelevant. A bit pathological, don't you think?

Unfortunately, in my position I don't have the luxury of becoming emotionally attached to products. My customers - who are primarily in the financial services sector - have zero tolerance for delays. Time is literally money for these people, and my workloads model their runtime environment (which is a huge target market for workstation vendors).

Bottom Line: I'm strongly recommending that my customers avoid Opteron-based workstations for demanding, multi-process, multi-tasking workloads, and I'm echoing these sentiments in my InfoWorld Test Center contributions on the subject.

RCK

wow, thats like the dumbest response ever. y dont you support your recommendations with facts and reasons? the opterons are the superior cpu's. the opteron 250 has been out for a while and intel has yet to make a cpu more powerful then it. the opterons beat the new xeons in many of the benchmarks, so i would call it neck to neck in performance, but which one is cheaper? which one is more readily available? in the poll at gamepc, it states that 90% of the people out there still prefer the opterons. y? people can trust amd; they know amd is cheaper; and performance is neck to neck, if not even better hten hte xeon. there is a reason why computer smart people mostly aim towards amd.

 

[clarkey01]

Very true Mik3y, i was about to say that to fanboy over there but you beat me to it.

RISE

 

[justly]

Originally posted by: woodscomp
Remember Gateway tried this a few years ago and it flopped. Why? Because the chipsets suck.

I sure am glad to know that it had nothing to do with Gateway using inadequate power supplies, overcharging, limiting its availability to Gateway country stores, marketing them as budget/economy systems, poor design, or anything else that Gateway was rumored to have done wrong at that point in their history. :roll:

 

[clarkey01]

Originally posted by: rck01
My customers - who are primarily in the financial services sector - have zero tolerance for delays.
RCK
--------------------------------------------------------------------------------


DrMrLordX
-
"If they don't tolerate delays, then you can't recommend Nocona-based systems to them!"

Yeah, back that up rck1 ? oh wait it was delayed. so err ? and what ?

Opteron : Fast all around Majority lead in benchmarks, FULLY 64-bit compatible and not some rushed reversed engineered copy-cat of a CPU, no FSB bottleneck , readily available, bright future with CLEAR roadmaps indicating duel core, unlike Intel?s where they just choose to say ? Me Too? at the last minute. Not to mention cheaper and scale a hell of a lot better then Xeons ever will.

 

[txxxx]

Originally posted by: rck01
Not sure what you're referring to - our review hasn't been published yet, so...

RCK

......I believe your full of crap to be honest. How will a 30-step x86 pipelined CPU somehow outperform a much smaller 12/14step x86-64 CPU in your so called benchmarks? HT alone cant save your Intel's. If anything, Intel's CPU's will be memory bandwidth strangled as you place more demand on them.

And as for scaling, I believe you really have some issues benchmarking, Opterons scale very well in these environments.

I do feel sorry for organisations that use your so called results, its clear there's Intel bias here.

 

[NewBlackDak]

The bad part is people buy products based on the BIAS "tests" thinking they're doing something good.
Financial services sector. I think of large databases, and something that needs outstanding FP performance. If that is the case, why would you use encoding benchmarks to show which is better? I'm sure the Xeon encodes faster than ultrasparcs. However, if you show a dual xeon box being superior to a dual UltrasparcIII for an Oracle DB because it encodes faster you're a friggin idiot. Same case here. I see no reason why someone who needs a datacenter, or number crunching server needs something that should be encoding DIVX all day.

 

[clarkey01]

"However, if you show a dual xeon box being superior to a dual UltrasparcIII for an Oracle DB because it encodes faster you're a friggin idiot"


that is the case..

 

[mamisano]

Hmmm, still no response to numerous questions.

 

[clarkey01]

exactly

 

[mamisano]

Hmmm, just tried to configure a new Dell Poweredge 1850 with Dual Xeon 3.4s and 8GB of DDR2 - Estimated ship date is 9/9. Talk about a paper launch. If the number 1 retailer of Intel chips can't even ship product, then it is Vaporware until then.

A similarly configured HP Proliant DL145 with Dual Operton 250s runs ~$500 less than the Dell too and can be increased to 12GB of ram, where the Dell is limited to 8GB.

Also, doesn't the Xeon use some sort of translation to address memory above 4GB? I know they added 64bit AMD extensions, but I know that there are some drawbacks to the implementation, especially in memory access.

 

[rck01]

Scarce or not, we're sitting here with the following lineup in our lab:

IBM IntelliStation A Pro (2.2GHz Opteron 248)
MPC NetFrame 600 (3.2GHz Prestonia)
HP xw8200 (3.4GHz Nocona)
Dell Precision 670 (3.6GHz Nocona)
HP xw6200 (3.6GHz Nocona)

All systems tested with 2GB RAM and identical SATA disks (no RAID allowed out of courtesy to the Opteron box, which doesn't have a RAID-capable controller).

Note: The IBM box is their second go-around. Their first box - which was pre-production - performed far worse. Their second attempt - which is a GA box that we let them tweak before submission - closed the gap, but only slightly. Still a full 36% slower across a mixture of tasks vs. the best Nocona (HP xw6200).

None of the above vendors are disputing our results. And yes, the list is in order from slowest to fastest.

RCK

 

[mamisano]

mixture of client/server database, workflow, multimedia playback and media encoding (components of our Clarity Studio suite). Though its fast on linear tasks (OfficeBench is 20-25% faster

So, you offer testing for Financial institutions with the above mentioned tests?

What tests are used for client/server database? You are testing WORKSTATIONS with only 2GB of Ram. Why test Database applications on it?

Multimedia playback and media encoding? Are you kidding me? Are the primary users of these WORKSTATIONS using these tasks?

Now, if you were testing CAD and WORKSTATION based software, then I can see your point.

I also notice that the IBM only comes with a 248 max instead of a 250...odd.

 

[rck01]

Actually, our primary business function is to monitor real-time performance for stock and equities trading workstations over time. We deploy a sophisticated array of metrics tracking agents and sample the runtime environment every 15-60 seconds for several weeks. The data is then aggregated to a central database for analysis through our custom Report Card templates.

Our workloads are modeled after nearly one billion sample records taken from numerous financial services institutions (some with over 30K employees). You'd be surprised what these people run on a daily basis (WinTV anyone?).

The test project is being conducted at the request of the InfoWorld Test Center, a consulting customer of ours and also a publication to which I personally contribute on a regular basis (check the masthead). As for the 248 quip, that's what IBM sent us (though the performance deltas we're seeing would likely not be negated by a minor uptick in Opteron clock frequency).

Also, everyone seems up in arms over the media encoding component. We added that to diversify our workload mixture so that it reflects *more* than just our financial services models - again, per request of the Test Center. Since another leading market for these systems is high-end multimedia content creation (this per the vendors' own sales data), it makes sense to test their handling of this type of task.

Regardless, the final score is derived by aggregating results from each of workloads. No matter how you slice it - individual workload scores or the average across a mixed set of tasks (including Database, Workflow *and* Media Encoding) - the Opteron is slower. Our methodology is sound and our results easily reproducible. Enough said. Now go read the article when it hits in a couple of weeks...

RCK

 

[justly]

I guess I should thank you for supplying more information, even though it is still quite vague.

The thing is you haven?t answered many of the questions that where asked of you.

Another thing is CSA Research apparently only makes upgrade recommendations, correct?

Since the IBM IntelliStation A Pro (2.2GHz Opteron 248) (I just picked this because it was the first one I looked for) comes in 6 configurations and all but one come with Ultra320 SCSI and the only one that uses serial an ATA drive comes with only 1GB of memory, so why are you not testing the system as configured.

By changing the system configuration you are not helping your customers unless it is being done to match prices. Or are you suggesting to your customers that they down grade from SCSI or SATA RAID?

So far your testing methodology doesn?t appear to help your customers make an informed decision based on price, standard configurations, or even a good custom configuration.

Even disregarding these configurations choices there is still the questionable use of media playback and encoding being part of the ?Clarity Studio suite? that is being used to determine an upgrade cycle for a workstation environment and your customers, customers that you claim ?who are primarily in the financial services sector?.

I would also like an answer to two questions.
1. Are these systems being tested for workstation use or server use?
2. How many concurrent apps are expected to be in use at a time? Before you answer the second question please refer back to what you said earlier ?Though its fast on linear tasks (OfficeBench is 20-25% faster), Opteron folds as we scale the number of concurrent workloads...?