Resolved [Barefeats] Old Mac Pro upgraded to 29% Faster Than New One?

[slughead]

I am not usually in the market for Xeons so forgive me if this is old news, but according to Barefeats, the 5,1 Dual Core Mac Pro can have its Xeons upgraded to be faster than the nMP!

Not only that, but this company called "editbuilder" is actually selling them pre-upgraded.

Dual 6 core Xeon X5690 3.46GHz
128GB PC1333 RAM
1TB Samsung Evo SSD drive (Fastest SATA SSD on the market)
4x4TB HD (550MBps at RAID-0) [ you clearly need to back this up immediately ]
GTX570 -just one card

$9,100 for this ridiculous machine, but it blows the rumored Geekbenches of the nMP away. [Edit: Apparently the application was run in 32 bit, Primatelabs did their own Geekbench using the same processors, to show a better estimation. The "frankenpro" by Editbuilder still dominates. To be clear, Geekbench is estimating using numbers from the exact same processors Apple is using for the nMP.]

The fastest nMP will reportedly use the 12 core Xeon E5-2697 v2 at 2.7Ghz - 3.5ghz Turbo.

Here are the upgraded Mac Pro benchmarks.

On Integer performance, the Upgraded Old Mac Pro is 73% faster than the leaked benchmarks of the old one and 29% faster than what Geekbench estimates the nMP will do--this apparently correlates well to CPU-intensive applications.

On floating point, the nMP barely wins, however, with a score 3% faster than the upgraded old one using only the leaked benchmarks.

I'm not saying this is a good value, those are some ridiculously expensive components. 16GB DIMMS aren't cheap (~$1,200 for 128GB?), the Processors are $1670 each (x2), and that SSD is $600 by itself.

What I'd like to see is them include a 2nd PSU and a couple of GTX780's in there!

 

[slughead]

Okay so apparently the benchmarks Barefeats was comparing it to were done on an "old" version of Geekbench which supposedly isn't as accurate. According to Geekbench, the score should be higher. According to those particular benchmarks, the upgraded "frankenpro" or whatever Editbuilder is selling is only 29% faster. I will update the above post accordingly.

Still interesting though, I think.

 

[crjackson2134]

Wow, that's insane... I'd love to have one of those, but the price is totally out of the range for me as a consumer. There are other companies that will swap out your processor tray with one that has the upgraded processors. It's expensive too but not like this one. Nice find!

 

[pastrychef]

Barefeats is also running 64 bit Geekbench while the leaked benchmark of the New Mac Pro is running 32 bit.

I think it's best to wait until the machines are actually released before making sensationalistic claims such as "Old Mac Pro upgraded to 73% Faster Than New One?"

 

[Gav Mack]

Barefeats is also running 64 bit Geekbench while the leaked benchmark of the New Mac Pro is running 32 bit.

I think it's best to wait until the machines are actually released before making sensationalistic claims such as "Old Mac Pro upgraded to 73% Faster Than New One?"

I agree, they may have had some pre release cans posting their scores on geekbench but that's just computational scores. The internal PCIe storage, perhaps even doubled filling the extra slot running raid 0 will have a huge part to play in usage in real life scenarios.

 

[slughead]

Barefeats is also running 64 bit Geekbench while the leaked benchmark of the New Mac Pro is running 32 bit.

I think it's best to wait until the machines are actually released before making sensationalistic claims such as "Old Mac Pro upgraded to 73% Faster Than New One?"

Geekbench already addressed this, apparently the gap is really more like 29%. I edited the OP accordingly.

You're right though, nothing is for certain until the experiment is done in real life with the final retail product.

Also, this was a question, not a claim. Moreover it was based on data. Flawed data, but data

 

[Gav Mack]

Geekbench already addressed this, apparently the gap is really more like 29%

You're right though, nothing is for certain until the experiment is done in real life.

Also, this was a question, not a claim. Moreover it was based on data. Flawed data, but data

It ain't over till the fat lady sings - or in this case the little black can with the warm air blowhole!

 

[666sheep]

Old news
https://forums.macrumors.com/threads/1663480/

 

[slughead]

Old news
https://forums.macrumors.com/threads/1663480/

I clearly didn't read that thread at the time, my bad.

I was under the impression that the single 12 core E5 Apple is using was better than any dual core setup of the previous generation. I was very surprised to learn that was not the case.

 

[barkmonster]

I agree, they may have had some pre release cans posting their scores on geekbench but that's just computational scores. The internal PCIe storage, perhaps even doubled filling the extra slot running raid 0 will have a huge part to play in usage in real life scenarios.

There's no way a PCIe SSD can make up for raw CPU power. It might increase swapfile efficiency if you've already used up all the internal RAM but it's nothing that can compensate for a slower CPU. There's Xeon-based Dells using the same CPU as the entry level 2013 Mac Pro that barely beat the 2.3Ghz Macbook Pro and adding a fast PCIe SSD won't change that.

Face it. It's Apple's form before function problem yet again and they're waving the fast I/O and for a lot of people, unnecessary GPU power, in the face of people like a shinny object in front of a kitten to distract them from that fact.

 

[Gav Mack]

There's no way a PCIe SSD can make up for raw CPU power. It might increase swapfile efficiency if you've already used up all the internal RAM but it's nothing that can compensate for a slower CPU. There's Xeon-based Dells using the same CPU as the entry level 2013 Mac Pro that barely beat the 2.3Ghz Macbook Pro and adding a fast PCIe SSD won't change that.

Face it. It's Apple's form before function problem yet again and they're waving the fast I/O and for a lot of people, unnecessary GPU power, in the face of people like a shinny object in front of a kitten to distract them from that fact.

I agree but let's wait a few more weeks till it's here first also with apps can justify the use of these GPU's over CPU. I've said that this first gen can is a semi pro product but I'll reserve doing so till all the benchmarks are in first with real time testing.

Until Apple do rethink their strategy with the second gen, at least when it comes out they'll be better support for LGA2011 in hackintosh land cos that's the only alternative left if you want to use OSX with more horsepower than they are currently offering.

 

[barkmonster]

I agree but let's wait a few more weeks till it's here first also with apps can justify the use of these GPU's over CPU. I've said that this first gen can is a semi pro product but I'll reserve doing so till all the benchmarks are in first with real time testing.

Until Apple do rethink their strategy with the second gen, at least when it comes out they'll be better support for LGA2011 in hackintosh land cos that's the only alternative left if you want to use OSX with more horsepower than they are currently offering.

I'm fairly underwealmed by the 2013 Mac Pro and with so many systems floating around a similar performance with quad core CPUs and Turbo Boost/Hyperthreading, it's disappointing that the entry level at it's current price isn't at the very least, the specs of the model above it.

 

[Gav Mack]

I'm fairly underwealmed by the 2013 Mac Pro and with so many systems floating around a similar performance with quad core CPUs and Turbo Boost/Hyperthreading, it's disappointing that the entry level at it's current price isn't at the very least, the specs of the model above it.

Me too a bit but for my case I'm more than happy with my 3,1 for now for what I do. My main gripe is it's not an all in one box but a neat can with a load of spaghetti and other boxes strewn about for extras. My 4/5,1 customers though who don't want to jump ship to windows are not amused. I'm currently on the hunt for used hex cores as a stop gap for two towers until Apple either rethink their strategy with can 2 or do nothing and desert the pro market for higher sales of semi pro gear.

Intel have made a few significant jumps with recent CPU designs, with the xeon one step behind the desktop chips the haswell models have excellent bang per buck and watt as well as far better floating point performance.

 

[thekev]

I agree, they may have had some pre release cans posting their scores on geekbench but that's just computational scores. The internal PCIe storage, perhaps even doubled filling the extra slot running raid 0 will have a huge part to play in usage in real life scenarios.

The PCIe storage just makes high bandwidth read from disk situations more convenient, as you can achieve that bandwidth without extra hardware. In cases where the data is being read from someplace else, it's not a real difference. Given the market it's aimed at, I don't expect many people to use the ssd as an excuse to install less ram. Given the amount supported, pageouts shouldn't be an enormous issue.

I agree but let's wait a few more weeks till it's here first also with apps can justify the use of these GPU's over CPU. I've said that this first gen can is a semi pro product but I'll reserve doing so till all the benchmarks are in first with real time testing.

The issue is that as far as I'm aware, there are no libraries that allow for GPGPU computation to be easily leveraged. I don't think the gpus account for a huge amount of the base price though. I doubt one D300 costs any more than the 5770 when they launched the last model. An additional 5770 was $250 through Apple. The problem is whether it can offer enough value to justify a starting price of $3000 over whatever rig it replaces.

 

[slughead]

The issue is that as far as I'm aware, there are no libraries that allow for GPGPU computation to be easily leveraged. I don't think the gpus account for a huge amount of the base price though. I doubt one D300 costs any more than the 5770 when they launched the last model. An additional 5770 was $250 through Apple. The problem is whether it can offer enough value to justify a starting price of $3000 over whatever rig it replaces.

Schiller actually apparently said that the Dx00 were a great deal at this price, or something to that effect. I don't believe for a second they're paying a lot for them, but I do believe they're charging for them. So they're going to claim, apparently, that the prices are a screaming deal.

I think they're sincere about wanting to offer more GPGPU solutions, but they're also serious about forcing users to pay for crap you don't need. Case-in-point: Try putting a 1TB drive into a Retina Macbook Pro... You can't do it at base config, they force you to buy a $1,800 computer and pay an additional $500 to upgrade from 512GB.

I totally agree on PCIe storage. It's clearly great for laptops, but it's just a more convenient way of having high-bandwidth storage in desktops. The same speed can be quite easily gotten through RAID0 SATA SSD. I'm actually planning on getting a SSD RAID for my current Mac Pro in the next 6 months.

 

[thekev]

Schiller actually apparently said that the Dx00 were a great deal at this price, or something to that effect. I don't believe for a second they're paying a lot for them, but I do believe they're charging for them. So they're going to claim, apparently, that the prices are a screaming deal.

It was a silly comment on his part, because they aren't standard naming conventions. There isn't really a good reason to assume extreme costs below the D700 level. My suspicion was that the base options started in the same cost range, and they just added a second one.

I think they're sincere about wanting to offer more GPGPU solutions, but they're also serious about forcing users to pay for crap you don't need. Case-in-point: Try putting a 1TB drive into a Retina Macbook Pro... You can't do it at base config, they force you to buy a $1,800 computer and pay an additional $500 to upgrade from 512GB.

They have always done that. It makes for simpler configuration matrix. I used to think they wanted to guide people toward standard configurations as much as possible. I figured the GPGPU solution was to be more forward thinking, but the upfront cost is quite high for what they offer. They chop down the card framebuffers, stuck to a quad cpu, no longer bundle keyboard and mouse, and the price still went up by $500. I probably sound negative, but it doesn't mean these won't be a worthwhile purchase for anyone.

 

[slughead]

It was a silly comment on his part, because they aren't standard naming conventions. There isn't really a good reason to assume extreme costs below the D700 level. My suspicion was that the base options started in the same cost range, and they just added a second one.

That was my exact thought. To their credit, The Firepros are quite expensive and would otherwise justify a high pricepoint. The W7000 (D300) are $650 each... granted the D300 have half the RAM of the W7000