I'd be interested in seeing tests of pro level graphics cards (Quadro & FireGL) in Solidworks and Autocad.
**I'd also be very interested in home-grown benchmarks for these programs (perhaps based on CAD models provided from academia or industry) since we've seen both Nvidia and ATI cheat in past benchmarks for gaming cards.**
Am I the only one that has never heard of the E3 series Xeon chips? Are they a replacement for the 5600 series, or where do they slot in compared to previous gen Xeons?
And yes, Dell RAm is laughably expensive. They wanted $430 to go from 4GB to 8GB in my M4500. I bought 8GB from Crucial instead for $97. They also have crazy prices for SSD's.
It would been nice to see some benchmarks from workstation class applications that make better use of the Quadro.
Don't know if it counts for this model range but the 3400 series of workstations are pretty hefty, the heaviest workstations I've ever had the 'pleasure' of lugging around anywhere. They put me in mind of the weight of a decent sized CRT monitor rather than a workstation. That said the build quality is very high on them and the devices do look like they can take a serious beating.
I was surprised how basic the heatsink is on this compared to an Optiplex 980 which is a mere business class desktop and not a workstation. That Optiplex has a much larger heatsink and uses a larger fan so potentially could run even quieter.
You didn’t seem to give any actual power consumption figures which would be useful especially at idle. I did read that the P/S efficiency is poor but actual measurements are more useful than just quoting the minimum rated efficiency.
4GB of RAM is a bit low for a system running 64-bit Windows 7, especially considering its intended to run memory hungry workstation applications. Running Maya, Houdini, After Effects, Premier, or even Photoshop, I personally wouldn't go with less then 8GB of memory in a new system. Most of these applications could easily eat up ~3GB of RAM running a single instance with a moderate workload. Hope it's not ridiculously expensive for the 8GB upgrade, it's hit or miss with Dell sometimes.
I appreciate you calling Dell out on the insane pricing structure, but I think even greater outrage was warranted. If Dell wants to add $200/machine to cover the cost of software certifications and support, I'm fine with that. That would be honest pricing. HOWEVER, that's not what they do. Instead they charge $345 to upgrade to a 1GB hard disk, which costs $70 at retail. That's just plain outrageous and *deserves* outrage. Don't be afraid to call it as you clearly see it!
Forgive me; I just took a pricing class where the mantra was "your customer is your enemy and your competitors are your friends who you cannot legally talk to." But the reason they do this is because the people who purchase this type of machine are often engineering firms. Revvit files are MASSIVE, so disk space is definitely a priority. But you have 2 types of engineering firms who buy these; large ones who order 25-50 at a time and who will likely be installing a pre-built image anyway, or small firms who are ordering less than 5 machines. The labor to build an image for 5 machines isn't worth the $1500 you'd spend upgrading the disk on those machines directly.
Customers who aren't willing to pay a high price for more disk space likely aren't willing to pay much of anything for more disk space, as they're just going to install their own disks anyway. So gouge the shit out of the customers who are. The biggest lesson I learned from that class is that these feature prices are very carefully calculated, and cost is only a consideration when the optimal price you can charge is less than the cost of the feature.
The page titled "Build, Noise, Heat, and Power Consumption" has almost no mention of actual power consumption.
Just a fleeting mention of 142 watts but it's not clear if you just added up the 80w tdp of the cpu and the watts for the video card to get this value.
Please get a Kill-a-watt type of device and measure power in idle and when doing something cpu intensive, at least.
It would also be nice to get the idle power usage and then replace the power supply with another power supply that you know how efficient it is, to actually determine the actual power efficiency of this 260 something power supply Dell provides.
I have a suspicion that 65% is for maybe 20% load but if the system uses 140 watts when in use, the efficiency should be a bit higher.
I'm sorry Dustin, what idle power numbers? I can only see the 142w which I'm assuming is load you measured but don't see any idle numbers. I would also second the request for a quick replacement using an 80+ efficiency PSU and retesting idle/load values.
Honestly the combination of POS PSU and RAID 0 with no backup for a workstation is downright laughable IMO.
I appreciate the effort that this site is taking to broaden its horizons beyond just gaming/enthusiast systems and appealing more to the IT side of computing.
That being said, this review contains many pointless graphs that should have been edited out, and just because Dustin was curious about how well this workstation runs CoD and SC doesn't mean you should let him run useless tests. It would have been much more beneficial to have him run the workstation benchmarks on all the boutique builds so he would have a reasonable baseline chart to compare how much better the nvidia 2000 can handle its intended applications than a high-end gaming system.
And yes Dell charges too much for memory and drives... I don't believe I've ever built a single virtual server host with factory upgraded RAM/HDD. Just base config them and save about $2k per host by buying your own memory and drives.
im my opinion, the only important benchmarks when testing a workstation are cinebench 11.5 and specviewperf. maybe use pcmark and 3dmark and cinebench 10. why test games?
the most important thing is to see is how a gaming machine with similar specs compares to this workstation, and this article fails at that.
I realize prices change on a daily basis, but maybe you could take a date-stamped snapshot of prices, and then normalize the scores of some of these test by system price. I think that would be really fascinating, especially for someone like me that edits a lot of video on a budget (as much as one can, dealing with video) - instead of raw x264 encoding speeds, give me [(raw x264 speed) / price].
At least, maybe provide this kind of graph once in a review, using whatever metric you think works best divided by price.
Our company has been using workstations for many years in our dellivery's. And we actually use them as gaming devices. The quadro 4000-5000 doesen't do that bad compared to geforce. However with quadro you will get a stable running card that lasts for 5 years, this is also true for workstations. I dont mind paying extra for that stabillity - in the projects we delliver hardware is only about 10% of the cost anyhow.
Last time we tried geforce we bought 35 workstations and 35 geforce cards, turned out that about 25% percent of the geforce cards was failing after just a few weeks. So in the end the small ammount of money we saved by using geforce in the purchase, we ended up loosing big time due to replacement of defective cards.
Hopefully we will see a review of the HP Z800 workstation. And i hope anandtech also compares graphic performance against game engines !! The difference between the old FX series and the new quadro series is BIG !
The best target quadro card would be 4000 since this is good money for your buck compared to performance. It's also still single slot and with low power req's
My own testing on quadro versus geforce shows that the 5000 card is not that bad compared to geforce 580 - And the quadro 6000 card would probably come close.
Would love to see anandtech run all the new quadro cards against geforce, i think there is a myth tha says quadro is bad when it comes to gaming. Yes the quadro drivers are not optimized for games and the old FX series was crap. But dont judge the new quadros based on the old FX series. Same goes for Xeon CPU's !
Sorry to say, but I've deployed several of these T1500/1600 series Precisions and the build quality is nothing like the 3400/3500. The 3000 series and higher are built like tanks with fantastic case design. The 1500/1600 series is more like a cheap Vostro version of a Precision with the outward appearance that is largely a plastic facade.
Not that this is an issue that ultimately would probably decide or break whether to purchase this or not. Just an FYI.
It's nice that Anandtech has decided to test workstations, but i think that the review still needs a lot of work. I work on broadcasting and all the computers we use are naturally workstations. Here's my comments on workstations benchmarking and on what i see important:
Skip the game benchmarks completly, if you're buying a workstation for professional use it's completly irrelevant information
Skip the comments about some parts costing more than what you can buy as components in PC shops. If you're buying workstations you're buying the possibility to have a tested system, that'll work and run when it's critical. Not only CAD people and engineers use work stations, TV-studios, broadcast in general, sound engineers etc. all use workstations.Price is rarely the #1 driving factor. But do keep comments about memory upgrades and other component upgrades costing too much from the manufacturer.
Run tests with the GPU & CPU cores stressed to 100% for 24 hours. If i'm rendering something i don't want the computer to die from heat, i want to know that the product is solidly engineered and will keep running when left on to do work.
If a workstation is configured in raid 0 it's not usable for critical situation. It's absolutely not possible for a workstation to die in tv-studios or sound stages for example. Raid 0 adds too much uncertainty. If you need the speed, you go SSD or use external storage, raid 0 is basically never seen. Raid 1 however, is.
Video encoding, rendering, specviewperf are relatively good tests. But try finding some other tests as well. Many video codecs are single threaded, some can be encoded multi threaded. I want to know how the computer manages doing single threaded and multithreaded tasks on video encoding. I also want to know stuff like how many 100-200mbits/s videostreams i can simultaneously write on the HDD. Weirdly i also want to know how quickly the system boots up, since if something crashes the computer, every second waiting for it to boot up costs a lot of money. How many sound tracks can i write simultaneously on the HDD with ProTools? How badly does the speed degrade when the HDD gets full?
Noise tests are very important as well. If the computer is not sitting in a machine room behind a KVM it'll be sitting close to the user. Noise can get really irritating very quickly, specially if there are many of the computers in the same space. Both load and idle tests should be done on noise. Saying things like "whisper quiet" take a completly new meaning when you put the workstation in a sound mixing enviorment for example, or a control room of any kind. What you take for "whisper quiet" in a home enviorment doesn't really mean a thing when reviewing workstations.
The computer will in some cases be installed in racks in a machine room, or a soundproofed rack closet. What kind of rack mounting brackets does the manufacturer provide? How deep is the computer? Supermicro enclosures for example are often so deep that you need to make sure you have deep enough rack closets. How well does the enclosure circulate air from the front to back (or back to front) when mounted inside a rack closet?
What kind of monitoring and managing capabilities has the manufacturer built in? Will i get e-mail alerts on failing parts? Will the BIOS be able to bootup the computer every morning at 7am for example? I might want to close the computer for the night, but i don't want to walk into a machine room to boot it up the next morning.
How many network cards do i have? Can i use load balancing on the 2 integrates NICs? Having 2 is better than 1, since often some computers can sit on several separated LANs.
When i open up the enclosure, is it designed so that when mounted with sliding rack brackets i can just pop the lid off and throw in a replacement HDD? I don't care if cables are neat or not, as long as i can access the system and replace a broken component if i have to.
Power supplies might or might not be a issue. Heat generation is always an issue.
Reliability is key when it comes to workstations in my opinion, and should be kept in mind when reviewing them.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
29 Comments
Back to Article
Michael REMY - Monday, May 2, 2011 - link
good test but why use old cinebench 10 instead the lastest 11.5 release ?mastard - Monday, May 2, 2011 - link
Why were there no test of workstation type applications? Quadro cards are not optimized or intended for games.JarredWalton - Monday, May 2, 2011 - link
Dustin ran SPECviewperf11; if you have any other specific requests for tests, let him/us know.RandomUsername3245 - Tuesday, May 3, 2011 - link
I'd be interested in seeing tests of pro level graphics cards (Quadro & FireGL) in Solidworks and Autocad.**I'd also be very interested in home-grown benchmarks for these programs (perhaps based on CAD models provided from academia or industry) since we've seen both Nvidia and ATI cheat in past benchmarks for gaming cards.**
Stuka87 - Monday, May 2, 2011 - link
Am I the only one that has never heard of the E3 series Xeon chips? Are they a replacement for the 5600 series, or where do they slot in compared to previous gen Xeons?And yes, Dell RAm is laughably expensive. They wanted $430 to go from 4GB to 8GB in my M4500. I bought 8GB from Crucial instead for $97. They also have crazy prices for SSD's.
It would been nice to see some benchmarks from workstation class applications that make better use of the Quadro.
DigitalFreak - Monday, May 2, 2011 - link
No, they're the replacement for the Xeon 3000 series single socket processors. The 5600 series will be replaced by the E5, but not until Q4 this year.Stuka87 - Monday, May 2, 2011 - link
Oh, and a small typo I believe on page 2 after the specview scores:"Quadro 2000 produces scores at least to three times higher than a GeForce GTX 480 would under this test."
Twirrim - Monday, May 2, 2011 - link
Don't know if it counts for this model range but the 3400 series of workstations are pretty hefty, the heaviest workstations I've ever had the 'pleasure' of lugging around anywhere. They put me in mind of the weight of a decent sized CRT monitor rather than a workstation. That said the build quality is very high on them and the devices do look like they can take a serious beating.smilingcrow - Monday, May 2, 2011 - link
I was surprised how basic the heatsink is on this compared to an Optiplex 980 which is a mere business class desktop and not a workstation. That Optiplex has a much larger heatsink and uses a larger fan so potentially could run even quieter.You didn’t seem to give any actual power consumption figures which would be useful especially at idle. I did read that the P/S efficiency is poor but actual measurements are more useful than just quoting the minimum rated efficiency.
dragonsqrrl - Monday, May 2, 2011 - link
4GB of RAM is a bit low for a system running 64-bit Windows 7, especially considering its intended to run memory hungry workstation applications. Running Maya, Houdini, After Effects, Premier, or even Photoshop, I personally wouldn't go with less then 8GB of memory in a new system. Most of these applications could easily eat up ~3GB of RAM running a single instance with a moderate workload. Hope it's not ridiculously expensive for the 8GB upgrade, it's hit or miss with Dell sometimes.TrackSmart - Monday, May 2, 2011 - link
I appreciate you calling Dell out on the insane pricing structure, but I think even greater outrage was warranted. If Dell wants to add $200/machine to cover the cost of software certifications and support, I'm fine with that. That would be honest pricing. HOWEVER, that's not what they do. Instead they charge $345 to upgrade to a 1GB hard disk, which costs $70 at retail. That's just plain outrageous and *deserves* outrage. Don't be afraid to call it as you clearly see it!Exelius - Tuesday, May 3, 2011 - link
Forgive me; I just took a pricing class where the mantra was "your customer is your enemy and your competitors are your friends who you cannot legally talk to." But the reason they do this is because the people who purchase this type of machine are often engineering firms. Revvit files are MASSIVE, so disk space is definitely a priority. But you have 2 types of engineering firms who buy these; large ones who order 25-50 at a time and who will likely be installing a pre-built image anyway, or small firms who are ordering less than 5 machines. The labor to build an image for 5 machines isn't worth the $1500 you'd spend upgrading the disk on those machines directly.Customers who aren't willing to pay a high price for more disk space likely aren't willing to pay much of anything for more disk space, as they're just going to install their own disks anyway. So gouge the shit out of the customers who are. The biggest lesson I learned from that class is that these feature prices are very carefully calculated, and cost is only a consideration when the optimal price you can charge is less than the cost of the feature.
mariush - Monday, May 2, 2011 - link
The page titled "Build, Noise, Heat, and Power Consumption" has almost no mention of actual power consumption.Just a fleeting mention of 142 watts but it's not clear if you just added up the 80w tdp of the cpu and the watts for the video card to get this value.
Please get a Kill-a-watt type of device and measure power in idle and when doing something cpu intensive, at least.
It would also be nice to get the idle power usage and then replace the power supply with another power supply that you know how efficient it is, to actually determine the actual power efficiency of this 260 something power supply Dell provides.
I have a suspicion that 65% is for maybe 20% load but if the system uses 140 watts when in use, the efficiency should be a bit higher.
Dustin Sklavos - Tuesday, May 3, 2011 - link
Actually, I did use a Kill-a-watt power meter. Those idle and load numbers are exactly how things worked out.7Enigma - Tuesday, May 3, 2011 - link
I'm sorry Dustin, what idle power numbers? I can only see the 142w which I'm assuming is load you measured but don't see any idle numbers. I would also second the request for a quick replacement using an 80+ efficiency PSU and retesting idle/load values.Honestly the combination of POS PSU and RAID 0 with no backup for a workstation is downright laughable IMO.
Dustin Sklavos - Tuesday, May 3, 2011 - link
You're absolutely right. The graphs were in the system, just didn't get posted. I've added them.7Enigma - Tuesday, May 3, 2011 - link
Dell are you kidding me? Of all the places where RAID 0 has it's place a workstation is about the worst place I could possibly think of.meorah - Tuesday, May 3, 2011 - link
I appreciate the effort that this site is taking to broaden its horizons beyond just gaming/enthusiast systems and appealing more to the IT side of computing.That being said, this review contains many pointless graphs that should have been edited out, and just because Dustin was curious about how well this workstation runs CoD and SC doesn't mean you should let him run useless tests. It would have been much more beneficial to have him run the workstation benchmarks on all the boutique builds so he would have a reasonable baseline chart to compare how much better the nvidia 2000 can handle its intended applications than a high-end gaming system.
And yes Dell charges too much for memory and drives... I don't believe I've ever built a single virtual server host with factory upgraded RAM/HDD. Just base config them and save about $2k per host by buying your own memory and drives.
ochentay4 - Tuesday, May 3, 2011 - link
im my opinion, the only important benchmarks when testing a workstation are cinebench 11.5 and specviewperf. maybe use pcmark and 3dmark and cinebench 10. why test games?the most important thing is to see is how a gaming machine with similar specs compares to this workstation, and this article fails at that.
rangerdavid - Tuesday, May 3, 2011 - link
I realize prices change on a daily basis, but maybe you could take a date-stamped snapshot of prices, and then normalize the scores of some of these test by system price. I think that would be really fascinating, especially for someone like me that edits a lot of video on a budget (as much as one can, dealing with video) - instead of raw x264 encoding speeds, give me [(raw x264 speed) / price].At least, maybe provide this kind of graph once in a review, using whatever metric you think works best divided by price.
mrmj2u - Tuesday, May 3, 2011 - link
Two Dell T7500 machines under my desk right now...Dual Quad Core W5590 processord @ 3.33Ghz
10TB HD Space using LSI RAID 9260-81
24GB RAM...
ATI Firepro V5700
Ah but its still my work machine ;)
zdzichu - Wednesday, May 4, 2011 - link
I'm surprised to see ATX form factor. After all, traditional desktop from Dell (Optiplex) sports BTX board for a long time.Tor-ErikL - Wednesday, May 4, 2011 - link
Good to see that Anandtech goes workstation ;)Our company has been using workstations for many years in our dellivery's. And we actually use them as gaming devices. The quadro 4000-5000 doesen't do that bad compared to geforce. However with quadro you will get a stable running card that lasts for 5 years, this is also true for workstations. I dont mind paying extra for that stabillity - in the projects we delliver hardware is only about 10% of the cost anyhow.
Last time we tried geforce we bought 35 workstations and 35 geforce cards, turned out that about 25% percent of the geforce cards was failing after just a few weeks. So in the end the small ammount of money we saved by using geforce in the purchase, we ended up loosing big time due to replacement of defective cards.
Hopefully we will see a review of the HP Z800 workstation. And i hope anandtech also compares graphic performance against game engines !! The difference between the old FX series and the new quadro series is BIG !
The best target quadro card would be 4000 since this is good money for your buck compared to performance. It's also still single slot and with low power req's
My own testing on quadro versus geforce shows that the 5000 card is not that bad compared to geforce 580 - And the quadro 6000 card would probably come close.
Would love to see anandtech run all the new quadro cards against geforce, i think there is a myth tha says quadro is bad when it comes to gaming. Yes the quadro drivers are not optimized for games and the old FX series was crap. But dont judge the new quadros based on the old FX series. Same goes for Xeon CPU's !
Tor-ErikL - Wednesday, May 4, 2011 - link
Company's like us doesent care if a graphic card costs 300USD or 900USD, we tend to look at other aspects than price when we buy ;)Fozzie - Thursday, May 5, 2011 - link
Sorry to say, but I've deployed several of these T1500/1600 series Precisions and the build quality is nothing like the 3400/3500. The 3000 series and higher are built like tanks with fantastic case design. The 1500/1600 series is more like a cheap Vostro version of a Precision with the outward appearance that is largely a plastic facade.Not that this is an issue that ultimately would probably decide or break whether to purchase this or not. Just an FYI.
Etern205 - Thursday, May 5, 2011 - link
be nice, if you spend a fortune on a machine to have at least a couple of USB 3.0 connectors in the front!Max- - Thursday, May 5, 2011 - link
It's nice that Anandtech has decided to test workstations, but i think that the review still needs a lot of work. I work on broadcasting and all the computers we use are naturally workstations. Here's my comments on workstations benchmarking and on what i see important:Skip the game benchmarks completly, if you're buying a workstation for professional use it's completly irrelevant information
Skip the comments about some parts costing more than what you can buy as components in PC shops. If you're buying workstations you're buying the possibility to have a tested system, that'll work and run when it's critical. Not only CAD people and engineers use work stations, TV-studios, broadcast in general, sound engineers etc. all use workstations.Price is rarely the #1 driving factor. But do keep comments about memory upgrades and other component upgrades costing too much from the manufacturer.
Run tests with the GPU & CPU cores stressed to 100% for 24 hours. If i'm rendering something i don't want the computer to die from heat, i want to know that the product is solidly engineered and will keep running when left on to do work.
If a workstation is configured in raid 0 it's not usable for critical situation. It's absolutely not possible for a workstation to die in tv-studios or sound stages for example. Raid 0 adds too much uncertainty. If you need the speed, you go SSD or use external storage, raid 0 is basically never seen. Raid 1 however, is.
Video encoding, rendering, specviewperf are relatively good tests. But try finding some other tests as well. Many video codecs are single threaded, some can be encoded multi threaded. I want to know how the computer manages doing single threaded and multithreaded tasks on video encoding. I also want to know stuff like how many 100-200mbits/s videostreams i can simultaneously write on the HDD.
Weirdly i also want to know how quickly the system boots up, since if something crashes the computer, every second waiting for it to boot up costs a lot of money.
How many sound tracks can i write simultaneously on the HDD with ProTools? How badly does the speed degrade when the HDD gets full?
Noise tests are very important as well. If the computer is not sitting in a machine room behind a KVM it'll be sitting close to the user. Noise can get really irritating very quickly, specially if there are many of the computers in the same space. Both load and idle tests should be done on noise. Saying things like "whisper quiet" take a completly new meaning when you put the workstation in a sound mixing enviorment for example, or a control room of any kind. What you take for "whisper quiet" in a home enviorment doesn't really mean a thing when reviewing workstations.
The computer will in some cases be installed in racks in a machine room, or a soundproofed rack closet. What kind of rack mounting brackets does the manufacturer provide? How deep is the computer? Supermicro enclosures for example are often so deep that you need to make sure you have deep enough rack closets. How well does the enclosure circulate air from the front to back (or back to front) when mounted inside a rack closet?
What kind of monitoring and managing capabilities has the manufacturer built in? Will i get e-mail alerts on failing parts? Will the BIOS be able to bootup the computer every morning at 7am for example? I might want to close the computer for the night, but i don't want to walk into a machine room to boot it up the next morning.
How many network cards do i have? Can i use load balancing on the 2 integrates NICs? Having 2 is better than 1, since often some computers can sit on several separated LANs.
When i open up the enclosure, is it designed so that when mounted with sliding rack brackets i can just pop the lid off and throw in a replacement HDD? I don't care if cables are neat or not, as long as i can access the system and replace a broken component if i have to.
Power supplies might or might not be a issue. Heat generation is always an issue.
Reliability is key when it comes to workstations in my opinion, and should be kept in mind when reviewing them.
ochentay4 - Friday, May 6, 2011 - link
you sir made the best comment in this article. i totally agree with every point you made.unnameduser11 - Wednesday, January 26, 2022 - link
why is the graphics card nvidia? i bought mine and it has intel hd graphics