Intel Xeon and AMD Opteron Battle Head to Head

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Or me.

I was truly surprised and amazed at what Intel did with Conroe.
I suspected lots of Hype.
But their engineers pulled it off.
The new Xeon is just a product of that.

AMD likely has its work cut out for it.

We will need to get some systems in for load testing for the things we do.
(VMWare Server is a biggy - so we can collapse about 10+ idling servers into 1 physical box.)

A little over a year ago we ceased order Intel's unless a clueless vendor refused to certify anythng but Intel. (Yeah, there were a few dense smaller software providers.)
 
If a dweebie Science/Engineering type were trying to build a very high-end workstation, would these “Server” processors provide any performance benefit over the latest “Workstation/Gaming” ones? Apps are Matlab, Maple, Solidworks, Xilinx ISE, etc.
 
There are many different types of 'servers' -- computation server,
file servers, etc. Since the article seems to gear toward the 'CPU'
power, I assume it's OK to ask for more computation related
benchmarks. The benchmarks did include linpack benchmark.
But obviously got it from Intel as it stated 'This software is Intel
optimized.' I'd love to see one that compiled from source using
a generic compiler (such as gcc) and not a hand-tuned version.
Architecture-wise, I just don't see how Intel's single (or double)
bus based structure can scale well when it comes to quad-core
and beyond. Please show us multi-threaded benchmarks to see
how well it scales up (say, in a dual dual-core system). Thanks
for your effort.

Ummm, if Intel releases Compilled versions of software to take advantage of their architecture then by all means they should be the versions tested.

Because..

A. It's the version that people buying these Intel servers will be running in the real world.

B. Not Intel's fault that AMD are not emphasizing optimised compilled versions for their processors.

Would you pit an IBM Power PC running software compilled in it's own native mode vs a x86 PC running software in emulation mode compilled for a Power PC? No.
 
Well, here is the thing for all the people compairing these benchmarks. These are for the laymen, IE the people not running Exchange Stress tests, or Citrix stress tests. OR sending 1k queries to a SQL DB..

I found the test very interesting, considering that fact that 1) Intel 5160 destroys the Opteron. Destroy meaning, that it is FASTER and Cheaper to run. (power)...

What we see here, is that a CPU, runs cooler, and is faster. For I am sure AMD has some cpu's not released to public, that can run faster, but not neccessarly cooler. Vice versus for run cooler, but prob not faster..

The killer here, is that - The same MB used, (Supermicro).. YOu can swap out the 5160, and go Quad core.. NEXT MONTH,,, that to me is a superior deal..

AND to finish off.... People with the benchmarks; lol.. you kill me.. If you go into ANY data center TODAY, for a company, looked at 10 servers, and just picked 10 random, servers, they are probably using only 10% of their processing power at any given moment.

To effectively benchmark any server, you would need a killer App to test. ANd server based apps, geared for the enterprise, to server >500 users.
1. would utilize a Number of hard drives >30. (unless you bootleg).
2. need a number of subordinate systems, to preform the tests...

Judging how long a divix conversion runs= SAME SYSTEM, same specs, obvious differene!!!
 
I can say for sure that the Socket F Opterons ARE available, I just built a dual Opteron 2214 Server yesterday (and I'm not even with a big time Server builder company.) If I can get ahold of a socket F motherboard from Tyan, populate it with dual Opteron Socket F's.. drop on a couple Cooljag Socket F blowers and 8gigs of ram... any of you can.

Its not in computer stores, No one with a small or medium sized store will carry these yet. You have to ask someone who works with the suppliers and can call them up to order them.

I think Toms hardware is either getting way too lazy or people just are not trusting them with new stuff anymore. I did hear they were very Intel slanted in reviews, juding from how they portrayed things so far doesn't dispel that rumor in my mind.

Just like them reviewing video cards that are already discontinued... Toms Hardware, bringing you Yesterday's Hardware news Today.
 
I can say for sure that the Socket F Opterons ARE available, I just built a dual Opteron 2214 Server yesterday (and I'm not even with a big time Server builder company.) If I can get ahold of a socket F motherboard from Tyan, populate it with dual Opteron Socket F's.. drop on a couple Cooljag Socket F blowers and 8gigs of ram... any of you can.

Its not in computer stores, No one with a small or medium sized store will carry these yet. You have to ask someone who works with the suppliers and can call them up to order them.

I think Toms hardware is either getting way too lazy or people just are not trusting them with new stuff anymore. I did hear they were very Intel slanted in reviews, juding from how they portrayed things so far doesn't dispel that rumor in my mind.

Just like them reviewing video cards that are already discontinued... Toms Hardware, bringing you Yesterday's Hardware news Today.

Hey Kid,

Socket F Opteron wouldn't have changed the outcome. Jesus Christ man.. you guys are like sore losers. See this is why choosing a company over another is bad. It makes people turn into aluminium helmet wearing conspiracy/FUd spreading nutjobs.

Core 2 > K8
Woodcrest > Opteron

Doesn't matter who's reviewing it, EVERY other site have found the same results.

Deal with it Mmmkay!
 
I can say for sure that the Socket F Opterons ARE available, I just built a dual Opteron 2214 Server yesterday (and I'm not even with a big time Server builder company.) If I can get ahold of a socket F motherboard from Tyan, populate it with dual Opteron Socket F's.. drop on a couple Cooljag Socket F blowers and 8gigs of ram... any of you can.

Its not in computer stores, No one with a small or medium sized store will carry these yet. You have to ask someone who works with the suppliers and can call them up to order them.

I think Toms hardware is either getting way too lazy or people just are not trusting them with new stuff anymore. I did hear they were very Intel slanted in reviews, juding from how they portrayed things so far doesn't dispel that rumor in my mind.

Just like them reviewing video cards that are already discontinued... Toms Hardware, bringing you Yesterday's Hardware news Today.

Hey Kid,

Socket F Opteron wouldn't have changed the outcome. Jesus Christ man.. you guys are like sore losers. See this is why choosing a company over another is bad. It makes people turn into aluminium helmet wearing conspiracy/FUd spreading nutjobs.

Core 2 > K8
Woodcrest > Opteron

Doesn't matter who's reviewing it, EVERY other site have found the same results.

Deal with it Mmmkay!


LOL-
 
@ElMoIsEviL: How can you be sure it wouldn't have changed anything?

thraxarious might be a fanboy, or not, but he states a relevant point. Why aren't Toms Hardware comparing the top of the line, which would be the socket F?
Who knows there might have been made a small adjustment in the switch, that makes a diffrence.

See this is why choosing a company over another is bad. It makes people turn into aluminium helmet wearing conspiracy/FUd spreading nutjobs.
But who is spreading the hate? The one who finds an obvius fault, in one of Toms Hardwares tests, or the one who dismiss it by calling the other guy a fanboy?
 
Good article, but its a moog point. New technology beats old technology. Big news there. :roll:

Let's reword this a bit...

Better technology beats already great technology. Newer/older this is a common, illogical argument used when one doesn't like to lose.
Right, however, I don't care that much; these are server chips and Quad Core (still initially expensive server chips) will follow and these are both too far from my PC case. K8 is really great like you say; got terribly hit by C2D but still holding tight.
Talking about technologies; it will be funny next year when Intel releases the single core Core2s as they will probably go head to head at least with the 805 :lol:
 
You need to examine the article a little more closely.

In regards to power, the CPU uses less power, but the memory chips required use far more power that totally off-set the lower power usage of the Chip. So there is no power savings vs AMD. You can't take components in isolation. This is why the author was careful to point this out in text and why I wished he had anotated the graphics since many people are not careful to fully read the article and just glance at the graphics.

If as you say all CPUs are running at 10%, then the fact the chip is faster is meaningless.

However, The fact the CPU is faster is not meaningless to me since we use Virtualization Technology to allow us to place those 10 low use boxes onto a single box.

Other systems are so over loaded they are placed behind layer-4 switches and allow multiple servers to handle requests.

Anyone who has worked in a datacenter knows that servers multiply like bunnies which makes real estate quite valuable. As they multiply they but a huge load on your power systems.

If we were to compare the XEON 5000 vs the XEON 5100 we have a huge win. There is a large power cut. This will allow me to place more servers in the datacenter w/o needing to increase my electrical capacity. Since the architecture is the same, the performance compared to the 5000 will be the same or better depending on where the constraint was located.

When comparing a XEON 5000 System vs an Opty System I will similar power usage. (Slightly Less in a 4 Dimm setup and slightly more in an 8 dimm setup) Also as you wisely noted, many server functions are not CPU bound but BUS bound and push the memory controllers hard. How do these react? Which system gives me more?
 
Good article, but its a moog point. New technology beats old technology. Big news there. :roll:
LameNoobMike, go to your dumb forum for trollz and take your BS with you.

+1

I know lets stop reviewing everything new since its obvious its gonna be better :roll:

Nice to see a comparison between the three (better than comparing it to AMD's new opteron which doesnt even exist), i thought it was a pretty well written article and fair.

Benchmarks for servers need to be updated though. Video encoding isn't a typical server application. Maybe draft up a huge oracle database and compare query times on a set of querys.
Oh yeah, and drop the win2k3 😱 that's shocking stuff 😛
 
I did hear they were very Intel slanted in reviews, juding from how they portrayed things so far doesn't dispel that rumor in my mind.

Retarded statement of the day.

Take a look at the cpu charts. The E6300 isnt even on there yet! It also took them ages to get the E6400 on there. They are the 2 best processors to come out of the core 2 arch. The two that destroy AMD's "value sector" as people are calling it.

Intel slanted, i think not.
 
Hey Kid,

Socket F Opteron wouldn't have changed the outcome. Jesus Christ man.. you guys are like sore losers. See this is why choosing a company over another is bad. It makes people turn into aluminium helmet wearing conspiracy/FUd spreading nutjobs.

Core 2 > K8
Woodcrest > Opteron

Doesn't matter who's reviewing it, EVERY other site have found the same results.

Deal with it Mmmkay!

I'm sorry, I am not any fanboy, as our company deals in Core2duo, Athlon, Opteron, Xeons, etc.. I learned a long time ago that if you chain yourself to any one thing you end up looking and sounding like a complete ass.

In the past, Athlon was one of the best choices for servers. I have customers come to me pricing them and asking to have me build them. In their cases, they have already done the research to place them in racks. they don't just go "LOL XEON KICKS OPTY BUTT!" If Opteron is what their research says to go buy, fine. They're looking at more than just one performance metric.

As we saw in the article the Opteron's memory bandwidth does outperform Xeon. If a company is using something that is VERY memory intensive (ie VMware servers?) the Opteron Socket F's might be the one for them. A lot of the time I am not given the information as I cannot be told what the systems are for. NDAs are really sticky things to violate.

The right tool for the right job. Toms hardware SHOULD be comparing Socket F opterons to the new Xeons as they are in the same release time frame. not something from over a year ago. That is my point. Releasing an article that talks about the old opterons at this stage when they are available is more FUD. I hate FUD from any source, Intel or AMD.
 
Hey Kid,

Socket F Opteron wouldn't have changed the outcome. Jesus Christ man.. you guys are like sore losers. See this is why choosing a company over another is bad. It makes people turn into aluminium helmet wearing conspiracy/FUd spreading nutjobs.

Core 2 > K8
Woodcrest > Opteron

Doesn't matter who's reviewing it, EVERY other site have found the same results.

Deal with it Mmmkay!

I'm sorry, I am not any fanboy, as our company deals in Core2duo, Athlon, Opteron, Xeons, etc.. I learned a long time ago that if you chain yourself to any one thing you end up looking and sounding like a complete ass.

In the past, Athlon was one of the best choices for servers. I have customers come to me pricing them and asking to have me build them. In their cases, they have already done the research to place them in racks. they don't just go "LOL XEON KICKS OPTY BUTT!" If Opteron is what their research says to go buy, fine. They're looking at more than just one performance metric.

As we saw in the article the Opteron's memory bandwidth does outperform Xeon. If a company is using something that is VERY memory intensive (ie VMware servers?) the Opteron Socket F's might be the one for them. A lot of the time I am not given the information as I cannot be told what the systems are for. NDAs are really sticky things to violate.

The right tool for the right job. Toms hardware SHOULD be comparing Socket F opterons to the new Xeons as they are in the same release time frame. not something from over a year ago. That is my point. Releasing an article that talks about the old opterons at this stage when they are available is more FUD. I hate FUD from any source, Intel or AMD.

VMWare performs better on C2D then on an Athlon64 FX-62... it's the same with Woodcrest vs. Opteron.

Memory bandwidth advantage or not. Intel's VT Technology is so far superior to AMD's offering. K8L should change this by introducing a Shared Caching mechanism.
 
I did hear they were very Intel slanted in reviews, juding from how they portrayed things so far doesn't dispel that rumor in my mind.

Retarded statement of the day.

Take a look at the cpu charts. The E6300 isnt even on there yet! It also took them ages to get the E6400 on there. They are the 2 best processors to come out of the core 2 arch. The two that destroy AMD's "value sector" as people are calling it.

Intel slanted, i think not.

Retarded would be saying "Toms authors side with intel to slam AMD any chance they get." I had heard this rumor, though thought it a bit far fetched. Though with a review that is designed to convey how intel Xeons do better than opterons but are using an older generation of them sounds slanted, or sloppy.

I don't know. I check in with Tom's every so often and have been for a few years. In the past it seemed new hardware always would show up there first. new samples that were not even available to system builders would be put into a glimpse at Tom's...

Now It just seems I have my hands on hardware thats out before toms even thinks of reviewing it. Whats the deal? Not enough people? partners no longer letting them see new hardware? General mallaise?
 
Could you link to any test of virtualization technologies?
VT VS Pacifica?
I personally haven't seen a single test of Pacifica, i think i recall something concerning VT (VS SW virtualization), but i don't have a link..
And how is a shared caching going to help with that?
 
Hey Guys,

I wanted to thank you all for the feedback and input...A lot of valid points here, and they won't go overlooked.

As many have mentioned perhaps it would have been nice to have some more server-related benchmarks, although our emphasis was mostly to give you guys a feel of what the CPU's are capable of, not necessarily solely server-related applications. However, it is still a valid point and it has been noted.

Also, a few of you have stated that newer technology is always better. I find that a very baseless arguement, and although it may be true in many instances, its not correct to generalize, especailly not in our case.

For those of you who think we are Intel biased, lets just say stay tuned for AMD's response.

Thanks again and keep the excellent feedback coming!

Cheers,

Sina
 
Yes,

I thought the tests should have been more like 4-way blade servers doing database and vmWare work. I still don't know which system is best for this. The tests where 2-way using desktop applications. (correct me if I am wrong)
 
I'm sorry but this is a sad example of a review. If Tom's wasn't able or didn't care to get a Socket F Opteron, ok - but they should have said so. But the article shows that the authors clearly don't even know that Socket F Opterons were released in the middle of August 2006! Sun, HP, IBM, Dell all offer Socket F Opteron based machines, but Tom's tests the previous generation and doesn't mention that fact or care about it.

Furthermore, the authors obviously have not seen any AMD server CPU roadmaps for a long time. I quote from the article:

"In mid-2007 AMD is expected to answer Intel's new threat with a line of faster Opterons utilizing the new Socket F, as well as support for DDR2 memory. Whether or not this will be enough to regain the performance lead has yet to be seen, but with no major architectural changes, it is expected that Intel will still retain the overall better performance per watt status."

Incorrect in all regards. Socket F and DDR2 have been introduced in August 2006. A new generation of Opterons is expected for 4Q2006 (65nm), and the next generation after that (Barcelona) is expected for mid-2007 with loads of architectural changes (K8L). Doing a little research takes time, yes, but it helps sometimes.

Another nice example of refusing to think is the following bit from the article (page 3):

"Also, the fact that cache uses less power than an integrated memory controller is a plus in a market where power consumption is a significant factor."

What? Could it be that if you integrate the memory controller in the CPU, you don't need one in the northbridge? The power spent in one place is the power saved in another place. I have never heard that a large cache in the CPU would make the memory controller in the northbridge obsolete. Holy cow.
 
People on this forum are stating that the power consumption is misleading because they state that the Intel-based system will consume an add'l 11 W per FB-DIMM. Looking at your power consumption curves, it appears as if you're showing full system power consumption, not just CPU consumption. That would imply that at least for 4 GB systems, the 5160 does have a clear power advantage over the Opty. But I can't tell whether we're being shown a full system-power number (e.g. at the wall), or a CPU-power number (how would you get that number, anyway?). A little clarification might go a long way.

Thanks.
 
Well, I think this isn't really aimed at the people buying servers. It's more of an informative article that's meant for the desktop/workstation crowd. Most of the people buying servers don't read this type of article, and just buy whatever CDW/Insight/etc. sells them from HP/Dell/IBM/etc....

Begging your pardon, but I am a person who buys servers. Part of an intelligent hardware purchasing decision, is determining what hardware fits your given requirements, and finding a suitable system within the budgets you are allocated (which in smaller businesses tends to mean building your own machines/buying custom-built machines from distributors). Not many small businesses can afford to pay 2-5 times the cost of a machine to get the "Basic" service contracts from IBM/HP/Dell. They run a small internal IT department who build/service systems, and run the network.

In addition, anybody worth their salt in IT who does purchasing should know the hardware/technology scene, which means reading technical news sites (inclusive of the reviews posted on) such as Tom's Hardware, TechRepublic, The Register, The Inquirer, Hexus amongst many, many others.

We run a very IT-heavy infrastructure on a limited budget, but we are running between 3 and 6 dual-Opteron servers (244's or 265's), old dual PIII 1GHz servers and 1U P4 servers in each of our 4 sites (supporting 50+ users).

In addition, some of the more technically inclined readers may run their own servers at home (and use them for different purposes, be it file serving/security/application server/testing machine/etc).
 
Sorry to hear that your IT department is not capable of making informed decisions.

I only work for very large organizations, some global in nature.
Yes, there are IT standards.
Those standards are based upon technical analysis.

Our servers only go into "Data Centers" and not behind people's desks.

A server does not go into the "Data Centers" without power and lan analysis.

There are often capacity guidelines so we know we can add 10 of model X or 5 or model Y.

I certainly hope nobodies life is lost when your server room goes down because of power-overloads or the fact that a critical server has coffee spilled upon it by a secretary.
It's OK if you like to do drugs, but please, don't do crack then come post on the forumz :lol: .

Anyway, where did you come up with all of the assumptions? Yes, of course our servers go in data centers! We don't jump every time AMD releases a more efficient CPU than Intel and vice versa, but we're not insane.

In terms of capacity planning, the server engineering team tends to just buy servers with as many CPUs in them as we need for hosting whatever application/db/whatever we are hosting. Or, for less intensive applications, we have servers w/ 8 CPUs and 16gb of RAM that host virtual servers.

There are no secretaries to spill coffee on the servers. The IT building I work in is highly secure (like bullet-proof glass, security officers and keycards to gain entrance to the building secure). But, thanks for playing a game of "let's make shit up"!