-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
conroe 64 bit benchmarks
- Thread starter Roughneck
- Start date
You can use whatever version of Vista you like. 32-bit Vista Beta 2 runs fine on 64-bit processors and this would only change for the release version if Microsoft deliberately stopped the 32-bit version running on a 64-bit processor, which I can't see any reason for them to do.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
vista, unlike win xp 64, will be completely properly supported as its not a stop gap short time nightmare of a OS to program for, it will be the new long term OS that everyone will have to support(its a nightmare to start supporting a new os for it to dissappear so soon, just not worth the cost or time to do). so vista 64 should by release be highly supported, drivers for anything you need, games should work and most/all apps there will be no reason to not run 64bit.
afaik the idea will be that vista will ship with either 2 dvd's one 64bit or one 32 bit, or more likely one dvd with the product code relating to what version you want. ie probo get two product codes, one that will tell install program to install 32 bit version and vice version. something like that anyway.
afaik the idea will be that vista will ship with either 2 dvd's one 64bit or one 32 bit, or more likely one dvd with the product code relating to what version you want. ie probo get two product codes, one that will tell install program to install 32 bit version and vice version. something like that anyway.
There is some 32bit/64bit comparisions here and suprisingly show that the conroes are slower in 64 bit than 32 bit.
Its bablefish so hard to read
but the graphs are clear.
http://babelfish.altavista.com/babe...olumn/latestparts/2006/07/14/663447-002.html?
Its bablefish so hard to read
but the graphs are clear.http://babelfish.altavista.com/babe...olumn/latestparts/2006/07/14/663447-002.html?
I wouldn't read too much into those benchmarks. Comparing 64-bit to 32-bit is not going to be "fair" for a long time yet. Whole raft of issues to resolve first... drivers, software, OSes...Devious said:There is some 32bit/64bit comparisions here and suprisingly show that the conroes are slower in 64 bit than 32 bit.
Its bablefish so hard to read
http://babelfish.altavista.com/babe...olumn/latestparts/2006/07/14/663447-002.html?
Soldato
- Joined
- 7 Aug 2004
- Posts
- 11,166
This is what iv started thinking about, question I cannot find the answer to:
'Is conroe a 'true' 64bit cpu like Amd64 ?'
or is it just a 32bit cpu with 64bit instruction sets 'bolted on' like it was in the p4 ?. The 'bolting on' of instruction sets is what made P4 so slow in 64bit, while AMD64 was designed for it from the ground up, with conroe being slower in 64bit that 32bit, i fear it will be a 'bolted on' kind of approach if this thread is making any sense
'Is conroe a 'true' 64bit cpu like Amd64 ?'
or is it just a 32bit cpu with 64bit instruction sets 'bolted on' like it was in the p4 ?. The 'bolting on' of instruction sets is what made P4 so slow in 64bit, while AMD64 was designed for it from the ground up, with conroe being slower in 64bit that 32bit, i fear it will be a 'bolted on' kind of approach if this thread is making any sense
Combat squirrel said:This is what iv started thinking about, question I cannot find the answer to:
'Is conroe a 'true' 64bit cpu like Amd64 ?'
or is it just a 32bit cpu with 64bit instruction sets 'bolted on' like it was in the p4 ?. The 'bolting on' of instruction sets is what made P4 so slow in 64bit, while AMD64 was designed for it from the ground up, with conroe being slower in 64bit that 32bit, i fear it will be a 'bolted on' kind of approach if this thread is making any sense
it worrys me 2
Bolt on or not P4's with EM64T are 'True' 64bit processors in every respect, they just dont have very good 64bit performance, in truth P4's dont have that good 32bit perfomance either.
Conroe is a full blood, 64bit ground up design, sure it takes elements from previous generation processors, but all chips are like that, AMD's K8 still has a lot of K7 concepts in it.
Its very important to realise that performance of code is very reliant on the compiler, generally the intel compilers give the best results for intel processors, but the big issue is that conroe is brand new, if code is optimized for AMD64, it may perform badly on Conroe EM64T, on the other hand if the complier is optimized for Conroe EM64T, the chances are the code generated would be good, and high performance.
There have been some woodcrest benchmarks on 64bit, and they looked pretty good, so you cant look at the results of 3 tests and say, oh look Conroe sucks at 64bit. If you looked only at Conroe's 32bit performance using Science mark (a test suite supposedly written by a former AMD worker), or Synthetic tests, you might think it sucks horribly. But in more 'real world' tests it is performing excellently.
64Bits always been like that anyway, I could probably run up dozens of tests on Sun Ultrasparc processors, which show 32bit applications outperforming 64bit ones. Its all down to appropriate use of code. There are so tasks where even 32bit is totally overkill, who needs a 32bit word processor. Its just current processors are so insanely fast (even P4's) that there is no point in programming in 16bit.
Its great that AMD64, Some P4's and Core 2 all support 64bit. Right now there are some tasks which need 64bit, and performance on a 64bit system is made clear. But if you just take a 32bit application, and 'compile' it in 64 bit chances are you'll either see a small performance hit, or in a perfect world, it would give exactly the same performance.
Does it matter if and AMD64 can render a raytraced image using 64bit code in 44 seconds if a Core 2 can render the exact same image, using 32bit code, and get the result in 30 seconds.
Conroe is a full blood, 64bit ground up design, sure it takes elements from previous generation processors, but all chips are like that, AMD's K8 still has a lot of K7 concepts in it.
Its very important to realise that performance of code is very reliant on the compiler, generally the intel compilers give the best results for intel processors, but the big issue is that conroe is brand new, if code is optimized for AMD64, it may perform badly on Conroe EM64T, on the other hand if the complier is optimized for Conroe EM64T, the chances are the code generated would be good, and high performance.
There have been some woodcrest benchmarks on 64bit, and they looked pretty good, so you cant look at the results of 3 tests and say, oh look Conroe sucks at 64bit. If you looked only at Conroe's 32bit performance using Science mark (a test suite supposedly written by a former AMD worker), or Synthetic tests, you might think it sucks horribly. But in more 'real world' tests it is performing excellently.
64Bits always been like that anyway, I could probably run up dozens of tests on Sun Ultrasparc processors, which show 32bit applications outperforming 64bit ones. Its all down to appropriate use of code. There are so tasks where even 32bit is totally overkill, who needs a 32bit word processor. Its just current processors are so insanely fast (even P4's) that there is no point in programming in 16bit.
Its great that AMD64, Some P4's and Core 2 all support 64bit. Right now there are some tasks which need 64bit, and performance on a 64bit system is made clear. But if you just take a 32bit application, and 'compile' it in 64 bit chances are you'll either see a small performance hit, or in a perfect world, it would give exactly the same performance.
Does it matter if and AMD64 can render a raytraced image using 64bit code in 44 seconds if a Core 2 can render the exact same image, using 32bit code, and get the result in 30 seconds.
Corasik said:
Just one point - isn't Intel's EM64T just a direct copy and implementation of AMD64 extensions anyway? If thats the case, then surely it boils down to the way the processor itself uses the extensions rather than the extensions themselves?
Yes, EM64T is based on AMD's 64bit enhancements to 80x86 instruction sets.
As you say its all down to implementation. P4's for instance perform considerably better with code that's optimized for its long pipeline and reducing the chance of branch misprediction.
It's always down to implementation, with all processors. Taking an example that wont irritate any AMD/Intel fans.... If you have a 32bit sun SuperSparc processor, and try to run software thats optimized for Hypersparc then your program will run slower than its best speed. Recompile with SuperSparc optimization and it will be faster. They are both 32bit sparc processors, but they implement it very differently.
Conroe is a lot more like an AMD than a P4 'Netburst', but its still a very very different beast, and there is still a lot of scope for optimization. The biggest thing with Conroe is that its wider, with 4 instruction decoders. Programs need to be optimized for this extra width. Sure some programs will use the extra width by 'good luck' (and intel would probably say carefull chip design), but programs designed knowing the Conroes architecture can make full use of it, and probably create code which makes best use of MicroOp fusion to squeeze in the 'bonus' 5th instruction more often.
As you say its all down to implementation. P4's for instance perform considerably better with code that's optimized for its long pipeline and reducing the chance of branch misprediction.
It's always down to implementation, with all processors. Taking an example that wont irritate any AMD/Intel fans.... If you have a 32bit sun SuperSparc processor, and try to run software thats optimized for Hypersparc then your program will run slower than its best speed. Recompile with SuperSparc optimization and it will be faster. They are both 32bit sparc processors, but they implement it very differently.
Conroe is a lot more like an AMD than a P4 'Netburst', but its still a very very different beast, and there is still a lot of scope for optimization. The biggest thing with Conroe is that its wider, with 4 instruction decoders. Programs need to be optimized for this extra width. Sure some programs will use the extra width by 'good luck' (and intel would probably say carefull chip design), but programs designed knowing the Conroes architecture can make full use of it, and probably create code which makes best use of MicroOp fusion to squeeze in the 'bonus' 5th instruction more often.
Soldato
- Joined
- 7 Aug 2004
- Posts
- 11,166
Fair points, see what your saying, HOWEVER that doesnt remove the worry that Vista 64 is highly likely to be 'written for' AMD64 64bit instruction sets (basically using code + techniques from Xp 64). Im sure conroe will beat it or keep up of course, but I REALLY hope they have put 'native' conroe EMT64 sets in Vista 64,(i.e. optimised for conroe and not just use AMD64) it would be much suckage if they didnt
Microsoft compilers are neutrally optimised for both AMD and Intel. The differences between the K8 and P8 architectures (from a compiler point of view) are really quite small and therefore over the next few years we will see compilers increasingly being optimised towards high IPC/short pipeline CPUs.
Conroe is a native 64-bit part. As was the P4 NetBurst. The reason 64-bit performed sub-par on the P4 was because the instructions were added late into the chip design and never got the full level of pipeline optimisation that they deserved. This has now been done with Conroe so there's no need to worry any more.
Conroe is a native 64-bit part. As was the P4 NetBurst. The reason 64-bit performed sub-par on the P4 was because the instructions were added late into the chip design and never got the full level of pipeline optimisation that they deserved. This has now been done with Conroe so there's no need to worry any more.
Yeah Vista itself wont really care if its running on Intel, or AMD, but the applications being run will, and intel still have a lot of pressure to convince programming houses to use the intel compilers. Loads of games for instance were 'optimized for P4'. Why is this, well intel actually sent programmers to companies like Verant (Everquest) and helped them do the optimization.
Quite a lot of games used to pop up with the Intel P4 logo when you loaded them, and most of those had the P4 optimizations. Conroe on the otherhand could be a different story, it will run nicely with AMD optimized code, and if the programmers do optimize for Conroe it will run even better. If enough people buy Conroe, then the programmers will optimize for it anyway, especially if intel decide to send programmers out to help again.
Quite a lot of games used to pop up with the Intel P4 logo when you loaded them, and most of those had the P4 optimizations. Conroe on the otherhand could be a different story, it will run nicely with AMD optimized code, and if the programmers do optimize for Conroe it will run even better. If enough people buy Conroe, then the programmers will optimize for it anyway, especially if intel decide to send programmers out to help again.