Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

One 2.93GHz vs Two 2.26GHz

M

mism

macrumors member
Original poster
May 25, 2010
83
0
Hello there, long time lurker, first time poster here. It feels a bit cheeky to kick off with a question (especially a long winded one, sorry) but I'm keen to clear a few things up before the, hopefully, imminent Mac Pro refresh. I know that once they are announced I'm going to have a very itchy trigger finger.

I'm about to make a career change and move into 3D visualization / CGI while continuing with a certain amount of 2D artwork for print. I use Cinema 4D, Photoshop, Illustrator, some Final Cut and Motion and I can see After Effects being added to that list at some point. Until now my trusty Mac Book Pro (2.33 Core 2 Duo 3Gb RAM) has served me well…until it comes to rendering or dynamic simulation in Cinema, or rendering in Motion or FInal Cut.

My eye has been on the "Two 2.26GHz Quad-Core Intel Xeon" Mac Pro, I'd bump the RAM to 8Gb to start with. But like many others I've been reluctant to pay what seems like over the odds especially when we want to believe that a new machine is round the corner.

Following (somewhat compulsively) the various threads on the new Mac Pros I've come to realise that while more cores will really help rendering in Cinema it may not be so useful for other tasks. For instance a quick test running a dynamic simulation and watching the Activity Monitor shows Cinema only using about 99% of CPU, compared to almost 200% when rendering, so I must assume the simulation is not multi threaded? Equally I've read that many other apps, like Final Cut or Motion, will use at most 4 cores and often just 2?

Questions:

I believe I am correct in thinking that each core will be seen as 2 cores by the system, so Cinema will have 16 render buckets on the go on an 8 core machine?

Considering the kind of work I've described what are people's opinions re the configurations below? Both are around my budget, I'll be getting a new monitor and Apple Care on top.
I realise that the configurations will change after the refresh (and the decision to wait or not will be made after WWDC) but this could make the argument all the more valid as it could be 6 fast cores vs 12 slower ones. I also understand that speed isn't only about GHz, although with pure rendering its a pretty good measure.


Two 2.26GHz Quad-Core Intel Xeon, 8Gb RAM = £2635 ($3399 on the US store)
36.16GHz assuming 16 logical cores
4.52 GHz when using 2 cores


One 2.93GHz Quad-Core Intel Xeon, 8 Gb RAM = £2472 ($3149 on the US store)
23.44GHz assuming 8 logical cores
5.86 GHz when using 2 cores


Thanks very much for your help.
 
Final Cut is single-core, only compressor is multi-core. After Effects can take advantage of 8 cores though so if you will be using it a lot, then the 8 cores should be worth it. Most apps are getting support for multiple cores sooner than later so when thinking about the future, 8 cores sound great. There is also '08 Mac Pro with 8 cores @3.2GHz for 3299$
 
If you've read the other threads, you probably came across my opinion already.

There was a discussion in the May 18th thread in particular where a number of people who couldn't wait, opted for 2.26x8 core machines which kind of baffles me. These machines really don't have a place in my opinion.

In my view, if your workload demands that much multi-threading, and time is money to the extent that every saved minute on a long render job counts, then buck up and at least get the 8x2.66. The time/money you save should easily justify the added expense and you won't be sacrificing clocks for cores. The 2.66x8 is probably the sweet spot for a multi-threading beast of a workstation.

The 2.26 is going to be and feel slower than most laptops at most tasks. Sure it might complete the big render job a few seconds or even a few minutes quicker than the quad, but ask yourself how important that is vs. all the other things you do.

The quad with hyperthreading is no slouch at multi-tasking. It's rare that an application will be able to stress this CPU to the max.

Another argument often used for the 2.26x8 is the added memory slots. I have two things to say on this... a) you can equip a quad with 3x4GB for only $450 from OWC b) if you really need more than 16GB of RAM, don't cheap out on a workstation that only runs at 2.26GHz... at least pop for the 2.66x8.

Finally, to say a 2.26x8 core is future proof is a bit crazy given what we know. Any day now, a Mac Pro with 3.33x6 is going to be released for the about the same money as the 2.26x8 and it will absolutely kick that machine to the curb. So don't try to convince me a 2.26x8 is future proof.

As far as the more distant future of multi-threading, technologies like OpenCL hold the promise of an order of magnitude or two improvement in multi-threaded task execution, and that relies on a good GPU. It could be that the best computer of the future is a 5GHz dual-core with two or more monster GPU's.

A 2.93x4 is going to be a great machine for a few years and you can get a refurb right now from the Apple store for $2549. The money you save can go into stuff that will have a REAL impact on your performance like an SSD and more memory.
 
VirtualRain's right.

In general, you have to look at your usage carefully. That means not just the applications used (find out the core utilization if you don't already know), but how much time is spent in each as well. If the majority of time is spent rendering, then yes, get the max. core count possible within budget.

But if other applications (i.e. single threaded and core limited multi-threaded - n= 2 or 4), then a fast Quad (or hex core when it actually ships), would make much more sense.

The funds saved can be put into addressing bottlenecks (upgrades) such as RAM, graphics, and drive throughputs.
 
I don't really see that big of a performance jump from 2.26 to 2.66, as we are only talking 400MHZ. I would think a better performance jump going from a 4 core to a 8 core ( Or 6 to 12 core ) . Usually where speed is more important is single threaded applications.

More apps will be written for multi-core processors in the future, making speed not as important as it once was, unless we are talking huge ghz speeds.
 
linuxcooldude said:
I don't really see that big of a performance jump from 2.26 to 2.66, as we are only talking 400MHZ. I would think a better performance jump going from a 4 core to a 8 core ( Or 6 to 12 core ) . Usually where speed is more important is single threaded applications.

More apps will be written for multi-core processors in the future, making speed not as important as it once was, unless we are talking huge ghz speeds.
Click to expand...

But you have to multiply that 400MHz by 8 as there are eight cores running at 2.66GHz then. That's 3.2GHz! 2.66GHz also goes up to 3.06GHz with Turbo while 2.26GHz stays in 2.53GHz, meaning that 2.66GHz will run circles around 2.26GHz in single and limited multi-threaded apps. 2.66GHz would narrow the gap between 2.93GHz quad in apps that cannot utilize +4 cores.
 
linuxcooldude said:
More apps will be written for multi-core processors in the future, making speed not as important as it once was, unless we are talking huge ghz speeds.
Click to expand...
When, and what though?

Not all applications will benefit from multithreaded operation (think word processing or other applications that rely on user input, such as keyboard entry).

As per when, software always trails hardware. And professional applications in particular, aren't typically on annual development cycles either, as they're too large (suites where there's multiple applications developed simultaneously that have to be completed for the release cycle). Most of it is still centered around FSB right now, and to maintain backwards compatibility, multi-threading may be based on a fixed core count (not n, where it can utilize the full core count, no matter the processor). Then there's specific features, such as those introduced with the Nehalem architecture that may not be utilized by much yet.

Since professional software development is slower (say 3yr cycles), there's a good chance that a new system will be required by the time the Nehalem architecture is commonly utilized by software. :eek: :(
 
Hellhammer said:
But you have to multiply that 400MHz by 8 as there are eight cores running at 2.66GHz then. That's 3.2GHz! 2.66GHz also goes up to 3.06GHz with Turbo while 2.26GHz stays in 2.53GHz, meaning that 2.66GHz will run circles around 2.26GHz in single and limited multi-threaded apps. 2.66GHz would narrow the gap between 2.93GHz quad in apps that cannot utilize +4 cores.
Click to expand...

Actually, I totally forgot about turbo mode...lol. So in single threaded apps, turbo mode would make a big difference in 2.26 vs 2.66. I see a lot of people who multiply the speed when they start talking more cores and I'm not sure or convinced if that is really correct. If we were talking about comparing it to a single core processor with no hyper-threading it might make sense.
 
linuxcooldude said:
Actually, I totally forgot about turbo mode...lol. So in single threaded apps, turbo mode would make a big difference in 2.26 vs 2.66. I see a lot of people who multiply the speed when they start talking more cores and I'm not sure or convinced if that is really correct. If we were talking about comparing it to a single core processor with no hyper-threading it might make sense.
Click to expand...

The CPUs are otherwise exactly the same, that's why the frequency is the only thing we can compare. That's the easiest method of comparing them. Sure we should look at some real world benchmarks but finding them takes time. With raw mathematics, 2.66GHz is ~15% faster when Turbo etc is excluded (2.26/2.66=0.8496...) and that's IMO significant for someone who is professional and uses it +8 hours a day. Let alone Turbo and single-threaded performance
 
Hellhammer said:
The CPUs are otherwise exactly the same, that's why the frequency is the only thing we can compare. That's the easiest method of comparing them. Sure we should look at some real world benchmarks but finding them takes time. With raw mathematics, 2.66GHz is ~15% faster when Turbo etc is excluded (2.26/2.66=0.8496...) and that's IMO significant for someone who is professional and uses it +8 hours a day. Let alone Turbo and single-threaded performance
Click to expand...

I could seeing comparing it to processors with different or odd core counts also. I should of kept the concept with what the poster is originally using it for. So 15% over a long period of time would be huge over a years time. I use mine in a different capacity, running multiple apps at the same time, some which are idle at any given time. No real number crunching or rendering type of thing.
 
VirtualRain said:
If you've read the other threads, you probably came across my opinion already.

There was a discussion in the May 18th thread in particular where a number of people who couldn't wait, opted for 2.26x8 core machines which kind of baffles me. These machines really don't have a place in my opinion.

In my view, if your workload demands that much multi-threading, and time is money to the extent that every saved minute on a long render job counts, then buck up and at least get the 8x2.66. The time/money you save should easily justify the added expense and you won't be sacrificing clocks for cores. The 2.66x8 is probably the sweet spot for a multi-threading beast of a workstation.

The 2.26 is going to be and feel slower than most laptops at most tasks. Sure it might complete the big render job a few seconds or even a few minutes quicker than the quad, but ask yourself how important that is vs. all the other things you do.

The quad with hyperthreading is no slouch at multi-tasking. It's rare that an application will be able to stress this CPU to the max.

Another argument often used for the 2.26x8 is the added memory slots. I have two things to say on this... a) you can equip a quad with 3x4GB for only $450 from OWC b) if you really need more than 16GB of RAM, don't cheap out on a workstation that only runs at 2.26GHz... at least pop for the 2.66x8.

Finally, to say a 2.26x8 core is future proof is a bit crazy given what we know. Any day now, a Mac Pro with 3.33x6 is going to be released for the about the same money as the 2.26x8 and it will absolutely kick that machine to the curb. So don't try to convince me a 2.26x8 is future proof.

As far as the more distant future of multi-threading, technologies like OpenCL hold the promise of an order of magnitude or two improvement in multi-threaded task execution, and that relies on a good GPU. It could be that the best computer of the future is a 5GHz dual-core with two or more monster GPU's.

A 2.93x4 is going to be a great machine for a few years and you can get a refurb right now from the Apple store for $2549. The money you save can go into stuff that will have a REAL impact on your performance like an SSD and more memory.
Click to expand...

This is absolute rot quite frankly.

I have a 2.26 Octad and its nothing like you describe. I can run a full 8 core render (with everyone of the cores maxed out in activity monitor) and my Safari still snaps open faster than I can see and runs super quick without any slowdown. It does not feel as slow as a Macbook Pro because I have one. I can quite happily continue to work quite intensively in photoshop while such things chug away in the background. Now I'm not saying that a Quad core with higher clock won't do the same but it absolutely won't process multi-core faster than mine and to suggest this is a slow 'OS' experience is just garbage. Most apps I use run to all the cores - even logic to my surprise uses all eight.

Now that's not to say if you CAN wait for the MP010 then you should - that makes absolute sense - some people have that luxury - they don't have paying jobs that need heavy lifting right now.

No machine is future proof. In 9 months time the MP 10 will be decried as old and people will start baying for an update 'where is the MP '11' they will cry. Please try to keep this rational.


Oh and lets not forget the 2.66 option on the UK Octo is a whopping £1141 EXTRA! Almost 50% of the cost of the Mac Pro - that's insane money for the marginal gains.
 
mism said:
Hello there, long time lurker, first time poster here. It feels a bit cheeky to kick off with a question (especially a long winded one, sorry) but I'm keen to clear a few things up before the, hopefully, imminent Mac Pro refresh. I know that once they are announced I'm going to have a very itchy trigger finger.

I'm about to make a career change and move into 3D visualization / CGI while continuing with a certain amount of 2D artwork for print. I use Cinema 4D, Photoshop, Illustrator, some Final Cut and Motion and I can see After Effects being added to that list at some point. Until now my trusty Mac Book Pro (2.33 Core 2 Duo 3Gb RAM) has served me well…until it comes to rendering or dynamic simulation in Cinema, or rendering in Motion or FInal Cut.

My eye has been on the "Two 2.26GHz Quad-Core Intel Xeon" Mac Pro, I'd bump the RAM to 8Gb to start with. But like many others I've been reluctant to pay what seems like over the odds especially when we want to believe that a new machine is round the corner.

Following (somewhat compulsively) the various threads on the new Mac Pros I've come to realise that while more cores will really help rendering in Cinema it may not be so useful for other tasks. For instance a quick test running a dynamic simulation and watching the Activity Monitor shows Cinema only using about 99% of CPU, compared to almost 200% when rendering, so I must assume the simulation is not multi threaded? Equally I've read that many other apps, like Final Cut or Motion, will use at most 4 cores and often just 2?

Questions:

I believe I am correct in thinking that each core will be seen as 2 cores by the system, so Cinema will have 16 render buckets on the go on an 8 core machine?

Considering the kind of work I've described what are people's opinions re the configurations below? Both are around my budget, I'll be getting a new monitor and Apple Care on top.
I realise that the configurations will change after the refresh (and the decision to wait or not will be made after WWDC) but this could make the argument all the more valid as it could be 6 fast cores vs 12 slower ones. I also understand that speed isn't only about GHz, although with pure rendering its a pretty good measure.


Two 2.26GHz Quad-Core Intel Xeon, 8Gb RAM = £2635 ($3399 on the US store)
36.16GHz assuming 16 logical cores
4.52 GHz when using 2 cores


One 2.93GHz Quad-Core Intel Xeon, 8 Gb RAM = £2472 ($3149 on the US store)
23.44GHz assuming 8 logical cores
5.86 GHz when using 2 cores


Thanks very much for your help.
Click to expand...

Right now

1) After Effects uses all cores when rendering. After Effects likes RAM but depends on your resolution. I can work easily with full HD.
2) FCP uses all the cores but only at 30-40% utilization (nobody knows why). FCS 4.0 is anticipated to be 64-bit and will fully utilise all the cores you have. It'll be like a mid-cycle speed boost. Motion uses the GPU.
3) Most 3D apps will use all cores right now (can't speak for Cinema 4d though)
4) Rest of CS5 suits will barely break any MP into a sweat these days (Pshop etc). I don't know what single core apps people are using that seems to need such high single clock speeds but for me working with similar apps to the ones you list, octo beats quad by a significant margin.

Pretty much ALL visual apps are built around the concept of low res pre-vis workflows. Only final output render needs the power, or put it another way you will never have enough power. You'll be waiting for renders with the 010 - maybe just 15% less waiting!

Quick and crude way of working out your horsepower is to simply multiple the speed by the cores (don't get fooled by the 'virtual cores' only count the real cores). Thus 8*2.26 = 18.08 Ghz. Compare this to the 4*2.93 = 11.76 - Quite a difference!

This won't be quite as simple when the 32nm CPU's arrive but basically if you do the same but then multiply again by .2 (20%) you'll be about right to account for any speculative gains in architecture. Thus a hexacore 3.33 (if you can afford one - likely to be expensive!) will be 3.33*6*02 = 23.976 GHz (relatively speaking) and so yes faster than the current octad.

However, you will see the current octad is very much a multi-core beast compared to the current Quad - this seems to needle some people - it will still be a beast no matter what Apple release in ten days or ten weeks time. Again, that's not to say I wouldn't have preferred to wait but needs must.
 
Thanks to everyone for the replies, a lot of food for thought. It seems there are strong arguments for both approaches.

I'm transitioning from being able to wait to needing to buy. I think I'll follow the advice someone posted in the 'May 18th' thread, see if there is an announcement at WWDC and then make a decision. Seems like a six core with a higher speed may be the sweet spot for me.
If nothing comes of WWDC then I'll keep my eye on the refurbs, they are frustratingly few and far between on the Apple UK store though.

Do you think an SSD as a boot drive be a worthwhile investment for my kind of work?
 
mism - I am one who had to finally purchase a current MP - very impressed with the latest version - I had a MP 1,1 - one of the best Macs I've owned. Do your analysis and then decide and purchase. Invest your time in growing your business - IMO a lot of time gets wasted in obsessing over tech specs.

Also - if you are interested in a new ssd PM me or check this Marketplace thread

https://forums.macrumors.com/threads/924082/
 
Thanks Apple Corps. Until now I've had the luxury of working away on my folio / site using my Mac Book Pro without deadlines breathing down my neck. Now that I'm ready to start pushing for work I really feel that a more powerful machine is a necessity (having just watched a render take 7 hrs).
WWDC seems like a good point at which to make a decision. They release something then great, if not then I bite the bullet knowing its that or wait with nothing else to go on but faith!

I think I'm too new to the forum to access that thread. The SSD was just a thought though, I'll have to see how my budget works out.
 
surferfromuk said:
This won't be quite as simple when the 32nm CPU's arrive but basically if you do the same but then multiply again by .2 (20%) you'll be about right to account for any speculative gains in architecture. Thus a hexacore 3.33 (if you can afford one - likely to be expensive!) will be 3.33*6*02 = 23.976 GHz (relatively speaking) and so yes faster than the current octad.
Click to expand...

The architecture is completely the same. The cores are still Nehalem based, hence no performance gain at all!

So the overall GHz count for the 3.33GHz hex core is 19.98.
 
surferfromuk said:
However, you will see the current octad is very much a multi-core beast compared to the current Quad - this seems to needle some people - it will still be a beast no matter what Apple release in ten days or ten weeks time. Again, that's not to say I wouldn't have preferred to wait but needs must.
Click to expand...

I don't think anyone disagrees about the Octad being a multi-core beast.

mism said:
Thanks to everyone for the replies, a lot of food for thought. It seems there are strong arguments for both approaches.

I'm transitioning from being able to wait to needing to buy. I think I'll follow the advice someone posted in the 'May 18th' thread, see if there is an announcement at WWDC and then make a decision. Seems like a six core with a higher speed may be the sweet spot for me.
If nothing comes of WWDC then I'll keep my eye on the refurbs, they are frustratingly few and far between on the Apple UK store though.

Do you think an SSD as a boot drive be a worthwhile investment for my kind of work?
Click to expand...

Waiting for WWDC if you can, is a wise idea. No matter what you end up with, it will probably blow your socks off! :)

An SSD is probably the best improvement you can make to any computer... definitely try to factor that into your budget... ahead of other things that will have less impact.
 
Transporteur said:
The architecture is completely the same. The cores are still Nehalem based, hence no performance gain at all!

So the overall GHz count for the 3.33GHz hex core is 19.98.
Click to expand...

The 5600s all have 12MB of L3 cache instead of 8MB. Even the quads.

BTW, was there supposed to be 32nm quads in the Xeon 3600 line? It seems there are only hexes, and Intel is keeping the 3500 line for the lower end?
 
J the Ninja said:
The 5600s all have 12MB of L3 cache instead of 8MB. Even the quads.
Click to expand...
There's been a few minor tweaks as well, such as Turbo Boost will give a greater boost, as well as an improved VT functionality IIRC.

J the Ninja said:
BTW, was there supposed to be 32nm quads in the Xeon 3600 line? It seems there are only hexes, and Intel is keeping the 3500 line for the lower end?
Click to expand...
No, no Quads in the 32nm SP parts (just a single hex core for now, with two others to be introduced at later dates).

So the Quads will still be 45nm.
 
Transporteur said:
The architecture is completely the same. The cores are still Nehalem based, hence no performance gain at all!

So the overall GHz count for the 3.33GHz hex core is 19.98.
Click to expand...

Oh, that's interesting data, basically means it's no faster than the 2.26 octad in multi-core mode.

...and from what I've seen on the net that the 6*3.33 is a $1300 chip so there is no way on earth it's going in an entry level MP.
 
surferfromuk said:
Oh, that's interesting data, basically means it's no faster than the 2.26 octad in multi-core mode.

...and from what I've seen on the net that the 6*3.33 is a $1300 chip so there is no way on earth it's going in an entry level MP.
Click to expand...

But most apps are still non-octo-core so Gulftown will crush 2.26GHz octo in those. And the difference in apps which can use 8 cores isn't big either.

Single CPU version (Xeon W3680) is 999$. The multi CPU version is 1663$ but base Mac Pro will remain single CPU so there is no need for the more expensive one. It's still almost 4 times more expensive than the current 2.66GHz found in low-end
 
Hellhammer said:
But most apps are still non-octo-core so Gulftown will crush 2.26GHz octo in those. And the difference in apps which can use 8 cores isn't big either.

Single CPU version (Xeon W3680) is 999$. The multi CPU version is 1663$ but base Mac Pro will remain single CPU so there is no need for the more expensive one. It's still almost 4 times more expensive than the current 2.66GHz found in low-end
Click to expand...

I think Shademaster's table says more than I can, but yes, a 6 core 3.33 would be a good machine. If it makes it to the base level entry spec is, imo, doubtful, given that there is a $1000 worth chips in the 2.26 multi-core and only $300 in the quad right now - that's basically saying that Apple are going to eat $700 per 6 core base machine they sell?
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back