Here’s a review of the new Blade 1500 64-bit workstation by Sun Microsystems. Can the UltraSPARC survive in a world with AMD64 systems that are twice as powerful and half as expensive? There is also some commentary on Solaris 8 in the review.
Sun Blade 1500 Review
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Two PCs for the price of two.
If there’s a platform that SUN SPARC users are to migrate to, I would think the first stop would be the G5 first before the AMD64. With UNIX underpinnings of OSX, seems more suited for the SUN crwod.
My university recently bought 15 Blade 1500 for our cad laboratory. I’m using Cadence software for for a project for my course of analog integrated circuit design. This machine really rocks!!!
Unfortunatly never had the luck to try an amd64…
Your suggestion of OSX as a migration path is valid, but how is it different from AMD64 with a selection of Linux or even BSD distros?
G5 with OSX probably is the best platform for image manipulating and will probably become the best platform for 3D graphic and video editing, but what you don’t understand is that sparc+solaris covers a niche of professional in cad/cam/cae. These applications don’t run on osx, but most of them have been ported to linux-x86, so probably it will be easier to switch to linux.
It seems like the Opteron is faster per MHz than the G5, and probably cheaper too, also it has 4 & 8-way capability, does G5 have that? The only reason to go for G5 should be to avoid direct competition from other opteron workstation builders like boxx.
@Krusty
Probably, the G5 doesn’t has AutoCAD, but it has VectorWorks & ArchiCAD…
I think ArchiCAD has more features and is better than AutoCAD..
@O
It seems like the Opteron is faster per MHz than the G5 <– Can u probe this?
The G5 has support for 16 way, but not the PowerMac..
but.. with Xgrid you can buy a few Xserves cluster nodes with infini band, i’ll be cheaper than a 8 way Opteron logic board.
@TLy
You can run linux, bsd distros in G5 & PowerMacs..
“Can the UltraSPARC survive in a world with AMD64 systems that are twice as powerful and half as expensive?”
I think some people, don’t understand that Mhz and price isn’t always as important as performance, usability, compatibility and workability like all SUN, SGI & Apple machines have…
A BOXX machine could be cheaper (for the suns & sgi, not Apple G5’s) but probably not equal in efficiency.
I’m spanish, sorry for my english
Throughout the article the author remarks that the Blade 1500 is slower than comparable Opterons or IA32 processors, but he also states that he couldn’t get most of the benchmarking programs to run. So, how does he KNOW that its slower?
As an aside, as an undergrad I worked with PIII 866 Linux boxes, as well as ultraSparc 10s. They both had a gig of RAM if I remember correctly, and the PIII ran KDE much faster than the ultraSPARCs. But, for an AI class I was required to write a genetic algorithm to solve a problem, and the Sun machine blew the PIII out of the water, even though the Ultra was a couple years older than the PIII. Take it for whatever you think it means.
Just my 2 cents.
Your suggestion of OSX as a migration path is valid, but how is it different from AMD64 with a selection of Linux or even BSD distros?
Tier 1 vendor support. There is no Tier 1 vendor offering Opteron workstations.
We are a Sun shop, mostly running Blade 150s at the moment, which are not very nice at all. We’re planning on getting a few 1500s, so this was a very interesting review.
I think what strikes me is that Suns are not “real Suns” any more, unless you spend a LOT. The Sun Blade 2000 has all the exotic stuff, like a blazing fibrechannel disk, but the 1500 is basically still a mid-range PC with a 64-bit SPARC chip and decent graphics. Mind you, you have to pay extra for the upgraded graphics too.
It does seem that Sun is seriously threatened at the low end, ie small servers and workstations. Now that 64-bit systems are available in the consumer space, running a variety of operating systems (including Solaris for AMD64 soon), I just don’t see how they can compete with these woefully expensive systems. You’re just not getting anything special any more, now that PC technology has come so far.
Sun is awesome at the high end, and Solaris is perhaps the best big-iron OS, but at the low end, it really looks like game over. Dual-Opteron workstation anyone?
I use UltraSPARC, Opterons, and Athlons. The problem with sun is that although they sucessfully formed SPARC International, Inc. to promote the SPARC they failed in the semiconductor industry.
I compare the SPARC to the Opteron to the Athlon servers. I rank them like this: 1) SPARC 2) Opteron 3) Athlon (poor Server CPU)
It really depends on what kind of OS you are using as well.
Solaris is lightning fast on the SPARC but its slow on the x86 (they claim they’ve improved it for solaris 10)
The SPARC is good for the enterprise. It’s no PC although it would be wise for sun tried to go into that market and failed.
The UltraSPARC has huge potential. They successfully sell seperate boards and CPUs (they also let you make your own boards) to wholesale customers. WHOLESALE. Sun keeps thinking volume. They have a CEO (I think) that does nothing but volume systems. They need to start offering CPUs for the retail. They need to start letting individual OEMs build their own machines and let other companies use the SPARC that aren’t big and huge.
Secondly, they have an expensive relationship with TI. I personally do not feel their partnership with TI makes the SPARC any cheaper. They perfer “friendship” rather than cutting costs.
Here’s my results.
Sun Problems:
1) Expensive to make the UltraSPARC due to their own decisions.
2) Unable to to get the ultraSPARC for retail use (not being able to sell 5,000 of them at a time)
3) Unable to get out the SPARC Boards for retail use (not being able to sell 5,000 of them at a time)
I’ve talked to their CEO before (I’m a shareholder) he seems to be taking all comments very seriously. He’s really wanting to turn the company around. Athlough they are not in any immediate trouble like everyone claims.
I personally feel that due to sun’s failure to get the SPARC in the mainstream and not in their own systems only they have no choice but to support the Opteron or the Itaninum will sneak up on their back and kill ’em. Although I think the Itanium sucks personally.
The UltraSPARC vs. x86 is not valid, just like how Intel Ghz = AMD ghz can’t be valid. It’s totally different environment.
Besides, some of those Sun machines sell for around $2,000. It’s remarkebly how cheaper sun machines are getting. I think we will see more in the future and great things from sun. I think the copmany is being turned around. Create an e-mail server natively on UltraSPARC Solaris (or FreeBSD or Linux SPARC) and then one for Windows on x86 (or *(BSD or Linux x86) and host 5,000 e-mail addresses that use high volume of system resources and tell me what kind of machine is better then. Sun does great things.
I read the article and I can’t say that I found a particularly stunning comparison between the Blade 1500 and a like-priced x86 solution. The best it says is that “it still falls short of a high-end P4 or dual Xeon solution”. Sun’s dual-Xeon product (V60x) sells for $6,600. It’s hard to compare prices because the total package (video cards, memory, etc.) are so different. Obviously the UltraSPARCs are going to be more expensive because they’re not produced in mass like Intel’s processors.
I don’t think Sun has ever really lit the world on fire with their desktops/workstations. They truly are their “worst” hardware offerings, and they ain’t bad at that. I don’t think that x86 has picked up market share in this area because of increased (more competitive) processor speed; it’s probably more a factor of having a solid “Unix-like” OS to run the important applications on.
As for the low end, I’m sure Sun’s recent venture with AMD will cover that base for them. They’ll probably become the #1 Opteron distributor. If that means UltraSPARC is out on the low end, then that’s just how it goes. As long as the applications port easily it probably won’t be much of a problem. But Sun has some big plans for the UltraSPARC in the future, so I think it’s here to stay.
JMO,
John
The UltraSPARC has huge potential. They successfully sell seperate boards and CPUs (they also let you make your own boards) to wholesale customers. WHOLESALE. Sun keeps thinking volume. They have a CEO (I think) that does nothing but volume systems. They need to start offering CPUs for the retail. They need to start letting individual OEMs build their own machines and let other companies use the SPARC that aren’t big and huge.
Secondly, they have an expensive relationship with TI. I personally do not feel their partnership with TI makes the SPARC any cheaper. They perfer “friendship” rather than cutting costs.
Hence the reason I said a while back for SUN to outsource their chip production to TSMC/UMC in Taiwan and assemble the computers in China.
Regarding OEM vendors, good suggestion, however, there needs to be software and to do that SUN needs to start approaching software companies and say, “how much will it cost to move your software natively to Solaris/JDS” and once they find out, sign a contract and cut a cheque covering the costs of the porting.
As for hardware support, as long as they have a standard base motherboard which with every integrated (excluding graphics card) and PCI-Express, then the only concerns should be supporting USB and Firewire addons, most which conform to some USB standard anyway.
As for the desktop, JDS is a great start, keep improving it and port it to Solaris SPARC and you will have a winner.
In terms of pricing, if they can get the price of a generic low end UltraSparc station minus screen down to around $700 then they will start selling them in larger numbers.
That was back when the UltraSPARC had a memory bandwidth advantage. If your algorithm was memory bandwidth limited, it would not be surprising for it to be faster than the PIII. However, it was starting with the P4 and later Athlon chips that the x86 world really began to crank up memory bandwidth. In a matter of three or four years, they went from 800MB/sec for PC133 memory, to 6.4GB/sec for the latest dual-channel DDR400 setups. Now, its the x86 machines that have the bandwidth advantage.
In regard to this:
I read the article and I can’t say that I found a particularly stunning comparison between the Blade 1500 and a like-priced x86 solution. The best it says is that “it still falls short of a high-end P4 or dual Xeon solution”.
————————————–
I would not buy an UltraSPARC to run x86 applications. I would use it to run SPARC applications. Like I said it’s a totally different ball park with UltraSPARC. I personally think the IBM POWER4/5 beats up the UltraSPARC just like how it beats up the G5. Oddly enough, It appears the SPARC is priced lower than the POWER4. Buy a $2,000 Sun Server and then buy a $2,000 Pentium 4 server and tell me which one is more scalable and powerful. Run the native software and compare. The articles all around don’t do sun justice. Like how ESR said Sun is dead. Most valuable comparisons is from actual sun customers.
“1) Expensive to make the UltraSPARC due to their own decisions.
2) Unable to to get the ultraSPARC for retail use (not being able to sell 5,000 of them at a time)
3) Unable to get out the SPARC Boards for retail use (not being able to sell 5,000 of them at a time)”
Excuse my lack of knowledge in this field, but why are they still using & developing their custom CPUs, at current volumes that must be a big extra cost, and looking at what is already available in volumes; power4, Opteron, Itanium, it doesnt make sense from neither a technical or economical perspective. Or is there some “must have” architecture in the sparcs, apart from their custom instruction set (i assume that they are not compatible with the power series)?
I read the article and I can’t say that I found a particularly stunning comparison between the Blade 1500 and a like-priced x86 solution. The best it says is that “it still falls short of a high-end P4 or dual Xeon solution”.
————————————–
I would not buy an UltraSPARC to run x86 applications. I would use it to run SPARC applications. Like I said it’s a totally different ball park with UltraSPARC. I personally think the IBM POWER4/5 beats up the UltraSPARC just like how it beats up the G5. Oddly enough, It appears the SPARC is priced lower than the POWER4. Buy a $2,000 Sun Server and then buy a $2,000 Pentium 4 server and tell me which one is more scalable and powerful. Run the native software and compare. The articles all around don’t do sun justice. Like how ESR said Sun is dead. Most valuable comparisons is from actual sun customers.
Also, to fully appreciate the full power of UltraSPARC the software has to be tune/optimised for it. For example, if one were to to an encoding comparision it would fairer for both sides to use all available optimisations and in the case of UltraSPARC one would take advantage of VIS.
Also, another thing that is not considered is this. RISC based architectures are not very tolerant of crappy compilers. With the x86 architecture you can get away with a bit of crappiness and the repocusions aren’t that great but like EPIC, RISC will definately see the negative results of crappy optimisations and compilation.
A case in point would be Sybase vs. Oracle. EVERY TPC bench mark that has give SUN a good price/performance has been using Sybase (if SUN had more money, purchasing Sybase would be the ultimate product for SUN to integrate with in the software stack and thus put them on a even footing with Microsoft in terms of a complete product line). Could someone here tell me about Sybase and why people still use Oracle? I really haven’t stuffed around that much with DB’s and just curious why one would use Oracle over Sybase.
You definitely have to use Sun’s compiler to unlock the power of a Sun box. GCC is horrible on most non-x86 platforms. Making use of Solaris 9’s different schedualers and the UltraSparc’s extentions, like VIS, help boost performance. I’ve also seen exceptional linear scalability with using Sun’s compilers, which is great because you can compile your software on a 1500 and it’ll scale out on a E15K just fine.
A lot of people have not had a the pleasure of owning a good Sun workstation. I’ve been using a dual proc Ultra60, with 2GB ram, an Elite3D-m6 card, sunpci card, and solaris 9 for a while now and I would not even consider going to another platform. The reason why, is that everything works that way it should, no compiling of the kernel, no screwing around with drivers, quake2 runs great, gnome is smooth, and the sun compiler works great. I use it for everything, and I like how things work properly. I can develop my software and run it on a bigger box at work, with no problems. The only thing that comes close is a good Mac. Which is what my wife uses and loves.
If you just want a desktop system to do every-day stuff, a Sun workstation is probably overkill. Sun has never been in the business of selling its workstations to the general consumers. They tools used my companies, universities, labs, and IT professionals. It’s what you can run on it and create with it that makes it valuable.
Sybase has worked closely with Sun on compiler optimizations. Oracle is a little more generic about it’s optimizations. People use it a lot because of the name brand and the number of products that integrate with it. Sybase is better, but Oracle has a larger marketing team. Oracle is basicly the M$ of the database world.
Yes, it is a superior processor to the AMD 64 and it is more solid and tested.
@O “It seems like the Opteron is faster per MHz than the G5”
I am not sure about that, as the opteron has a peak floating point performance of 4 gflops/s at 2 Ghz, comparing to a G5 which has 8 gflops/s at 2 Ghz!!! It’s twice the perfornance of the Opteron……
A lot of people have not had a the pleasure of owning a good Sun workstation. I’ve been using a dual proc Ultra60, with 2GB ram, an Elite3D-m6 card, sunpci card, and solaris 9 for a while now and I would not even consider going to another platform. The reason why, is that everything works that way it should, no compiling of the kernel, no screwing around with drivers, quake2 runs great, gnome is smooth, and the sun compiler works great. I use it for everything, and I like how things work properly. I can develop my software and run it on a bigger box at work, with no problems. The only thing that comes close is a good Mac. Which is what my wife uses and loves.
Yes, I’ve had the pleasure, the only downsides are these, crap desktop and insufficient software. Not crappy little applications like “Card Maker” but things like Photoshop, Indesign and so forth. The desktop could be fixed via porting JDS (GNOME etc) to Solaris, however, Id don’t hold my breath.
If you just want a desktop system to do every-day stuff, a Sun workstation is probably overkill. Sun has never been in the business of selling its workstations to the general consumers. They tools used my companies, universities, labs, and IT professionals. It’s what you can run on it and create with it that makes it valuable.
True, having used a Blade 150, the only thing I would suggest, apart from the software suggestions above, is to replace the video chip with a Radeon, apart from that, I was a pretty happy chap.
Sybase has worked closely with Sun on compiler optimizations. Oracle is a little more generic about it’s optimizations. People use it a lot because of the name brand and the number of products that integrate with it. Sybase is better, but Oracle has a larger marketing team. Oracle is basicly the M$ of the database world.
Well, I hope the two do team up and start convincing the IT industry that Oracle isn’t the only “enterprise” database on offer.
I’m not sure his comments are valid slamming Lite-On and Seagate for Sun’s choices for CD-Rom and Harddrive. Personally I have had a much better user experience from Lite-On CD-Roms over Sony CD-Roms. Also, I’ve had better reliability from Seagate drives over Maxtor and WD drives.
I did a lot of research before buying Seagate or Lite-On. There were no reliability problems with modern Seagate drives and Lite-On is an OEM manufacturer for many other labeled CD-Rom drives. Some you may go out and buy a more expensive CD-Rom than the Lite-On but chances are it’s made by Lite=On. You pay a lot more for just because of the name on it.
So basically I’d have to disagree with his opinions there.
> with Xgrid you can buy a few Xserves cluster nodes with infini band, i’ll be cheaper than a 8 way Opteron logic board.
And? distcc works on any architecture (could even use other architectures as compiling nodes than the client’s architecure). There was even an article here on OSNews if I remember correctly.
Probably, the G5 doesn’t has AutoCAD, but it has VectorWorks & ArchiCAD…
I think ArchiCAD has more features and is better than AutoCAD..
Apples and oranges. ArchiCad is well-suited for fast architectural design of medium complexity, AutoCad is mostly for mechanical enigneers and architectural and civil engineering design of any size. While ArchiCad is nice, Autocad is a monster.
I’m spanish, sorry for my english
Your English is excellent! I only wish more people were able to speak/write as well as you do.
@Gabriel Ebner
Yes, i know u can do this on any platform, clustering isn’t new…
But, with Xgrid is very very easy… I think is clustering for dummies.
And.. I don’t know any 1U box with a Dual Opteron 2Ghz inside….
Anyway, a dual Opteron usually costs 1000$+ more than a PowerMac G5…