Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
AMD Discussion
- Thread starter koyoot
- Start date
- Sort by reaction score
Xeon E5-2699 v4 system gets 4918 on that test - right?
Xeon E5-1650v2 gets 934 - right? So it's a bit faster than a hex-core MP6,1 with a 3½ year old Xeon.
And i7-6800K gets 1120 pts, in the same benchmark.
So its actually better than newest 6C/12T CPU that Intel delivered.
No, it's the same. 1120 and 1136 are the same. Don't spout ******** about "better" when the differences are within the error bars on the tests.
And somehow, you didn't acknowledge that the E5-2699v4 is 430% faster than the Ryzen. More alternative facts?
First, you came here with 22C CPU, and Ivy Bridge-E CPU and intentionally completely left Broadwell and Haswell architectures from comparison, and you accuse me about alternative facts. Because they were not showing reality you wanted it to be(AMD doing worse than Intel)?
Secondly, why do you think that AMD will perform worse than this, when they will release 24 and 32C versions of this CPU? Those are coming Q2/2017.
And lastly, you are threadcrapping.
Wow - are you butt hurt?
I mentioned Ivy Bridge EP (not -E) because that's what Apple is currently selling in the MP6,1. Seems very reasonable to me to point out that your Ryzen screen grab showed performance just a little better than the 3½ year old CPU in the MP6,1.
The E5-2699v4 is a Broadwell chip - I didn't "intentionally completely" leave Broadwell out of the comparison. Check your facts.
Let me know when you have screen grabs of Cinebench 15 running those 24C and 32C Ryzen chips. Until then, vaporware.
Sorry if you think that it's "thread crapping" to interrupt your long-running AMD commercial with a reality check.
So why you used deliberately 22C CPU, that is server, and compare it to desktop, Mainstream/HEDT CPU?
Its like me trying to show how bad intel is, by comparing 32C/64T Naples chip to i7-6900K.
If you do not know what it is: its logical fallacy.
Wake me up, when you will bring proper comparisons, that are in context. Until you will turn to tech enthusiast like I am, you, and Tuxon will both be running around idea, that I am doing commercial for AMD.
If I am doing commercial for AMD, then what you two are doing? Are you payed by Intel or Nvidia to downplay AMD as brand on forums?
Because Apple uses "server" CPUs in the MP6,1 (this is the Mac Pro forum, after all). I compared the top scoring Xeon (which Apple could be using, if they hadn't decided to ignore the MP6,1) to your screenshot. I also compared your screenshot to the closest MP6,1 CPU (which is also a server CPU). But you got all butt hurt.
Why would you compare a non-existent AMD server CPU to a desktop chip?
Wake me when Naples shows up.
So why you will not compare 6C Haswell and Broadwell CPU to the scores of 6C/12T 65W Ryzen chip?
Will it show that Haswell/Broadwell is also not that faster from 6C Ivy Bridge CPU, by your omission?
So maybe I will point out it myself.
Haswell-EP:
Aiden, you are being dishonest in your comparison.
And if anyone asks:
E5-1650v4. Higher clocked CPU barely edges out Ryzen Eng Sample.
But I know, you have to show to everyone how biased shill I am.
Last edited:
I'm not sure if you realise that Mac runs x86-64, which means Apple may use AMD Opteron parts in Mac Pro as a second source with minimal disruption to the software ecosystem.
Neat.
38% faster multi core performance than Mac Pro with 6 Core Xeon W3690.
17% faster multi core performance than new Quad Core i7-7700K (not overclocked).
Approximately 78% single core performance of i7-7700K (no OC).
Last edited:
Of course I am! Thats why I also provide pre-release information about Nvidia upcoming products!
Design and Analysis of an APU for Exascale Computing.
http://www.computermachines.org/joe/publications/pdfs/hpca2017_exascale_apu.pdf
Let me give you an example. AMD is preparing right now 3 designs of an APU. Already there is floating around 35W, mobile APU, with 4C/8T CPU clocked at 3.0/3.3 GHz, and 12 CU. Its not bad for 35W package.
However AMD is also targeting HPC market with the APUs, and they want to "create" a market for this sort of hardware.
There are two other designs.
Compact one, that can scale from 45 to 95W package, with 4C/8T CPU, and 16CU GPU, plus HBM2, on package. However the big daddy: 16C/32T+4096 GCN cores, and 16 GB of HBM2.
This is from technical point. What is the use of those? All of APUs are connecting GPU with CPU through Infinity Fabric, so everywhere where you need as low as possible latency, and instant data delivery, to compute units is important. They will be important also for data analytics. Machine learning, embedded markets. If software will follow this is the place where we will see use for the APUs.
Good times ahead for HPC markets. AMD, Intel and Nvidia have very good product roadmaps here.
http://www.computermachines.org/joe/publications/pdfs/hpca2017_exascale_apu.pdf
Let me give you an example. AMD is preparing right now 3 designs of an APU. Already there is floating around 35W, mobile APU, with 4C/8T CPU clocked at 3.0/3.3 GHz, and 12 CU. Its not bad for 35W package.
However AMD is also targeting HPC market with the APUs, and they want to "create" a market for this sort of hardware.
There are two other designs.
Compact one, that can scale from 45 to 95W package, with 4C/8T CPU, and 16CU GPU, plus HBM2, on package. However the big daddy: 16C/32T+4096 GCN cores, and 16 GB of HBM2.
This is from technical point. What is the use of those? All of APUs are connecting GPU with CPU through Infinity Fabric, so everywhere where you need as low as possible latency, and instant data delivery, to compute units is important. They will be important also for data analytics. Machine learning, embedded markets. If software will follow this is the place where we will see use for the APUs.
Good times ahead for HPC markets. AMD, Intel and Nvidia have very good product roadmaps here.
What relevance does a 20 megawatt computer have for a Mac Pro thread?
Last edited:
I could see Apple using all of these APUs. They have long favored GPUs and small form factors, which is what you can get with an APU. Given AMD's willingness to make custom products, I am sure Apple could mix and match CPUs and GPUs as they see fit.
The biggest issue I see is we don't know how competitive AMD will be in efficiency. Would Apple switch to AMD if it can't use them in their laptops? Probably not.
When it comes to the mac pro, I shudder to think about the cooling required for a ~140 W processor and ~250 W GPU on the same package.
Ahem... that APU made from 16C/32T CPU +4096 GCN core GPU will have 150 and 180W TDP's...
.Part taken from the quoted PDF:
Vega is optimized for higher frequency, so it has backlash, similar to Nvidia Pascal. You can clock GP104 chip at 1000MHz to put it into 100W TDP package. Same thing happens with Vega.
Let me give you a hint: in current, experimental form of silicon, for the 95W package, the APU has 3.2/3.5 GHz clocks, and 1.3 GHz core clock for the GPU.
You may say that is very low. First Ryzen Sample, that has been tested was clocked at 2.8/3.2 GHz, and ended up being 3.6/4.0
. Raven Ridge APUs are slated for 2H 2017, and first ones are Mobile, because that is the biggest markup, and most competitive product AMD can bring, so they focus here. The APUs for desktop will come even in Q4 2017. So we have a lot of time to wait for them.And one last bit. Because of the design of the current, Raven Ridge APUs will clock higher than 4C/8T Ryzen CPUs. Up to 4.0 GHz on CPU and 1.5 GHz on GPU.
P.S. Also expect very high price, compared to what Ryzen chips have. Up to 399$, for highest clocked, best binned NPU
. This is not a typo .It does not have to be 20 megawatt. For example, I could very easily see a 450W trash can MP made from 3 150W APUs, with 48 cores, and triple GPU's.
It depends on the scale you are looking for. Your need may not be satisfied with this, some other person will be perfectly satisfied with 30TFLOPs FP32 of compute power from 450W TDP.
Last edited:
Ahem... that APU made from 16C/32T CPU +4096 GCN core GPU will have 150 and 180W TDP's...
Part taken from the quoted PDF:
Vega is optimized for higher frequency, so it has backlash, similar to Nvidia Pascal. You can clock GP104 chip at 1000MHz to put it into 100W TDP package. Same thing happens with Vega.
Let me give you a hint: in current, experimental form of silicon, for the 95W package, the APU has 3.2/3.5 GHz clocks, and 1.3 GHz core clock for the GPU.
You may say that is very low. First Ryzen Sample, that has been tested was clocked at 2.8/3.2 GHz, and ended up being 3.6/4.0
And one last bit. Because of the design of the current, Raven Ridge APUs will clock higher than 4C/8T Ryzen CPUs. Up to 4.0 GHz on CPU and 1.5 GHz on GPU.
P.S. Also expect very high price, compared to what Ryzen chips have. Up to 399$, for highest clocked, best binned NPU
I mean you can down clock the components on an APU to fit in whatever thermal envelop you want but then you are missing out on quite a bit of performance.
In some cases even shedding 100 MHz from core, can result in 100W lower power draw.
Compare 7970 GHZ edition, with D700. Same GPU's, one clocked at 1000 MHz, second one clocked at 850 MHz, and the difference around 100W lower power consumption of D700
.
Note that the Cinebench numbers have a lot of noise....
Code:
Intel XEON E5-1650 v2 1 6 3.50 1.39 10.45 124 934
Intel XEON E5-1650 v2 1 6 3.50 0.00 10.73 0 959
Intel XEON E5-1650 v2 1 6 3.50 1.49 10.44 133 933
Intel XEON E5-1650 v2 1 6 3.50 1.43 10.47 128 936
Intel XEON E5-1650 v2 1 6 3.50 0.00 10.36 0 926
Intel XEON E5-1650 v2 1 6 3.50 0.00 9.79 0 875
Intel XEON E5-1650 v2 1 6 3.50 1.53 10.62 137 949
Intel XEON E5-1650 v2 1 6 3.46 0.00 10.50 0 939
Intel XEON E5-1650 v2 1 6 3.50 1.54 10.78 138 964
Intel XEON E5-1650 v2 1 6 3.50 1.48 10.22 132 914
Intel XEON E5-1650 v2 1 6 3.50 1.50 10.47 134 936
Intel XEON E5-1650 v2 1 6 3.50 0.00 10.38 0 928
Intel XEON E5-1650 v2 1 6 3.00 1.77 10.64 158 951
Intel XEON E5-1650 v2 1 6 3.50 1.16 10.73 104 959
Intel XEON E5-1650 v2 1 6 3.50 1.44 10.08 129 901
Intel XEON E5-1650 v2 1 6 3.50 1.52 10.49 136 938
Intel XEON E5-1650 v2 1 6 3.50 0.00 10.45 0 934
Intel XEON E5-1650 v2 1 6 3.50 1.50 10.59 134 947
Intel XEON E5-1650 v2 1 6 3.50 0.00 10.26 0 917
Intel XEON E5-1650 v2 1 6 3.50 0.00 10.36 0 926
Intel XEON E5-1650 v2 1 6 3.50 1.33 9.12 119 815
Intel XEON E5-1650 v2 1 6 3.50 1.52 10.59 136 947
Intel XEON E5-1650 v2 1 6 3.50 0.00 10.37 0 927815 to 964 for the MP6,1's hex core - that's a lot of deviation. Best not to make too big of a deal about small differences if the benchmark numbers aren't reproducible.
[doublepost=1487553344][/doublepost]
The 450 watt on the MP6,1 is the rated output from the PS. You'd have to downclock the APUs significantly to fit the 450w total power envelope.
Whether statically downclocked from the factory, or dynamically throttled, doesn't really matter. It will never be able to run at 450w to the APUs without a larger power supply. (But the MP6,1 could easily handle a larger power supply if it were more efficient - more watts with the same heat.)
Last edited:
Just catching up with this discussion, but wondering why is this relevant for Mac Pro or Mac in general. Until there's a replacement for Thunderbolt or Apple stops supporting extensions of any kind, Apple will keep using Intel processors, for good or bad. The APU could be nice, but has Apple made ANY indication that it's interested in that market? I can't find anything, and I'm happy to be corrected, in the current offerings on hardware & software side suggesting that Apple is interested in anyone except mass market home & office users.
Yes, it would be interesting, but how many of those who would appreciate 8 cores would be willing to give up Thunderbolt, which you'd have to do if used a non-Intel processor?